article
stringlengths 4.26k
146k
| summary
stringlengths 31
3.31k
| section_headings
stringlengths 9
553
| keywords
stringlengths 0
1.24k
| year
stringclasses 13
values | title
stringlengths 20
281
| article_length
int64 1.05k
36.8k
| summary_length
int64 14
689
|
|---|---|---|---|---|---|---|---|
Nontuberculous mycobacteria (NTM) have been reported to be increasing worldwide and its geographic distribution differs by region. The aim of this study was to describe the epidemiology and distribution of NTM in the eastern part of China. Sputum samples were collected from 30 surveillance sites for tuberculosis drug resistance test from May 1,2008 to December 31,2008. Identification was performed using a biochemical test, multiplex PCR and GenoType Mycobacterium CM/AS assay. A total of 1779 smear positive clinical isolates were obtained, of which 60 (3. 37%) were NTM. Five species/complex of NTM were identified; M. intracellulare was the predominated species (68. 33%), followed by M. abscessus-M. immunogenum (13. 33%), Mycobacterium spec. (10. 00%), M. Kansasii (6. 67%) and M. peregrinum-M. alvei-M. septicum (1. 67%). M. intracellulare was the main species of NTM in the eastern part of China and clinical physicians should pay more attention to NTM induced pulmonary disease.
Nontuberculous mycobacteria (NTM) were observed soon after Koch’s discovery of Mycobacterium tuberculosis [1]. However, only until the 1950s NTM were defined as ‘atypical’ or ‘anonymous’ mycobacteria [2]. It is well known that more than 100 species of NTM are ubiquitously distributed in the environment, fresh and salt water, soil and biofilms [3,4, 5]. Pulmonary disease caused by NTM has gained increased attention in the world and several studies indicate that NTM incidence is increasing [6,7, 8]. Kozo Morimoto et al. estimated that the prevalence rate of pulmonary disease caused by NTM was 33–65 per 100,000 [9]. Meanwhile, due to inappropriate treatment and high treatment failure [10,11], the mortality of NTM caused lung disease was high at around 30% [12]. Therefore, efficient detection and regular monitoring of NTM is crucial. However, reporting NTM disease to the government is not compulsory according to the infectious disease control policy in China [13]. Thus, knowledge about the epidemiology and distribution of NTM causing pulmonary disease is limited in China, especially in the countryside. Several studies have shown that different NTM species exhibit varied pathogenicity and have different antibiotic susceptibility patterns [14,15]. Meanwhile, it is often seen that pulmonary disease caused by NTM were misdiagnosed as multi-drug resistant tuberculosis (MDR-TB), especially in the developing countries with a high burden of M. tuberculosis disease [16]. Because in such countries, most pulmonary symptoms resembling mycobacterial disease is presumed as M. tuberculosis, but NTM is often resistant to first-line anti-TB drugs, subsequently treated for multidrug resistant (MDR) disease. Our study will also establish proper assay procedure for improving the diagnostic accuracy for NTM caused lung disease. Conventional identification of NTM at the species level is primarily based on phenotypic characteristics as biochemical tests are not only time-consuming but also error prone [17]. However, the molecular biological method has been applied more commonly, and it facilitates the detection of NTM from clinical samples. Amplification of the 16S rRNA was chosen to provide the positive control when evaluating Mycobacteria by PCR and Rv0577 was a genotypic marker for the M. tuberculosis complex (MtbC) [18]. In our study, we chose both of them to differentiate MtbC from Mycobacteria other than MtbC species (MOTT). Besides that, recently DNA strip assays for the identification of Mycobacteria to the species level have been developed, GenoType Mycobacterium CM/AS assay (Hain Lifescience, Nehren, Germany) is one of them. This assay is based on reverse hybridization of a PCR product to a nitrocellulose strip with immobilized probes for different mycobacterial species and shows high concordance with 16S rRNA and biochemical tests [19]. We performed this assay to differentiate NTM species of the samples from tuberculosis suspicious patients to assist clinical diagnosis.
Sputum samples were consecutively collected from 30 tuberculosis drug resistance surveillance sites in Jiangsu province, China, during May 1,2008 to December 31,2008. Only patients with suspicious tuberculosis symptoms, such as cough for at least 2 weeks and abnormal chest X-ray manifestation, were recruited. All samples were derived from the lungs and at least one of three samples per patient were smear positive by the Ziehl-Neelsen method. Then, two of the three samples were chosen to be inoculated on the Löwenstein-Jensen (LJ) medium for culture. Finally, a total of 1779 clinical isolates were obtained. All samples collected were anonymized. This study was approved by the Institute Ethics Committee of Jiangsu Provincial Center for Disease Control and Prevention. DNA from mycobacterial culture was extracted following procedure. For each sample, one loop of cultures was suspended in 400ul TE buffer, boiled at 95°C for 30 minutes, then followed by ice-bath for 5 minutes and centrifugation at 12000×g for 5 minutes. Finally, 200μl of the DNA supernatant was used for further testing, while the remainder was stored at −20°C. As a preliminary screening, p-nitrobenzoic acid (PNB) and thiophene carboxylic acid hydrazine (TCH) was used for NTM identification at first. Growth on LJ medium containing PNB indicates that the bacilli do not belong to MtbC. In order to distinguish MOTT from MtbC, all of the MtbC isolates identified by PNB were tested by 16S rRNA and Rv0557 again. Finally, we used GenoType Mycobacterium CM/AS assay for further identification to species/complex level. The GenoType Mycobacterium CM/AS assay was performed according to the instructions of the manufacturer.
We collected 1779 positive cultures, from May 1,2008 to December 31,2008. The flow chart of NTM identification was shown in Fig. 1. After screening by PNB and TCH resistant test, 72 samples were classified as NTM and 1707 samples belonged to the MtbC. For those MtbC samples determined by PNB and TCH method, multiplex PCR of 16S rRNA and Rv0557 was carried out for confirmation. Finally, we obtained 106 strains including NTM (n = 72) and MOTT (n = 34) to perform GenoType Mycobacterium CM/AS assay. The CM/AS assay is based on a multiplex PCR targeting species-specific DNA regions combined with a reverse hybridization format (DNA strip). The specific patterns are composed of obligatory and additional facultative stainings that can be visually identified by clear-cut hybridization signals on the membrane strips. After an interpretation, sixty out of 106 strains were identified to species/complex level, forty five MtbC strains and one strain showed non species-specific lines were excluded (Fig. 1). Therefore, the rate of NTM was 3. 37% (60/1779) in Jiangsu province. The band patterns of all NTM determined by GenoType Mycobacterium CM/AS assay are shown in Table 1. Five kinds of species/complex were identified, including M. abscessus-M. immunogenum, M. intracellulare, M. Kansasii, M. peregrinum-M. alvei-M. septicum and Mycobacterium spec. The percentages of each species are shown in Table 2. The most dominant NTM was M. intracellulare which accounted for 68. 33% of the 60 isolates in the study. M. abscessus-M. immunogenum was the next most prevalent species (8 isolates, 13. 33%), followed by Mycobacterium spec. (6 isolates, 10. 00%), M. Kansasii (4 isolates, 6. 67%) and M. peregrinum-M. alvei-M. septicum (1 isolate, 1. 67%). In order to investigate the geographical distribution and frequency of NTM, we plotted NTM distributions in 13 cities of Jiangsu province (Fig. 2). Except for in Changzhou, Taizhou and Zhenjiang, M. intracellulare had an extensive distribution throughout the province and was most frequent in Yancheng, followed by Huai’an and Suzhou (Fig. 2). Besides that, M. abscessus-M. immunogenum was present in five neighboring cities in the southeastern part of the province and M. Kansasii was only found in 3 cities located along the Yangzi River. Only one isolate of M. peregrinum-M. alvei-M. septicum was found in Yangzhou city. The members of Mycobacterium spec. was detected in three different cities, Yancheng, Suzhou and Zhenjiang, located in the eastern part of Jiangsu province.
Worldwide, pulmonary disease caused by NTM is increasing [10,20] and has captured more awareness and interest among the isolates of all species of mycobacteria. However, there is no evidence of direct transmission of NTM between humans. Due to this, NTM is not a notifiable condition in many countries and remains unmonitored by governmental agencies. Our study revealed that the overall proportion of NTM isolates from whole specimens was 3. 37%, slightly lower than the mean rate in Shanghai, China [21] and much lower than reports from Europe [22]. We also elucidated the distributions of NTM species to analyze, for the first time, the geographical character in the eastern part of China. M. intracellulare was the dominant strain and was almost evenly distributed in this area. Meanwhile, M. abscessus-M. immunogenum and M. Kansasii were restricted to several adjacent cities of Jiangsu province. The distributions of NTM species varied by region and may have a profound impact on the prevalence of pulmonary NTM disease. For M. avium complex (MAC), the most common NTM, infections in immunocompetent patients are principally pulmonary [3]. The average treatment failure rates of MAC was as high as 20–40% [11]. Another work in Japan indicated that the overall mortality rate was 28. 0% and the mortality for untreated MAC patients was 10% higher than for treated patients [12]. As a member of the MAC, M. intracellulare was the dominant strain in Jiangsu province in accordance with previous studies. In Korea, Jae Kyung Kim et al. found that M. avium complex was the most common NTM and M. intracellulare accounted for 51. 3% of all specimens [23]. Other research conducted in east Asia showed that M. avium complex bacteria were also the most frequent isolates (13%–81%) and the most common cause of pulmonary NTM disease (43%–81%) [24]. In addition, M. intracellulare was the most frequently isolated strain in South Africa and Australia, [6]. The high frequency of M. intracellulare reported in different studies may be due to extensive distribution in the environment, especially in potable water [25]. When we focused on the geographical distribution of M. intracellulare, we found it almost evenly distributed in the province, although absent in three cities (Zhenjiang, Taizhou and Changzhou). The underlying causes of the absence seen in these cities is not clear, but the three cities located in the southern part of Jiangsu province have a relatively lower prevalence of M. tuberculosis [26]. Considering the samples were from tuberculosis suspicious patients, we presumed that the low epidemic situation of tuberculosis was one factor. M. intracellulare has been reported in association with HIV infection [12] as well as with increasing frequency in the non-AIDS population [27]. Our study for the first time described the current situation of NTM caused lung disease and the species proportions in the eastern part of China, where a lower prevalence of HIV infection exists [28]. In our study, TB suspects are not high-risk populations for HIV infection, such as injecting drug users, and therefore HIV status was not detected for each subject. The second most frequently isolated NTM in our study is M. abscessus-M. immunogenum complex. According to the interpretation chart provided by the CM/AS manufacturer, we couldn’t identify between these two species because they share the same line probe bands. Previous work suggested that M. abscessus was one of the most frequent species of rapid growers and usually concerned with skin, soft tissue and pulmonary infections [29]. As an example, it was the third most frequently isolated NTM species in Taiwan and the second most frequently isolated in South Korea [6]. M. immunogeum has been identified in metalworking fluids and has been shown to be highly correlated to hypersensitivity pneumonitis [30]. In our study, M. abscessus-M. immunogenum was found to be restricted to the southeastern part of Jiangsu province. M. kansasii often produces infiltrates or cavities in the upper lobes of immunocompetent patients [31]. In South America, M. kansasii was the second most isolated NTM after MAC, accounting for 19. 8% of all NTM [6]. But in Jiangsu province it was not the prevalent species according to our study and showed very limited geographic spread. We encountered the occasional presence of other rare species in this study, such as M. peregrinum-M. alvei-M. septicum. This species appeared only in one region and the isolate was too scarce to include in analysis however. Besides that, there were six specimens that failed with the CM/AS assay, only identifying as Mycobacterium spec. This may be due to the limited discriminative ability of this assay. Similar results have been seen in previous studies, where cross reaction among NTM species was supposed as the reason for the discrepant results [32,33]. Considering the rate of MDR TB in Jiangsu province is higher than the epidemiological situation of all of China [26], there was a high possibility of misdiagnosis of NTM in clinic. Usually, NTM is resistant to first-line anti-TB drugs, so misdiagnosis leading to inappropriate treatment can result in poor outcome. Our study could be beneficial for distinguishing NTM from M. tuberculosis and promoting valid clinical diagnoses of NTM. According to the American Thoracic Society (ATS) document on NTM diagnosis [34], clinical symptoms and manifestation for TB suspicion in combination with laboratory identification increases diagnostic accuracy of Mycobacterium caused lung disease. However, several shortcomings of our study should be mentioned. Firstly, according to the criteria for subject inclusion, those subjects with pulmonary symptoms for less than 2 weeks would be ignored and the actual prevalence for NTM caused pulmonary disease would likely be underestimated. In addition, we did not follow-up patients therefore we could not analyze NTM treatment outcomes. In summary, we performed a new reverse hybridization technique to illustrate the NTM species distribution from sputum specimens in the eastern region of China and established a procedure to identify and confirm NTM. Given the clinical challenge, further knowledge of the epidemiology of NTM in Jiangsu province is needed and the varying distribution of NTM species by region might have a profound and lasting impact on prevalence of pulmonary NTM disease. In addition, research efforts should be directed towards areas that will lead to strategies to prevent, predict, and improve treatment of NTM disease. | Nontuberculous mycobacteria (NTM) exist ubiquitously in the environment and cause many kinds of diseases including pulmonary infection. Despite this, NTM does not match compulsory report policy in many countries, such as China. Thus, the epidemiology of NTM is generally unknown. Furthermore, misdiagnosis of nontuberculous mycobacterium disease as multi-drug resistant tuberculosis (MDR-TB) frequently occurs in clinical settings because of similar clinical manifestations. Therefore, elucidating the epidemiology and distribution of NTM species is important and may have a profound and lasting impact on the prevalence of pulmonary NTM disease. In our study, we enrolled smear-positive sputum samples during 2008 from Jiangsu province in the eastern region of China. Traditional biochemical tests and molecular biological methods were performed to distinguish NTM isolates to species/complex level. For the first time, we provide a snapshot of the epidemiology and geographic distribution of NTM in Jiangsu province. The proportion of NTM was 3. 37% of all the Mycobacterium isolates and the species of NTM differed by area. | Abstract
Introduction
Methods
Results
Discussion | 2015 | The Epidemiology and Geographic Distribution of Nontuberculous Mycobacteria Clinical Isolates from Sputum Samples in the Eastern Region of China | 3,613 | 263 |
|
Ribosomal RNAs (rRNAs) in budding yeast are encoded by ~100–200 repeats of a 9. 1kb sequence arranged in tandem on chromosome XII, the ribosomal DNA (rDNA) locus. Copy number of rDNA repeat units in eukaryotic cells is maintained far in excess of the requirement for ribosome biogenesis. Despite the importance of the repeats for both ribosomal and non-ribosomal functions, it is currently not known how “normal” copy number is determined or maintained. To identify essential genes involved in the maintenance of rDNA copy number, we developed a droplet digital PCR based assay to measure rDNA copy number in yeast and used it to screen a yeast conditional temperature-sensitive mutant collection of essential genes. Our screen revealed that low rDNA copy number is associated with compromised DNA replication. Further, subculturing yeast under two separate conditions of DNA replication stress selected for a contraction of the rDNA array independent of the replication fork blocking protein, Fob1. Interestingly, cells with a contracted array grew better than their counterparts with normal copy number under conditions of DNA replication stress. Our data indicate that DNA replication stresses select for a smaller rDNA array. We speculate that this liberates scarce replication factors for use by the rest of the genome, which in turn helps cells complete DNA replication and continue to propagate. Interestingly, tumors from mini chromosome maintenance 2 (MCM2) -deficient mice also show a loss of rDNA repeats. Our data suggest that a reduction in rDNA copy number may indicate a history of DNA replication stress, and that rDNA array size could serve as a diagnostic marker for replication stress. Taken together, these data begin to suggest the selective pressures that combine to yield a “normal” rDNA copy number.
Ribosomal RNAs (rRNAs) in budding yeast are encoded by ~100–200 repeats of a 9. 1kb sequence arranged in tandem on the long arm of chromosome XII, the ribosomal DNA (rDNA) locus. Each 9. 1kb unit is composed of a region encoding a pre-35S rRNA that gives rise to 25S, 18S and 5. 8S rRNAs (transcribed by RNA polymerase I), a small 5S rRNA (transcribed by RNA polymerase III), and two intergenic spacers (IGS1 and IGS2). IGS2 contains an rDNA origin of replication, rARS, and a cohesin associating sequence (CAR). IGS1 contains a replication fork barrier site (RFB) to which Fob1 binds; the binding of Fob1 inhibits DNA replication in the direction opposite to 35S rDNA transcription, preventing the head on collision of transcription and replication machinery [1]. IGS1 also contains a bi-directional RNA polymerase II promoter, E-pro, whose activity is normally suppressed by the binding of Sir2, an NAD-dependent histone deacetylase [2]. Two major features of the rDNA locus give it the unique potential to sense the environment and tune cellular response–high instability, and the wide range of copy number variation it can accommodate. Both features have been extensively studied, particularly in budding yeast, however molecular mechanisms of regulation of instability and copy number and their dependence on one another are not well understood. Because of the tandem nature of the repeats in the rDNA array, the high rates of transcription it needs to accommodate, and the difficulty of replicating repetitive sequences, the rDNA array is highly prone to double stranded breaks (DSBs) and homologous recombination, which leads to loss of repeats at a rate as high as 1 copy per cell division [3,4]. In order to maintain rDNA copy number, the cell has at least two independent mechanisms to restore lost rDNA copies by amplification of repeats. These include unequal sister chromatid exchange, which occurs as a result of clearance of cohesin from rDNA by RNA polymerase II mediated transcription, and rolling circle replication [2,5, 6]. Copy number of rDNA repeat units in eukaryotic cells is maintained far in excess of the requirement for ribosome biogenesis. In yeast, only about half of the ~150 rDNA repeats are actively transcribed to meet the translational needs of the cell [7]. Although only about half of these repeats are actively transcribed, it is known that in budding yeast, reduced copy number makes cells more sensitive to DNA damage [3]. The extra, untranscribed copies are thought to contribute to stability and integrity of the locus by acting as a ‘foothold’ for repair enzymes and for contacts with other regions of the genome [3,8–10]. Human population genome sequencing data analysis has revealed correlations between rDNA copy number and the expression of several genes encoding chromatin-modifying proteins [11], further supporting the idea that the effects of rDNA copy number go well beyond rRNA production for ribosome biogenesis. These could involve, for example, differential recruitment of chromatin-modifying proteins to rDNA arrays with variable copy number. Disproportional binding of chromatin-modifying proteins to the rDNA locus could result in altered concentrations of these proteins throughout the rest of the genome, affecting chromatin environments and transcription genome-wide [12–14]. The hypervariablity of the rDNA locus combined with its ability to titrate various factors gives it the potential to act as a sensor of various conditions or stresses and in response, be a rapid and reversible source of variation for adaptation at the cellular and organismal levels. Several genetic factors have been identified that affect copy number. Of these, some, for example, Rtt109, a histone acetyltransferase, regulate rolling-circle replication of rDNA repeats [2], and others, like Rrn10, may act by regulating RNA polymerase I transcription [15]. On the other hand, some proteins like Sir2 and Fob1 are part of the pathway that senses loss of rDNA repeats and feeds back into regulation of recombination and replication, thereby ensuring that the array does not contract beyond a certain size [5,16]. Although the advantages of having additional copies of rDNA repeats, and the existence of mechanisms to regulate the amplification of rDNA repeats are becoming evident, very few studies have focused on the factors that determine “normal” copy number. Ide et al. [17] first reported that the rDNA array is hypersensitive to perturbations in initiation of DNA replication, and that rDNA array size could modulate cellular response to such perturbations. A more recent study by Kwan et al. [18] also showed that rARS activity affects rDNA copy number as well as genome-wide DNA replication dynamics. Although these studies support the idea that the rDNA array could act as a sensor and source of adaptive response to stress, identification of the mechanisms that determine normal rDNA array size requires an unbiased screen for factors that affect rDNA copy number. To measure rDNA copy number in an accurate manner that is easily adaptable to high-throughput methods, we developed a droplet digital PCR (ddPCR) based assay to measure rDNA copy number in yeast and used it to identify essential genes involved in the maintenance of rDNA copy number. We screened the 787 mutants of the yeast conditional temperature-sensitive (ts) mutant collection of essential genes [19] for mutants with altered rDNA copy number. Our data revealed that low rDNA copy number is associated with compromised DNA replication. Consistent with this, we found that subculturing yeast under two separate conditions of DNA replication stress causes a contraction of the rDNA array independent of the replication fork blocking protein, Fob1. Additionally, we show that a smaller rDNA array enables cells to complete DNA replication in a timely manner, likely by freeing scarce replication factors for use by the rest of the genome, thereby allowing cells to survive and adapt to DNA replication stress. Finally, the loss of rDNA repeats in thymic tumors derived from mice with reduced expression of mini chromosome maintenance 2 (MCM2), a key component of the MCM2-7 DNA replicative helicase, also suggests that rDNA copy number may be used as an indicator of past stress, and that past replication stress may be diagnosed by a contracted rDNA array. Taken together, these data begin to suggest the selective pressures that combine to yield a “normal” rDNA copy number.
Rapid and accurate high-throughput characterization of the genetic determinants of rDNA copy number has thus far been impeded by the lack of simple, sensitive assays to measure copy number. Previous methods of determining rDNA copy number involved measuring changes in the size of chromosome XII by Pulsed Field Gel Electrophoresis (PFGE) and subsequent Southern blotting [20]. More recently, qPCR has been used to measure copy number [21,22]. However, adapting qPCR to high-throughput experiments is difficult because of the requirement for a standard curve for each experiment, and its ability to detect only relatively large changes in copy number. ddPCR is a DNA molecule-counting technique that directly counts the absolute number of DNA molecules in a sample. In ddPCR, targets of interest are partitioned into ~20,000 nano-droplets and amplified to endpoint with TaqMan probes as in qPCR. The concentrations of the targets are then determined by counting the number of fluorescently positive and negative droplets in the sample. Thus, the fluorescence signal in a qPCR is converted from an analog signal into a digital one in ddPCR, eliminating the need for standard curves and allowing the determination of target copy number on an absolute scale with high precision [23]. In our assay, primers and fluorescent probes specific to the 25S coding sequence of the rDNA repeat, and to a stable, essential, single copy control gene, TUB1 were designed (Fig 1A). Absolute copy numbers of each target in the sample are obtained by a duplexed ddPCR reaction, and rDNA copy number per haploid genome calculated from their ratio. The assay was validated by measuring rDNA copy number in multiple independent isolates of two wild-type laboratory yeast strains, BY4741 and W303, and strains with 20–110 rDNA copies, in which rDNA copy number has been previously measured using PFGE [3]. BY4741 has ~150 rDNA copies, and W303 has ~250 rDNA copies, and the strains with 20–110 rDNA copies were also easily distinguishable (Fig 1B). To estimate the range of technical error, we measured rDNA copy number in 8 technical replicates each of both wild-type strains, BY4741 and W303. In our assay, technical error is within 10%, and often within 5% (Fig 1C). Since each reaction is partitioned into ~20,000 droplets, technical error in an individual reaction can be calculated based on the droplet data from that well. Technical error in ddPCR comes mainly from errors due to sub-sampling, and partitioning into droplets. In a good assay, this total technical error should be close to the standard error of the mean, and in our assay, closely matches the error from true technical replicates (Fig 1C). Therefore, we use the errors calculated for each individual reaction as an estimate of technical error, which eliminates the requirement for multiple technical replicates per sample. Biological replicates, however, are critical due to natural biological variability in rDNA copy number. Finally, we also used our ddPCR assay to confirm an increased rDNA copy number in 3 independent isolates each of rtt109Δ, mms22Δ, rrn10Δ, and ctf4Δ strains, which were previously reported to have expanded rDNA arrays [2] (Fig 1D). To screen for essential genes involved in the maintenance of rDNA copy number, we used the ddPCR assay to measure rDNA copy number in high-throughput format in the yeast ts mutant collection of essential genes [19]. This collection contains 787 ts strains, covering 497 (~45%) of the 1,101 essential yeast genes, with ~30% of the genes represented by multiple alleles. The mutants, along with wild-type controls, were grown at the permissive temperature (room temperature), and then shifted to the restrictive temperature (37°C) for 3 hours. Following this, genomic DNA was isolated for ddPCR. A distribution of rDNA copy number across the ~1200 strains screened is shown in Fig 2A. The mean rDNA copy number of wild-type strains was 95 ± 12. 175. Mutants with significantly higher or lower rDNA copy number were identified based on thresholds set by variation in rDNA copy number in wild-type controls (Fig 2A). Of the ~1200 strains screened, 288 had significantly altered copy number (p < 0. 05). The top 89 hits in each category, low and high copy number, were further validated by measuring rDNA copy number in two additional biological replicates (S1 Table). These top hits were also tested for aneuploidies in chromosomes XII and XIII, on which the ddPCR targets reside (S1 Table). Only 5 hits, YDR311W (tfb1-1), YGR002C (swc4-4), YIL126W (sth1-2), and YLR186W (emg1-1), and YNR053C (nog2-1) had a ratio of Chromosome XII to Chromosome XIII that may be representative of aneuploidies that could affect rDNA copy number measurements, and these mutants were excluded from all further analyses. Additionally, rDNA copy number was also measured at the permissive growth temperature to ensure that the restrictive temperature itself did not confound our results. In fact, we found that the hits with altered rDNA copy number after growth at the restrictive temperature also had similarly altered copy number even at the permissive temperature (S1 Table), suggesting that the effects on rDNA array size are likely a result of prolonged propagation in the mutant background. Gene Ontology (GO) enrichment analyses revealed a striking difference in the pathways associated with low or high copy number. Most of the hits with high copy number were mutants in transcription and RNA processing genes, however none of these GO terms were significantly enriched in our hits relative to their background frequencies in the ts mutant collection. On the other hand, the hits with low rDNA copy number were significantly enriched for DNA replication mutants (p<0. 001) (Fig 2B). Further examination of the “low copy number” hits revealed that the most significantly enriched hits were mutants in subunits of DNA polymerases α, δ, and ε, and various replication initiation complexes, such as the Mini chromosome maintenance 2–7 (Mcm2-7) complex, the Origin Recognition Complex (ORC), and the Cdc7-Dbf4 complex. The “low copy number” hits were also significantly enriched for mutants in subunits of the cohesin complex, well known for its multiple roles in maintaining chromosome integrity, including DNA replication and repair [24–26]. Therefore, this data suggested that compromised DNA replication may select for contraction of the rDNA array. Recently Kwan et al. [27] reported that unsolicited, and stably maintained rDNA copy number variations resulting from exposure to lithium acetate were prevalent in yeast mutant collections generated by standard transformation protocols. Although it is still unclear whether exposure to lithium acetate induces unsolicited rDNA copy number variations or simply selects for pre-existing copy number variations, it is possible that the copy number changes in at least a subset of our hits are a result of transformation protocols, and unrelated to the gene mutation. As a preliminary validation of the correlation between the gene mutation and the change in rDNA copy number in our hits, we analyzed the copy number changes in mutant strains of genes with multiple ts alleles, or multiple isolates of the same ts allele. We found that for most genes with validated rDNA copy number changes, a majority of the multiple mutant strains had a similar alteration in rDNA copy number (S1 Table). This is a strong indication that at least for a subset of genes, the mutation itself is likely responsible for the altered rDNA array size. To further validate the findings from our screen, we subcultured yeast cells under two separate conditions of DNA replication stress by a) plating them on medium containing 150mM hydroxyurea (HU), a ribonucleotide reductase inhibitor, or b) depleting Pol1, the largest subunit of DNA polymerase α, using a diploid strain homozygous for the GAL-POL1 allele, containing galactose inducible Pol1 [28]. Both of these conditions caused a contraction of the rDNA array after 50–75 generations, independent of Fob1 (Fig 3). Additionally, rnr1Δ mutants lacking Rnr1, the major isoform of the large subunit of ribonucleotide-diphosphate reductase also have contracted rDNA arrays (S1 Fig). These findings led us to hypothesize that since the rDNA is one of the most unstable and hypervariable loci, and difficult to replicate, DNA replication stresses select for smaller rDNA arrays, perhaps because cells with lower copy number can complete DNA replication in a timely manner and continue to propagate. This raised two main questions– 1) What are the consequences of having a contracted array? 2) Does a smaller rDNA array help cells survive and adapt to DNA replication stress, and if so, how? To address these questions, we used the strain with galactose inducible Pol1 to generate cells with normal or low rDNA copy number as in Fig 3B by growing the strain in medium containing either high or low galactose (high or low DNA polymerase α) respectively. Given the isolate to isolate variability in rDNA copy number, we generated and used 3 independent isolates from each condition for all our studies. Ide et al. [3] reported that low rDNA copy number makes cells more sensitive to DNA damaging agents, such as UV and methyl methanesulfonate (MMS). We found that wild-type, but not fob1Δ isolates with low rDNA copy number generated by Pol1 depletion also exhibited mild sensitivity to bleomycin and UV (S2 Fig). Notably, both wild-type and fob1Δ isolates with low rDNA copy number grew better than those with normal copy number under conditions of DNA polymerase α depletion (Fig 4A). However, selection under altered levels of DNA polymerases is known to induce general genome instability, and often results in mutations that restore normal expression levels [29,30]. Therefore, we wanted to ensure that the survival of the low copy number isolates generated by selection under low levels of DNA polymerase α was not due to chromosomal abnormalities or other mutations which enable expression of normal levels of Pol1. Song et al. [29] showed that trisomy of Chromosome XIV (on which POL1 resides) is very rare after 50–75 generations of selection under low levels of DNA polymerase α. Additionally, we tested for deletions in the GAL1-10 promoter in the normal and low copy number isolates by a PCR-based strategy [30], and found that the GAL1-10 promoter was intact after 50–75 generations of selection (S3 Fig). Additionally, strains with low rDNA copy number (20–110 rDNA copies, [3]) are also less sensitive to HU than corresponding wild-type cells (~200–250 rDNA copies) (Fig 4B) supporting the hypothesis that low rDNA copy number is advantageous to cells under conditions of DNA replication stress. To further characterize how low rDNA copy number helps cells survive DNA replication stress, we measured the fraction of cells in S-phase in asynchronous cultures of both low and normal copy number isolates under conditions of both high and low levels of DNA polymerase α by flow cytometry. We found that with low levels of DNA polymerase α, the high copy number isolates show a higher fraction of cells in S-phase compared to their low copy number counterparts (Fig 5A and 5B, S4A and S4B Fig). This shows that under conditions of DNA replication stress, cells with smaller rDNA arrays are able to proceed through S-phase and complete DNA replication in a timely manner, allowing them to continue to propagate. To test the role of the rDNA array in this rescue, we grew these isolates in high/low galactose (high/low levels of DNA polymerase α respectively) medium containing nicotinamide (NAM), a Sir2 inhibitor. Sir2 represses rARS firing and Sir2 inhibition by NAM will cause more rARS firing [31]. One possible effect of increased rARS firing is the titration of replication factors away from the rest of the genome, which could make the normal copy number strain grow poorly, even with high levels of DNA polymerase α. Another consequence of increased rARS firing in NAM is that the distance each replication fork has to travel in the rDNA will be shorter, which might instead rescue the normal copy number strain with low levels of DNA polymerase α. We found that NAM caused the isolates with normal rDNA copy number to grow poorly with low levels of DNA polymerase α (Fig 5C, S4C Fig). We also observed that the isolates with low rDNA copy number with low levels of DNA polymerase α exhibited mild sensitivity to higher concentrations of NAM (Fig 5C, S4C Fig). Since higher concentrations of NAM could also inhibit other sirtuins, including Hst3 and Hst4, which are known to participate in DNA replication [32], it is possible that the phenotypes observed in NAM are due to a combination of increased rARS firing and additional DNA replication stress imposed by the inhibition of Hst3 and Hst4. Altogether, this suggests that titration of already scarce replication factors by the rDNA is the primary issue faced by the normal copy isolates under conditions of DNA replication stress. Our data show that in yeast, under conditions of DNA replication stresses, a smaller rDNA array provides a selective advantage to cells. Since persistent DNA replication stress resulting from mutations that affect DNA replication machinery have been associated with both cancer [33] and human developmental syndromes [34], we wondered whether these human conditions would also be associated with genomes containing contracted rDNA arrays. To explore this, we used a previously established MCM2-deficient mouse model of cancer, generated by integration of a tamoxifen-inducible form of Cre recombinase downstream of the Mcm2 coding sequence and expression via an internal ribosome entry site (IRES) [35]. In this model, homozygous Mcm2IRES-CreERT2/IRES-CreERT2 (MCM2-deficient) embryos or mouse embryonic fibroblasts have reduced MCM2 expression, approximately one-third of wild-type levels. Although these mice develop normally, MCM2-deficiency results in stem cell deficiencies in multiple tissues, and ultimately, cancer, primarily lymphomas [35]. Moreover, it was also recently reported that the 45S rDNA repeats accumulate high levels of DNA damage in the form of DSBs, both in these MCM2-deficient mice [36], as well as in mouse hematopoietic stem cells (HSCs) with reduced MCM expression [37]. Therefore, we used array-based Comparative Genome Hybridization (aCGH) and Next Generation Sequencing (NGS) to study rDNA copy number in thymic lymphoblastic lymphoma (TLL) tumors derived from MCM2-deficient mice. We performed aCGH on 8 TLL tumors derived from Mcm2IRES-CreERT2/IRES-CreERT2 mice on the 129/Sv (6 of 8 tumors) or 129/Sv x C57Bl/6 F1 (2 of 8 tumors) genetic backgrounds using the NimbleGen 720K whole-genome tiling arrays [38]. All samples were assessed relative to DNA from non-tumorous tissue derived from wild-type littermates of the same genetic background. We also performed NGS on 2 TLL tumor samples derived from Mcm2IRES-CreERT2/IRES-CreERT2 mice on a 129/Sv x C57Bl/6 F1 genetic background and control thymic DNA samples from 2 wild-type littermates. Comparison of aCGH probe log2 ratios for TLL tumor DNA and DNA from 3 clonally derived neural stem cell cultures along with the NGS data confirmed a loss of approximately half of the 45S rDNA repeats in MCM2-deficient tumors (Fig 6). Although the loss of rDNA repeats may simply be a byproduct of tumorigenesis, and not a direct consequence of MCM2 deficiency, these data suggest that under conditions of genotoxic stress, the rDNA may be highly susceptible to DSBs, and while both gain and loss of rDNA repeats may occur frequently in the face of the DNA damage, losses may be selected for under conditions of DNA replication stress. This suggests that rDNA array size may be used as a molecular diagnostic marker for past DNA replication stress.
The rDNA array is the most unstable, and hypervariable genetic locus in the yeast genome. It is the most highly transcribed genomic locus, difficult to replicate owing to its repetitive nature, and comprises over 60% of chromosome XII, and ~10% of the yeast genome. This unique locus, therefore, has the potential to act as the “canary in the coalmine”, being particularly sensitive to stresses that disrupt genome integrity, and acting as a source of cellular adaptation, relatively simply, through variation in array size. In wild-type yeast cells, rDNA repeat copy number is maintained at ~150 per haploid cell. However, owing to the tandem arrangement of the repeats, and the associated instability, repeats are lost at a relatively high rate of up to ~1 per cell division [4]. Therefore, there must be mechanisms in place to a) sense rDNA copy number at each cell division, and b) trigger an appropriate response so that the array size can be maintained at wild-type levels. Studies so far have revealed many pathways that contribute to the maintenance of the rDNA locus, but the fundamental question of what determines normal repeat copy number remains unanswered. Our screen has identified essential proteins involved in maintenance of the rDNA locus based on screening 45% of essential yeast genes; in the future, we can extend this to nearly 75% by screening additional mutant collections [40]. Our data show that DNA replication stresses select for cells that have a smaller rDNA array, and a smaller array helps cells adapt to replication stress. Data from several studies support our findings–i) contraction of the rDNA array can rescue yeast temperature-sensitive mutants of the Origin Recognition Complex, the key replication initiation complex [17], ii) a large deletion of the rDNA array can rescue the synthetic growth defects of rif1Δ cells lacking either MRX or Ctf4-Mms22 activity [41], iii) a polymorphism in the rARS that results in a weakly replicating rDNA array causes a contraction of the array and promotes DNA replication in the rest of the genome [18], and iv) coordination of DNA replication and recombination at the rDNA is critical to maintain the integrity and size of the rDNA array [42]. Our data supports the previously reported role of extra, untranscribed rDNA repeats in protecting cells from DNA damaging agents [3]. In addition, our data suggests that while extra, untranscribed rDNA repeats are essential for allowing co-transcriptional DNA damage repair, they become detrimental to cells under conditions of DNA replication stress. Each rDNA repeat unit consists of a DNA replication origin, the rARS, and the rDNA array represents roughly one-third of all replication origins in the yeast genome [43]. At any given time in S-phase, only a fraction of licensed origins are fired, and this is thought to be achieved by limiting the pool of initiation factors available [44]. Recently, Foss et al. [43] presented evidence to support the long standing idea that, in yeast, repetitive rDNA compete for origin firing with unique genomic sequences (in yeast, the rest of the genome), and tipping the balance in favor of repetitive rDNA can lead to replication gaps, or underreplicated regions, throughout the rest of the genome. Our data also show that under conditions of DNA replication stress, cells with a smaller rDNA array are able to complete DNA replication and proceed through the cell cycle. This, combined with the growth defects of the cells with normal rDNA copy number in NAM, is in agreement with Foss et al. ’s observation that replication gaps produced by deletion of SIR2 can be suppressed by decreased origin activity within the rDNA. Taken together, these data suggest that the repetitive rDNA array may be particularly sensitive to DNA replication stresses, and while DNA damage repair could result in a gain or loss of repeats, the loss of repeats provides a selective advantage by liberating scarce DNA replication factors for use by the rest of the genome. Finally, we speculate that the size of the rDNA array may be used as an indicator of past stress. This has interesting implications for human disease. The link between persistent DNA replication stress and tumorigenesis is well established [33]. Additionally, mutations in key components involved in initiation of DNA replication, such as ORC1, ORC4, ORC6, CDT1, CDT6, and CDC45, have been reported to be the cause of Meier-Gorlin syndrome, a primordial dwarfism syndrome [34,45]. Our data suggest that these human conditions will be associated with genomes containing contracted 45S rDNA arrays. Xu et al. [46] and Wang and Lemos [47] recently showed through bioinformatic analyses of whole genome sequencing data from various cancers that 45S arrays are often lost in cancer. These cancer genomes show evidence of a hyperactive mechanistic target of rapamycin (mTOR) pathway [46] and p53 mutations [47]. Xu et al. further discovered that mouse HSCs lacking PTEN, a negative regulator of mTOR, also had contracted 45S rDNA arrays, and these cells were more sensitive to DNA damaging agents such as bleomycin, MMS, and X-rays. Interestingly, this DNA damage sensitivity was independent of mTOR activity, and mainly attributed to low rDNA copy number. Although PTEN is a phosphatase widely known for its role as a tumor suppressor, nuclear PTEN is essential for maintaining genome stability by dephosphorylating MCM2, and modulating replication fork progression under conditions of DNA replication stress [48], suggesting that the loss of 45S rDNA repeats observed in the Pten-/- mouse HSCs could be attributed to DNA replication stress. DNA replication stress resulting from reduced MCM expression was also shown to drive functional decline in aging mouse HSCs, and caused accumulation of phosphorylated γH2AX at the rDNA [37]. Our findings on the effect of DNA replication stress on the yeast rDNA predict that persistent DNA replication stress would select for a contraction of the rDNA array in each of these cases, and this is in fact what we observe in the tumors from MCM2-deficient mice. Consistent with previous reports, mouse embryonic fibroblasts derived from our MCM2-deficient mice also show increased levels of DNA damage at the 45S rDNA repeats and sensitivity to UV [49]. This is compelling evidence to suggest that DNA replication stress may be diagnosed by rDNA copy number, and cells with low copy number may be more susceptible to common DNA damaging chemotherapeutic agents. Therefore, rDNA copy number may prove to be an important indicator in human disease. Altogether, these results provide insight into the role of the rDNA locus in acting as a sensor and source of rapid, reversible adaption to general genomic stress through a plastic array size determined by selective cues from the environment.
All yeast strains used are listed in S3 Table. Unless otherwise stated, all growth was carried out in YPD (1% w/v yeast extract, 2% w/v peptone, 2% w/v dextrose) at 30°C. For the experiment in Fig 3A, cells were grown in YPD containing 150mM hydroxyurea. The GAL-POL1 strains were grown in YPR (1% w/v yeast extract, 2% w/v peptone, 3% w/v raffinose) medium containing either high (0. 05% w/v) or low (0. 005% w/v) levels of galactose as indicated. Genomic DNA was isolated using the YeaStar Genomic DNA Kit (Zymo Research). DNA concentrations were measured on a Qubit Fluorometer using the Qubit dsDNA HS Assay (Invitrogen). For ddPCR, 0. 005ng genomic DNA was used per 20μL reaction. Primers and probes used are listed in S4 Table. Duplexed ddPCR was performed according to the manufacturer’s protocol (Bio-Rad). Briefly, master mixes containing primers and probes for rDNA and TUB1, genomic DNA, and the restriction endonuclease EcoRI-HF (New England Biolabs, Inc.) were prepared and aliquoted into Eppendorf twin. tec plates. Reaction mixtures were incubated at room temperature for 15 minutes to allow restriction digestion of genomic DNA prior to droplet generation. Droplets were cycled to endpoint and subsequently read using the QX200 droplet reader. Quantification was performed using the Quantasoft software. Standard deviation (SD) for each individual reaction was calculated using the formula Standarddeviation= (CImax-CImin) / (2×1. 96) (1) where, (CImax—CImin) is the 95% Confidence Interval for the ratio of absolute copy number of rDNA and TUB1 in each reaction, with both assays multiplexed in the same well, as generated by Quantasoft. The top 89 hits of each category, low and high copy number, were cherry-picked, and genomic DNA was isolated from them in high-throughput format as described above. The genomic DNA was diluted to 0. 5ng/μL and 0. 01ng used per 20μL reaction for karyotyping. Partial karyotyping to obtain relative number of copies of Chromosomes XII and XIII was done using ddPCR. ddPCR assays for Chromosome XII and XIII were designed and duplexed to determine absolute copy number of each, which was subsequently used to calculate the ratio of chromosome XII relative to chromosome XIII. Primers and probes used, and their relative positions on chromosome arms are listed in S4 Table. Strains were struck out on to appropriate medium from glycerol stocks. Single colonies were picked and re-streaked on to fresh plates every 2–3 days (approximately 25 cell divisions). At each subculture, additional single colonies were also used to isolate genomic DNA for ddPCR. All growth was at 30°C. Cells were diluted to a starting OD600 of 0. 1 and 5 more 5-fold serial dilutions (unless otherwise mentioned), following which 4μL of each dilution was spotted on to plates containing the appropriate medium. Plates were incubated at 30°C for 2–3 days and photographed. 3 isolates each with normal or low rDNA copy number were generated by growing wild-type or fob1Δ GAL-POL1 strains on YPR medium containing either high or low levels of galactose for 50–75 generations. Following confirmation of rDNA array contraction by ddPCR (S2 Table), the isolates were each inoculated into liquid YPR medium containing either high or low levels of galactose. 100μL culture was collected at various time points. Cells were fixed using 70% v/v ethanol and treated with RNaseA, following which their DNA was stained using Sytox Green (1μM, at room temperature, in the dark, for at least 30 minutes). In parallel, wild-type haploid yeast cells for use as ploidy reference were also grown, fixed, treated with RNaseA, and stained with CellTrace Violet (1μM, at 37°C, in the dark, for 20 minutes, followed by 2 washes with 1x PBS) prior to staining with Sytox Green to allow distinction between reference and test strains. Before cytometric analysis, 20μL of the fixed and stained reference sample was added to 1mL of each fixed and stained test sample. DNA content analysis was performed on an EC800 Analyzer (Sony Biotechnology). DNA content modeling was performed using FCS Express 6 Plus (De Novo Software). aCGH, NGS and subsequent data analysis were performed as previously described [38,39]. | Eukaryotic genomes contain many copies of ribosomal DNA (rDNA) genes, usually far in excess of the requirement for cellular ribosome biogenesis. rDNA array size is highly variable, both within and across species. Although it is becoming increasingly evident that the rDNA locus serves extra-coding functions, and several pathways that contribute to maintenance of normal rDNA copy number have been discovered, the mechanisms that determine optimal rDNA array size in a cell remain unknown. Here we identify DNA replication stress as one factor that restricts rDNA copy number. We present evidence suggesting that DNA replication stress selects for cells with smaller rDNA arrays, and that contraction of the rDNA array provides a selective advantage to cells under conditions of DNA replication stress. Loss of rDNA copies may be a useful indicator of a history of replication stress, as observed in a mouse model for cancer. | Abstract
Introduction
Results
Discussion
Materials and methods | chemical compounds
dna-binding proteins
carbohydrates
galactose
organic compounds
dna damage
fungi
polymerases
dna replication
dna
mammalian genomics
molecular biology techniques
research and analysis methods
proteins
chemistry
molecular biology
animal genomics
genetic loci
yeast
biochemistry
biomolecular isolation
dna polymerase
eukaryota
organic chemistry
nucleic acids
dna isolation
genetics
monosaccharides
biology and life sciences
physical sciences
genomics
organisms | 2017 | DNA replication stress restricts ribosomal DNA copy number | 8,787 | 201 |
Mycobacterium tuberculosis (Mtb) mutants lacking rv1411c, which encodes the lipoprotein LprG, and rv1410c, which encodes a putative efflux pump, are dramatically attenuated for growth in mice. Here we show that loss of LprG-Rv1410 in Mtb leads to intracellular triacylglyceride (TAG) accumulation, and overexpression of the locus increases the levels of TAG in the culture medium, demonstrating a role of this locus in TAG transport. LprG binds TAG within a large hydrophobic cleft and is sufficient to transfer TAG from donor to acceptor membranes. Further, LprG-Rv1410 is critical for broadly regulating bacterial growth and metabolism in vitro during carbon restriction and in vivo during infection of mice. The growth defect in mice is due to disrupted bacterial metabolism and occurs independently of key immune regulators. The in vivo essentiality of this locus suggests that this export system and other regulators of metabolism should be considered as targets for novel therapeutics.
Tuberculosis continues to be a major global health threat. Mycobacterium tuberculosis (Mtb) is estimated to infect 2 billion people worldwide, or one-third of the world’s population [1]. Despite its clinical importance, key aspects of tuberculosis (TB) pathogenesis are still not understood, including predictors of whether exposure will lead to active versus latent disease. Only 5% of exposed individuals will go on to develop active disease, whereas the remaining 95% will develop latent disease but remain susceptible to reactivation [2]. Therefore, Mtb is able to survive during periods of reduced growth and has the capacity to regrow rapidly. How does this happen? Mtb orchestrates growth arrest in response to stresses encountered in the host, and lipid metabolism, a central part of Mtb’s life inside the host, likely plays an integral role [3,4]. While the synthesis and degradation of lipids has been extensively studied, their transport is far less well understood. We have previously identified two genes, rv1411c and rv1410c, that form an operon and are conditionally essential for survival in the mouse [5,6]. Rv1411c encodes the mycobacterial lipoprotein LprG, which can bind triacylated phospholipids such as phosphatidylinositol mannoside (PIM) and lipoarabinomannan (LAM) [7] and is necessary for normal surface display of LAM [8,9]. However, unlike other known mycobacterial lipoproteins, LprG is in an operon with a putative integral membrane transporter, Rv1410, a member of the major facilitator superfamily (MFS) [10,11]. Both genes are required in M. smegmatis for normal cell wall composition and for efflux of toxins such as ethidium bromide [12]. LprG has structural homology to the mycobacterial lipoprotein LppX, which, along with the transporter MmpL7, is required for the outer membrane localization of phthiocerol dimycocerosate (PDIM), a virulence-associated lipid [13,14]. By analogy, LprG and Rv1410 might function together to position mycobacterial lipids in the cell wall. LprG binds to PIM and LAM, two TLR2 agonists, and is predicted to transport these two lipids to the cell surface of Mtb [7]. Deletion of LprG limits TLR2 activation and blocks certain aspects of phagolysosomal fusion [7,8] but it is unclear whether these effects account for the strong in vivo growth attenuation of LprG-deficient bacteria. Loss of TLR2 does not alter growth of Mtb in vivo [15] and loss of key components of the TLR-induced signaling cascade, such as MyD88, actually worsen infection due to effects of IFN-γ-mediated activation of macrophages [16,17]. Thus, mislocalization of two TLR2 agonists, PIM and LAM, would not be expected to cause the significant in vivo attenuation observed upon loss of LprG function. Given that LprG binds several classes of Mtb lipids, at least in vitro, we instead posited that the distribution of other lipids within Mtb is affected by the loss of LprG and Rv1410. To investigate this possibility, we used unbiased lipidomic analysis to examine the abundance and distribution of lipids in Mtb. We found that disruption of LprG-Rv1410 function leads to an increased level of intracellular TAG. Furthermore, LprG co-crystallizes with TAG and transports TAG between lipid membranes in vitro, implying that these proteins are required to transport TAG out of the cytoplasm. This conclusion is supported by data showing that overexpression of LprG-Rv1410 increases release of TAG into the culture medium of broth-grown Mtb. Finally, loss of LprG-Rv1410 function is associated with Mtb growth attenuation both in vitro during carbon restriction and in vivo. Thus, we identify LprG as a physiologically important TAG transporter. This previously unknown mechanism likely explains the surprising presence of recently discovered TAG in the outer membrane of Mycobacterium smegmatis [18]. Further, because TAG plays an essential role in Mtb metabolism [19,20], we propose that an altered metabolic state is likely ultimately responsible for the significantly reduced survival and virulence of LprG-Rv1410 mutants in the host.
Our previous studies suggested that LprG and Rv1410 act together in processes that affect cell wall integrity [12]. Thus, we hypothesized that mutations that disrupted one or both genes would have similar phenotypes. We used two different strains: an Mtb H37Rv strain with a transposon insertion in rv1410c (rv1410c: : Tn; Mut1) [5] and a strain from which we deleted both genes by allelic exchange (ΔlprG-rv1410c, Mut2) (Fig 1A and S1 Fig and S1 Methods). To analyze lipids produced by each strain, we performed comparative lipidomic analyses [21]. Briefly, we extracted cell-associated lipids from Mtb during stationary growth phase and analyzed the total lipid content by normal phase high performance liquid chromatography-mass spectrometry (HPLC-MS). We performed initial analyses in both positive and negative ion modes, but, because we found fewer reproducible lipid alterations in negative mode, we focused further experiments on positive mode analysis. We conducted all experiments in biological triplicate and assigned ions appearing in two or more analyses with nearly identical mass-to-charge ratio (m/z) and retention time as ‘molecular events. ’ By generating mass spectrometry intensity ratios for many thousands of paired events from wild-type and genetically altered mycobacteria, this method provides organism-wide descriptions of lipid change and is capable of identifying individual changed lipids. For analysis, we aligned all events with equivalent mass and retention time across datasets to enable detection of changed events, which are defined as having an intensity value altered by 2-fold with a corrected p value < 0. 05. Among 7487 total events detected, 309 (4. 1%) met these criteria. Previous analyses estimate the rate of false positive changed lipids as being less than one percent [21]. The percentage of changed events detected here (4. 1%) is above method-specific variance and similar to the percent change detected between the Mtb strains H37Rv and W Beijing [21]. The absolute number of changed events exceeded our ability to identify them, so we used previously validated criteria to group and prioritize events for molecular identification [22,23]. In ranking events of known mass but unknown chemical identity, we prioritized those with highest absolute fold-change and lowest p values. Further, in normal phase chromatography, lipid classes are separated in time, but alkylforms within one class show nearly identical retention time with mass intervals characteristic of unsaturations (H2) or alkyl (CH2) chain length variants. In this way, the large list of changed ions can be grouped into a smaller number of lipid classes, even in the absence of knowing the chemical name of each lipid. Within each class, the intensity values of each member can be individually assessed to see whether they change with the same direction and magnitude after mutation. Therefore, as a third criterion for highlighting events of biological interest, we sought lipid classes that showed parallel changes among all events corresponding to the alkylforms in a given lipid class. Combining these three criteria, we observed a striking pattern for 32 events that all belonged to the same lipid class (Fig 1B, blue). These 32 events increased in intensity with similar magnitude in the knockout strain and matched masses of TAG alkylforms in the MycoMass database. Ion chromatograms showed that the events had the expected retention time of TAG (3–4 min) as annotated in the MycoMap database [21]. The structures of TAG alkylforms with a combined lipid length and saturation in the three chains of 48: 0 (m/z 824. 77), 58: 0 (m/z 964. 94), and 59: 1 (m/z 976. 93) corresponded to the expected total lipid lengths of mycobacterial TAG. In order to formally rule the identity of an event as TAG, we performed collision induced mass spectrometry (CID-MS) and detected the fragments of a protonated adduct of C58 (Fig 1C) and other ions in the series (S2 Fig) as TAG having the expected mass intervals of C16 and C26 fatty acyl units as well as the ions corresponding to the expected diacylglyceride units (m/z 691. 66, m/z 551. 50). Because Rv1410 is a predicted integral membrane protein and LprG is a lipoprotein that is shed into culture medium [24], we hypothesized that Rv1410 functions upstream of LprG in regulating lipid transport across the plasma membrane. To validate this, and to control for potential differences unlinked to the disrupted locus, we performed comparative lipidomics analyses on Mut1 and Mut2 strains to assess the overlap in lipid perturbation phenotypes and focused our analysis on lipids that increased in both mutant strains. By this more stringent analysis, increases in TAG remained the major lipid alterations in both mutant strains. Individual TAG species with differing length (C56-62) and unsaturation (0–2) were detected at statistically significantly elevated levels in both mutant versus wild-type strains across repeat experiments as compared to other detected lipids (Fig 1D and S3 Fig). The absolute fold-change in signals were large, more than 50-fold for most alkylforms, with increases in signal intensity for TAG in Mut1 and Mut2 compared to WT ranging from 2- to 100-fold depending on the TAG alkylform and strain evaluated. Upon complementation with a single copy of the lprG-rv1410c operon (ΔlprG-rv1410c L5: : lprG-rv1410c, Comp2), we detected reduced amounts of TAG species compared to Mut2 for every alkylform (Fig 1E). Previous studies have shown that the outer layers of the mycobacterial cell wall contain TAG [25,26]. Further, as shown by reverse micelle extraction of mycobacterial surface lipids, TAG represents a significant proportion of non-covalently associated lipids in the outer membrane of mycobacteria [18]. Based on our lipidomics analysis, we hypothesized that LprG and Rv1410 represent a formerly unknown mechanism to transport TAG from the cytoplasm to the outer membrane. Previous work on the structure of LprG revealed a large central cavity that can accommodate triacylated lipid species, including triacylated phosphatidylinositol (Ac1PIM2) [7]. Based on a model in which Rv1410 transports TAG across the cytosolic membrane and then transfers TAG to LprG in the cell wall, we hypothesized that LprG should bind similarly to TAG. As predicted, co-crystallization with TAG (tripalmitoylglyceride, TAG 48: 0) revealed that the complex (Fig 2A) contains TAG, which is seated similarly to Ac1PIM2 in the LprG binding site (PDB, 4ZRA). Briefly, the alpha-beta fold is maintained and consists of 10 anti-parallel ß-strands apposed by 6 α-helices that define a central cavity whose entrance is lined by multiple loops (Fig 2A). The TAG glyceryl group is located at this entrance between loops L1 (aa 65–72) and L2 (aa 95–98). Compared to the position of Ac1PIM2 bound to LprG, the orientation of TAG is conserved: the hydrophilic glyceryl headgroup interacts with Ser72 at the cavity entrance. The acyl chains form interactions with the side chains of hydrophobic residues within the cavity. However, while the three acyl chains of Ac1PIM2 are all bound within the cavity, only two of the TAG palmitoyl chains (sn1 and sn2) are similarly buried; the third acyl chain (sn3) is exposed to the solvent (Fig 2B). The first ten carbons of the sn3 palmitoyl chain form van der Waals contacts with the hydrophobic sidechains at the cavity entrance, including Leu95, Leu115, Phe123 and Ile129. Furthermore, the sn3 chain is close to two hydrophobic grooves formed by Phe123-Leu115-Ile100 and Leu73-Leu71-Ile68-Val64 (Fig 2B). This proximity suggests that these grooves could facilitate the binding of TAG with longer acyl chains. The lid of the cavity, composed of a helix-loop motif (α2-loop1- α3-loop2), is flexible and was found in two conformations that define distinct cavity volumes. The volume of the open form with TAG bound is 1425 Å3 and the closed form is 895 Å3 due to a major inward movement of the cavity lid (aa 129–134). In the closed form, the hydroxyl group on the side chain of Tyr130 is shifted 4. 5 Å toward the center of the cavity, as measured from the hydroxyl oxygen atom on the side chain. Comparing the helix-loop motif in the closed and open forms, the side chain of Tyr130 blocks the binding site for the sn2-palmitoyl chain of TAG. Therefore, different conformations of this helix-loop motif may help accommodate the binding of diverse triacylglycerides. Given the predicted localization of the two proteins, we hypothesized that Rv1410 transports TAG across the inner membrane and passes it to LprG, which then transfers the lipid to the outer membrane. Thus, LprG should not only bind to TAG but facilitate movement of TAG between lipid membranes, analogous to the suggested role of LppX in PDIM localization to the outer membrane [13]. To test the lipid transfer ability of LprG we used purified proteins in a vesicle-based assay to measure TAG movement between lipid membranes (S4 Fig) [27]. A protein-dependent increase in fluorescence indicates the net movement of fluorescently labeled TAG (dioleoyl-sn3-nitrobenzoxadiazole-C6 glyceride; NBD-TAG) from donor vesicles containing partially quenched NBD-TAG to acceptor vesicles without TAG and thereby demonstrates whether a protein can both extract TAG from and release it into lipid membranes. We found that LprG was able to transfer lipids in a protein dose-dependent manner (Fig 2C). Also, LprG-V91W, a mutant containing a bulky mutation located within the hydrophobic cavity, transferred TAG at a decreased rate and yield (Fig 2C and 2D) [7]. The marked reduction in activity is consistent with the reported lower binding affinity of LprG-V91W for triacyl lipids. In contrast, the LprG homologue LprA had no detectable activity (Fig 2C and 2D), consistent with our prediction that it cannot bind TAG due to a smaller binding pocket that accommodates only diacylated lipids [7]. Thus, LprG can transfer TAG between lipid membranes, as predicted from its hypothesized role as a lipid transport protein. Recent data suggest that TAG make up a significant proportion of non-covalently associated lipids in the outer leaflet of the outer membrane of mycobacteria [18]. If TAG is transported from the cytoplasm into the cell wall we predicted that it might be shed into culture medium of broth-grown cultures and be detectable by MS. To test this, we compared the quantities of TAG in culture medium derived from both mutant and wild- type strains. Lipidomics analysis of culture medium collected from the WT, Mut2, and Comp2 cells in logarithmic growth phase revealed low signals for lipids shed into the culture medium with few relative differences between the strains in both positive and negative mode analysis. In retrospect, this was not surprising given that we were able to detect differences in total cell lipids between strains from stationary phase, but not log-phase, cultures grown in broth medium. We therefore asked instead whether overexpression of LprG-Rv1410 would lead to increased quantities of TAG in culture medium (Fig 3A and S5 Fig). Indeed, we found a greater signal intensity corresponding to TAG in the culture medium from the overexpression strain (OE) as compared to the LprG-Rv1410 mutant (Mut2) (Fig 3B, upper panel). Importantly, OE had increased TAG in culture medium despite having equivalent to less total cell TAG as compared to Mut2 (Fig 3B, lower panel). The increased shedding of TAG into culture medium was confirmed using extracted ion chromatograms that monitor TAG (58: 0) ion counts across all genetically altered and WT mycobacteria (Fig 3C). Furthermore, a trend toward higher normalized TAG signals in the culture medium of OE was seen for each of the individual TAG alkylforms and the difference reached statistical significance for many of the alkylforms. This conclusion was further supported by low detection of the highly abundant inner membrane associated phospholipids, phosphatidylethanolamine (PE) and cardiolipin (CL), indicating that inner membrane and cytosolic lipid did not substantially contaminate the filtered cultured medium. Increased TAG in culture medium of OE is unlikely to be a non-specific effect of overexpression since amounts of control lipids were not increased in culture medium of OE. The identity of TAG and control lipids in culture medium was confirmed by MS/MS fragmentation analysis. Similar results were consistent across two additional experiments that measured TAG accumulation in culture medium of OE across a range of growth conditions (S6 Fig). Complementation of the LprG-Rv1410 operon restores virulence of LprG-Rv1410 mutants in mice (S7 Fig) demonstrating that the virulence defect is attributable to loss of function at this specific locus. Previous work suggests that 1) LprG promotes the activation of the innate immune receptor TLR2 through export of lipoglycans such as PIM and LAM [7,28] and 2) plays a role in LAM display [8,9]. Both hypotheses attribute the attenuation of the LprG-Rv1410 mutant in mice to an altered interaction with host immune cells, namely phagocytes, leading to enhanced clearance [8,9]. We reasoned that if either immune mechanism accounted for the strong attenuation seen during infection [29], then loss of innate and acquired immune functions that occur downstream of both of these processes should rescue growth of LprG-Rv1410 mutants in vivo. To examine the role of the innate immune response, we infected nos2-/- and phox-/- mice that are compromised for oxidative killing mechanisms and neutrophil function, respectively. Oxidative killing by nitric oxide generated by inducible nitric oxide synthase (iNOS; nos2) in macrophages and neutrophils is a well-documented mechanism for Mtb control in mice [30–32]. Similarly, NADPH oxidase (Phox; phox) is required for reactive oxygen species (ROS) killing of mycobacteria by neutrophils and macrophages downstream of TLR signaling in response to infection [33,34]. C57/Bl6 wild-type and mutant mice were infected with a 3: 1 mixture of Mut1 and the matching H37Rv WT background control. Mut1 was attenuated in the lung and spleen in all mouse strains, as assessed by colony forming units (cfu) (Fig 4A and 4B and 4C). Furthermore, loss of generalized oxidative killing also failed to rescue growth of Mut1 in phox -/- mice administered the iNOS inhibitor aminoguanidine (Fig 4D). Thus, the attenuation of a mutant lacking Rv1410 activity is independent of these innate immune control mechanisms. The onset of the adaptive immune response initiates secondary control mechanisms against Mtb growth. Specifically, the pro-inflammatory cytokine IFN-γ is critical for Mtb infection control in both mice and humans [35–37]. Importantly, Mtb can downregulate host immune responses by dampening classical macrophage activation and IFN-γ mediated effector functions via TLR2 signaling [38]. If the absence of lipid bound-LprG in LprG-Rv1410 mutants results in attenuation due to a shift towards classical macrophage activation, then we predicted that the growth of these mutants should be rescued in ifn-γ -/- mice. We infected ifn-γ -/- mice with mixed Mut1 and WT Mtb and found that the mutant strain was again attenuated relative to WT even though three times the infective dose of Mut1 vs. WT was delivered (Fig 4E). IFN-γ is only one of many effectors that might play a role downstream of adaptive immunity. To more comprehensively test the role of the host adaptive immune response in controlling infection with LprG-Rv1410 mutants we used rag-/- or severe combined immunodeficiency (SCID) mice, which both lack functional B and T lymphocytes. Rag-deficient mice succumbed to disease at 3 weeks post infection with WT bacteria, whereas Mut1-infected mice were still alive at 5 months post infection when the experiment was terminated (Fig 5A). At the end of the experiment, three Mut1-infected mice were moribund with numerous acid-fast bacilli in their lungs (Fig 5B). This was not specific to the bacterial mutant, the mouse strain or the route of infection, as SCID mice infected with the Mut2 strain by the low-dose aerosol route also survived for an extended time (Fig 5C). Thus, the attenuation of the LprG-Rv1410 mutant is seen in mice with variously altered immune responses (S1 Table), including innate and downstream adaptive mechanisms that are thought to be relevant to PIM and LAM action. If loss of LprG-Rv1410 interaction with host immunity does not explain lack of growth of LprG-Rv1410 mutants in vivo, why are these bacteria so markedly attenuated? One clue comes from longitudinal histopathologic examination of organs harvested from immunocompromised mice infected with LprG-Rv1410 mutants. We sacrificed a subset of Mut2-infected SCID mice at time points matching the time of death of WT- and Comp2-infected animals. At earlier time points (60 days) there were few bacteria and little pathologic change in tissues of Mut2-infected lungs, by gross and microscopic pathology (Fig 5D and 5E). Mut2-infected lungs showed only early evidence of thickening of the alveolar septa due to an influx of macrophages (Fig 5D, middle and inset). Histiocytic consolidation, suppuration, and necrosis associated with bacterial replication and/or cytotoxic response were not evident in Mut2-infected mice, as seen in WT and Comp2-infected mouse lungs at the time that mice became moribund (Fig 5D, left and right). Cfu from the lungs of SCID mice infected with the LprG-Rv1410 mutant were 2 log-fold lower in mice infected with Mut2 at matched time points compared to both WT- and Comp2-infected mice (Fig 5F). However, after more prolonged infection, bacterial load increased systemically and the histopathology in Mut2-infected moribund mice was similar to that seen at early time-points with WT or Comp2 Mtb strains at time of death. Cfu from the spleen of Mut2- infected mice when moribund, although still lower than for WT and Comp2 when moribund, were nearly 1x106. At six months post-infection aggregates of viable and necrotic macrophages containing numerous acid-fast bacilli were present in the lung, spleen, heart, and kidney of Mut2-infected SCID mice, consistent with the perimortem changes described above for WT- and Comp2-infected SCID mice at 50–60 days post-infection (S8 Fig). These results suggest that Mtb lacking LprG-Rv1410 induce the same host reaction as WT but take longer to accumulate the bacterial mass necessary to induce pathologic changes. Comparison of the growth trajectory as measured by cfu over time across the three strains suggests Mut2 grew more slowly during infection, again pointing away from host response as the major mechanism by which LprG-Rv1410 affects Mtb virulence. Although LprG-Rv1410 mutants grow slowly in the host environment, they exhibit no growth defect in vitro in standard culture medium (Fig 6A and S2 Table). Furthermore, we were unable to detect reproducible differences between wild-type and LprG-Rv1410 mutant strains with in vitro infections of several murine monocytic cell lines. One potential explanation for this dichotomy is that Mtb utilizes lipids as a carbon source during infection, while glycerol is the predominant carbon source provided in artificial medium in vitro [39]. Thus, we tested whether the growth of Mtb mutants would be slower when grown in vitro with lipids as the sole carbon source. Indeed, Mut2 had a reduced growth rate when cholesterol was the primary carbon source (Fig 6B and S2 Table). Growth rate attenuation was even more striking on propionate, a toxic byproduct of cholesterol and fatty acid metabolism [40,41] (Fig 6C and 6D and 6E and S2 Table). The growth defect was not reversible with complementation of either LprG or Rv1410 alone (Fig 6C). Vitamin B12 (VitB12) supplementation also failed to rescue the growth of Mut2, suggesting that the deficiency is unlikely a result of a defect in the methylmalonyl pathway of propionate detoxification (S9 Fig). Therefore, the LprG-Rv1410 mutant displays an in vitro growth defect when grown on lipids similar to those seen during infection. We hypothesized that the growth defect of Mut2 when the carbon source is restricted to lipids is the result of intracellular TAG accumulation and that blocking TAG transport via the loss of LprG-Rv1410 function enhances TAG accumulation. If altered levels of intracellular TAG in LprG-Rv1410 mutants can account for their growth defect in vivo, we predicted altered growth kinetics of LprG-Rv1410 mutants in vitro due to modified TAG levels. Given the large numbers of genes encoding both TAG synthases and lipases in mycobacteria [42,43], genetically altering TAG metabolism proved difficult, so we opted to modulate TAG levels pharmacologically by inhibiting TAG hydrolysis. We predicted that preventing TAG breakdown would exacerbate the growth defect of Mut2. Treatment with tetrahydrolipostatin (THL) decreases TAG hydrolysis by lipases such as LipY, which is thought to mobilize stored TAG as an energy source during the transition out of hypoxia [20]. During growth with cholesterol as the sole carbon source, THL at 5 μg/ml was bacteriostatic during the first 5 days of culture for all strains tested (Fig 6B). The growth rate of the WT and Comp2 subsequently recovered and, interestingly, mirrored the growth rate of Mut2 in the absence of THL. In contrast, the Mut2 treated with THL did not grow throughout the 7-day culture period. Conversely, TAG hydrolysis by lipases is thought, via β-oxidation, to provide acetyl Co-A for both lipid synthesis via FasI and anapleurosis of the TCA cycle [44,45]. Therefore we predicted that bypassing the need for TAG hydrolysis should at least partially relieve the LprG-Rv1410-dependent growth defect. To test this, we supplemented the medium with acetate, which has been shown to rescue the growth of Mtb mutants with defects in the methylcitrate cycle or in cholesterol metabolism when grown in the presence of propionate or cholesterol, respectively [46,47]. As predicted, acetate partially rescued the growth of Mut2 on propionate in both minimal and rich medium (Fig 6D and 6E and S2 Table). Similarly, the addition of glycerol and/or oleic acid rescued growth kinetics of Mut2, which under these conditions were indistinguishable from those of WT and Comp2 strains (Fig 6F and 6G and S2 Table). Furthermore, we determined that the LprG-Rv1410 mutant has decreased susceptibility to THL in medium supplemented with glycerol and oleic acid (S10 Fig and S1 Methods). Therefore, although inducing TAG accumulation mimics or exacerbates the loss of LprG-Rv1410 function, mimicking TAG hydrolysis by supplementing TAG breakdown products such as fatty acids and glycerol has the potential to relieve this stress and restore growth. Growth rescue by acetate, glycerol, and free fatty acids (FFA) is likely mediated by anapleurosis of the TCA cycle, but in the context of TAG accumulation, increased levels of these metabolites may also override inhibitory effects of intracellular TAG by relieving numerous potential negative feedback mechanisms, on processes ranging from fatty acid synthesis to cell cycle regulation and cell division, that are only beginning to be described in mycobacteria [48–52].
Mycobacteria, and Actinomycetes in general, are unique not only in their capacity to synthesize and store large quantities of TAG, but also in their ability to catabolize TAG as an energy source during starvation [19,53] [54]. TAG accumulates within lipid droplets in the bacterial cytoplasm [42,55] and is associated with slow growth and antibiotic tolerance (52). TAG in the form of lipid droplets provides energy via β-oxidation of the acyl chains [45,56] and TAG seems to serve a structural role as a major component of the outer leaflet of the outer membrane [18]. Like PDIM, TAG can also serve as a sink to alleviate the accumulation of potentially toxic propionyl-CoA [46,57]. Here we show loss of LprG and Rv1410 disrupts steady-state levels of intracellular TAG in Mtb. This could result in multiple cellular defects, consistent with the numerous roles that TAG is proposed to play in mycobacterial physiology and metabolism. Similar to a recent report [58], we show that loss of LprG-Rv1410 results in marked slowing of growth in vitro when Mtb are restricted to lipid carbon sources, such as cholesterol, that are likely utilized during infection. Others have shown this gene defect causes immune-activating defects in vitro, but several lines of evidence suggest that the metabolic defects associated with LprG-Rv1410 loss play a larger role in growth attenuation in vivo. In the host, lipids represent the major carbon sources available to the pathogen and these conditions favor TAG production [59]. We propose that, under such conditions, wild-type Mtb can either transport TAG out of the cytoplasm into the outer layers of the cell wall or store intracellular TAG, the subsequent hydrolysis of which generates energy via anapleurosis of the TCA cycle and enables growth (Fig 7A). Indeed, the growth defect of the LprG-Rv1410 mutant can be titrated by chemical modulation of TAG lipolysis using THL or by providing alternative catabolites, such as acetate and glycerol. The modulation of the growth defect under such conditions is thus likely a result of altered metabolism. In our model cytosolic TAG accumulation restricts bacterial growth and LprG-Rv1410-mediated transport of TAG into the cell wall functions as an adjunct to lipases to eliminate cytosolic TAG. In mycobacteria stresses such as acidic pH, carbon restriction, acid stress and hypoxia induce shifts in central carbon and lipid metabolism in Mtb that lead to intracellular TAG accumulation [4,19,42,44]. Under these conditions there has typically been an association of intracellular TAG with an altered growth state. We present a model by which TAG may chemically or physically participate in growth regulation during the carbon restriction experienced in the host. One possibility is that cytosolic TAG levels in mycobacteria contribute to negative feedback mechanisms known to regulate lipid metabolism in other bacterial species. For example, FFA act as regulatory molecules in E. coli lipid biosynthesis [60] and recent data suggest that FFA regulate lipid synthesis by regulating fatty acid synthase (FasI) in both mycobacteria and corynebacteria [48,61,62]. FFA have also recently been shown to bind transcription factors that regulate the MmpL family of transporters, some of which transport specific classes of outer membrane lipids and are also associated with lipid biosynthesis [63]. If FFA released by TAG hydrolysis act in a similar fashion, TAG transport out of the cytosol and into the outer membrane could allow for sequestration of these lipids as a means to regulate central carbon and lipid metabolism in mycobacteria. One alternative explanation for our findings is that loss of LprG-Rv1410 results in increased TAG production. However, increased TAG production occurs normally during the course of wild-type Mtb infection, as seen with lipid body formation, and is therefore unlikely to account for the attenuation of LprG-Rv1410 mutants in mice. The model presented here suggests that TAG transport by LprG-Rv1410 is critical under conditions that favor increased TAG production and storage during cellular stress. During growth resuscitation after hypoxic stress, intracellular stores of TAG are hydrolyzed, providing acetyl-CoA to fuel the TCA cycle. A similar dynamic likely occurs during carbon restriction. We propose that titration of growth in response to carbon restriction is partly a function of intracellular TAG levels and that these levels are modulated by TAG export and/or hydrolysis. In the background of LprG-Rv1410 loss, this balance is shifted towards TAG accumulation and therefore bacterial growth arrest is sustained (Fig 7B). It has been proposed that one way in which mycobacteria can co-catabolize multiple carbon sources is by segregating lipid catabolism into different compartments within the cell, including the periplasmic space [64]. One possibility is that TAG transport by LprG-Rv1410 contributes to carbon source segregation in mycobacteria. Our model can account for the attenuation and slow growth rate seen with the LprG-Rv1410 mutant in SCID mice and the overall lack of growth rescue irrespective of host immune status. In this study we set out to uncover the mechanism responsible for the in vivo attenuation reported for both LprG and Rv1410 mutants in mice [5]. To date, much of the data on the role of the lprG-rv1410c operon has centered on LprG, an experimentally more accessible molecule. For the phenotypes we have tested in Mycobacterium tuberculosis and Mycobacterium smegmatis [12], LprG mutants essentially phenocopy Rv1410 mutants, suggesting that these proteins act in concert to perform an essential function during growth within the host. Here we report that LprG binds TAG and is capable of mediating lipid transfer between vesicles. Thus, it appears likely that Rv1410 is responsible for the transport of TAG across the cytosolic membrane, allowing LprG to access TAG, although we cannot measure this directly. This model predicts that mislocalization of TAG would contribute to the increased levels detected in LprG-Rv1410 mutants. Our studies evaluating TAG released into culture medium show that overexpression of LprG-Rv1410 increases the presence of TAG, further supporting our hypothesis that TAG is a quantitatively significant substrate of the LprG-Rv1410 transport system. Is TAG transport the only role for the LprG-Rv1410 system? Other reports show that various lipids can bind to LprG in vitro and that cells that lack this transporter system display reduced levels of surface LAM [8,9]. TAG, PIM and LAM are triacylated lipids that are all proven to bind and occupy the large cavity of LprG, suggesting that the system operates to control the localization of triacylated lipids. [8,9]. However, upon disruption of the LprG-Rv1410 locus, steady-state quantities of TAG change more than 50-fold, which is much greater than the changes observed for the immunogenic phospholipids PIM and LAM. Demonstration of differential localization of TAG within the cell wall will be necessary to confirm our model that LprG-Rv1410 physically transports TAG. Nevertheless, our data suggest that the conditional essentiality of the lprG-rv1410c operon in vivo is a direct consequence of its non-redundant role in modulating intracellular TAG and, thereby, lipid metabolism and growth regulation. TAG storage has long been a defining feature of mycobacterial infection. TAG in Mtb has served primarily as a marker for stationary, dormant, or latent states, all of which have implications for drug effectiveness, drug resistance, and disease progression. Yet, we have only rudimentary knowledge about how TAG contributes to the processes leading to these states and overall cellular metabolism at the host-pathogen interface, an understanding of which is critical to the development of new preventive and therapeutic approaches to infection control. We propose that TAG transport critically regulates intracellular TAG levels during certain growth states within the host [19,42,55,65]. Certainly, lipid metabolism is central to growth regulation in mycobacteria. Our model suggests that synthesis, transport, and hydrolysis of TAG are balanced in a manner that regulates growth rate, assigning a far more complex role for TAG in Mtb cellular homeostasis than has been previously suggested. Clearly, though, regulation is more than just abundance, and future studies will need to evaluate the subcellular distribution of TAG and other lipid mediators in order to determine their collective roles in lipid homeostasis and potential growth regulation.
The Harvard Medical School (HMS) IACUC approved the animal care and use protocol 03000. The HMS IACUC adheres to the Public Health Service Policy and Animal Welfare Act. The virulent H37Rv strain of Mtb was used for all experiments unless otherwise noted. Cells were routinely cultured in Middlebrook 7H9 broth and supplemented with 10% (vol/vol) OADC (Middlebrook), 0. 2% glycerol, and 0. 05% Tween 80. For cells destined for lipidomics analysis, Tween-free media was used. Cells were maintained at 37°C with shaking at 100rpm. Lipids were extracted and analyzed as previously described from Mtb during either logarithmic growth phase in 7H9 liquid medium (O. D. 0. 8 +/- 0. 1) or stationary growth phase from 7H9 agar plates (1. 5% w/v Bacto-Agar) [21]. For extraction of lipids from culture medium, approximately 50 mL of culture medium obtained after centrifugation of Mtb grown in Tween-free medium was filtered twice with 1 μm average pore-size filters (Millipore). Lipids were extracted by addition of 6 N HCl (0. 3% final v/v) followed by addition of 140% v/v ethyl acetate for 30 minutes with gentle rocking. Extracts were centrifuged and the upper organic phase collected, pooled, and evaporated to dryness using a Genevac system (SP Scientific). Lipid masses per sample were measured and dried lipids were resuspended to 1mg/mL in 1: 1 chloroform: methanol. 5 μg to 20 μg of total lipid per technical and biological run were injected for HPLC/MS analyses. Sample titration was necessary to determine sample load for optimal detection and quantitation of TAG using normal phase chromatography. Lipids events were detected and defined as previously described using positive mode HPLC/MS [21], and lipidomes were analyzed using XCMS algorithm in R (Scripps) and GenePattern Software (Broad Institute). See also supplemental experimental procedures. C-terminally 6xHis-tagged recombinant proteins were cloned, overexpressed, and purified largely as described (Suppl Experimental Methods) [28]. TAG transfer assays were performed as previously described [27]. Briefly, donor vesicles containing 450 nmol/mL phosphatidylcholine (PC) and 14 nmol/mL of fluorescently labeled NBD-TAG (dioleoyl-sn3-nitrobenzoxadiazole-C6 glyceride, custom synthesis by Avanti Polar Lipids) were mixed 1: 1 with acceptor vesicles containing 2400 nmol/mL of PC either with or without LprG, LprG-V91W (LprG-VW), or LprA and buffer for a final volume of 100 μL. Plates were incubated at 37°C and fluorescence was monitored as a function of time with a 7620 Microplate Fluorometer (Cambridge Technology, Watertown, MA) using 460-nm excitation and 530-nm emission. The fluorescence representing the total TAG content of donor vesicles was determined by disrupting donor vesicles with isopropanol and measuring the final fluorescence. Percent TAG transfer was determined by subtracting the no-protein control and dividing by total TAG content [27]. See also supplemental experimental methods and S4 Fig. Purified LprG was produced as described previously [7]. Prior to the crystallization, a three-molar excess of powdered tripalmitoyl glyceride (TAG) was directly added to LprG. The mixture was incubated on ice for 1 h. Crystals were grown in 0. 1 M sodium acetate buffer pH 4. 5 with 25% (w/v) PEG3350. Data were collected to a resolution of 1. 8 Å at the Argonne National Lab (beamline 23-ID). Processed with HKL2000, the space group was C2 and cell dimensions were a = 95. 7 Å, b = 71. 6 Å, c = 61. 8 Å, α = 90°, β = 106. 6°, γ = 90°. Structural solution was obtained by MOLREP, using the apo-LprG structure (PDB: 3MH8) as an initial model. The structure was built and refined with coot, CCP4 and PHENIX. The final R-work and R-free were 20. 4% and 23. 4%, respectively. Six- to eight-week-old mice were purchased from The Jackson Laboratory (Bar Harbor, ME). The various mouse strains and their relevant immunological characteristics are shown in S1 Table. Mice were housed under sterile conditions in an ABSL3 facility, and all animal experiments were performed under an animal protocol approved by Harvard University. Mycobacterial strains were grown in 7H9 with OADC (Middlebrook), 0. 2% glycerol and Tween 80. Approximately 1x106 cfu were administered for IV infections. All strains used for mouse infections (WT, Mut1, Mut2, Comp2) were confirmed PDIM positive by TLC. For aerosol infections cells in log-phase growth were sonicated and diluted 1: 100 in PBS, and approximately 100–300 cfu administered. For competition experiments the mutant Rv1410: : Tn (Mut1), and wild- type, H37Rv (WT), strains were mixed at a ratio of 3: 1 (Mut1: WT). WT is unmarked and Mut1 is kanamycin-resistant. For all experiments, spleens and lungs were harvested and plated for cfu on 7H10 agar or Middlebrook 7H9+1. 5% Bacto Agar (Difco) containing OADC enrichment (Middlebrook) and 0. 2% glycerol with and without kanamycin (25 μg/mL). Three to five mice per group were harvested at the indicated time points. A subset of phox-/- received the reagent aminoguanidine (AG) ad lib. For survival of rag-/- mice, 14 mice per group were infected with WT or Mut1 and followed out to five months. For survival of SCID mice, double lprG-rv1410c knockout (Mut2) and complement strains (Comp2) were generated as described (S1 Fig and S1 Methods). Five mice per group were infected by aerosol and followed for 6 months. An additional 10 SCID mice were infected with Mut2 and followed longitudinally with matched sacrifices at time of death of WT infected mice and Comp2-infected mice. At time of death complete necropsy, histopathological evaluation, and cfu enumeration from lung and spleen homogenates were performed. See also supplemental experimental methods. Mycobacteria were cultured as described above. For carbon restriction, Sauton’s minimal media containing 0. 5 g/L asparagine, 1. 0 g/L KH2PO4,2. 5 g/L Na2HPO4,50 mg/L ferric ammonium citrate, 0. 05g/L MgSO4•7H20,0. 05 g/L CaCl2,0. 01 mg/L ZnSO4 and 0. 05% Tyloxapol (vol/vol) was used with addition of either cholesterol (0. 01% w/v) or propionate (10 mM) with or without supplementation with acetate (5 mM), glycerol (0. 2% v/v), or both, as indicated. Cholesterol-containing media was prepared from 500x stocks made by dissolving cholesterol in 50: 50 ethanol/Tyloxapol. | Of the estimated 2 billion people worldwide currently infected with Mycobacterium tuberculosis (Mtb), surprisingly few go on to develop active tuberculosis (TB) disease. The vast majority, 95 percent, of infected individuals develop latent TB, remaining infected but without disease. Despite its importance in global health, the question of what determines whether an infected individual will develop active or latent TB remains largely unanswered. Changes in how Mtb grows in response to stressors presented by the host environment likely play an important role in this process. In particular, the manifold ways in which Mtb synthesizes, degrades, and transports lipids dictates its growth in an infected host. Here, we show that lipid transport is an important function of two TB genes known to be required for Mtb’s ability to cause disease in the mouse model of infection. Using a variety of genetic and biochemical techniques, we found that the products of these genes prevent the cytosolic accumulation of a lipid associated with non-growing Mtb under the metabolic conditions it encounters during infection. Our results indicate an important role for the metabolism of Mtb in its ability to orchestrate a productive infection and cause disease. | Abstract
Introduction
Results
Discussion
Materials and Methods | 2016 | Mycobacterial Metabolic Syndrome: LprG and Rv1410 Regulate Triacylglyceride Levels, Growth Rate and Virulence in Mycobacterium tuberculosis | 11,577 | 292 |
|
Cryptococcus neoformans and Cryptococcus gattii are responsible globally for almost one million cryptococcosis cases yearly, mostly in immunocompromised patients, such as those living with HIV. Infections due to C. gattii have mainly been described in tropical and subtropical regions, but its adaptation to temperate regions was crucial in the species evolution and highlighted the importance of this pathogenic yeast in the context of disease. Cryptococcus gattii molecular type VGII has come to the forefront in connection with an on-going emergence in the Pacific North West of North America. Taking into account that previous work pointed towards South America as an origin of this species, the present work aimed to assess the genetic diversity within the Brazilian C. gattii VGII population in order to gain new insights into its origin and global dispersal from the South American continent using the ISHAM consensus MLST typing scheme. Our results corroborate the finding that the Brazilian C. gattii VGII population is highly diverse. The diversity is likely due to recombination generated from sexual reproduction, as evidenced by the presence of both mating types in clinical and environmental samples. The data presented herein strongly supports the emergence of highly virulent strains from ancestors in the Northern regions of Brazil, Amazonia and the Northeast. Numerous genotypes represent a link between Brazil and other parts of the world reinforcing South America as the most likely origin of the C. gattii VGII subtypes and their subsequent global spread, including their dispersal into North America, where they caused a major emergence.
Cryptococcosis is a life-threatening mycosis with high lethality rates, especially in underdeveloped countries [1]. Infection occurs via the respiratory route by inhalation of infectious propagules (desiccated yeast cells or basidiospores) of Cryptococcus neoformans and C. gattii, frequently spreading to the central nervous system causing meningoencephalitis, with a lethality rate of up to 70% within three months after diagnosis [1,2]. C. neoformans is a cosmopolitan and primarily opportunistic agent, comprising the major molecular types VNI, VNII (VNB), VNIII and VNIV. By contrast, C. gattii infects mainly otherwise immunocompetent hosts, although a previous study suggests that some immune profile deficiency not detected by routine tests may predispose immunocompetent individuals to meningoencephalitis by C. gattii [3]. Besides the well-known outbreak in North America, C. gattii infections occur in large areas of the Amazon region and in the semi-arid Northeast region of Brazil [4,5, 6,7], being the major molecular types VGI, VGII, VGIII and VGIV. The molecular types of both species have been recently described as new species [8]. To enable a clear connection to previous published work this report maintains the two species concept with its molecular type-based nomenclature. C. gattii VGI and VGIII had been the primary cause of human and animal infections until 1999 in North America, when isolates of the molecular type VGII were reported as the cause of an outbreak affecting hundreds of healthy humans and animals in British Columbia, Canada. This outbreak lineage subsequently spread to the Pacific Northwest (PNW) of the USA in the following years [9]. Alternatively, based on one clinical case reported from the 1970s, which described a VGII isolate NIH444 from Seattle (USA), it could be suggested that the VGII outbreak lineage was already present in the temperate region several decades before its emergence on Vancouver Island [10]. However, the genotype of this isolate is very different from the Vancouver Island outbreak lineages [11] making in unlikely to be the source of the Vancouver Island outbreak. Later on, PCR-fingerprinting, Amplification Fragment Length Polymorphism (AFLP) analysis and Multilocus Sequence Typing (MLST) identified three distinct clonal lineages (subtypes) responsible for the majority of cases in the PNW [9,12]: VGIIa, the most common genotype, VGIIb, the less common [13], and VGIIc, a subsequently identified genotype with a confined geographic distribution [12]. Following this, the ISHAM working group of the International Society for Human and Animal Mycology (ISHAM) on genotyping of C. neoformans and C. gattii proposed a standardized MLST scheme, using six housekeeping genes and the IGS1 region as method of choice for strain subtyping to obtain comparable subtyping results worldwide [14]. MLST confirmed the same three major genotypes within North America [11,15]. The emergence of infections by C. gattii VGII in temperate regions initiated a pursuit of the origin of the Vancouver Island outbreak strains. One hypothesis is the occurrence of same-sex mating from an Australasian population, giving rise to a virulent genotype, which was subsequently dispersed [16,17]. However, a study using coalescence gene genealogy, phylogenetic and recombination analysis suggested that it may alternatively have emerged from a highly-recombining C. gattii population in the native rainforest of Northern Brazil, subsequently dispersed out of the original tropical area, reaching North America [18]. Similarly, two recent population genetic analyses using Single Nucleotide Polymorphism (SNP) analysis based on whole genome sequence data provided additional evidence that the PNW strains originated from South America [11,16]. Based on the above mentioned findings the present work aimed to assess the genetic variability within the Brazilian VGII population and to gain new insights related to the population structure, its origin and global dispersal from the South American continent.
One hundred and forty five Brazilian clinical and environmental isolates of the major C. gattii molecular type VGII identified by URA5-RFLP analysis [19] stored in the Culture Collection of Pathogenic Fungi, at the Oswaldo Cruz Foundation, Rio de Janeiro, and in the Research Collection of the Adolf Lutz Institute, São Paulo, Brazil were studied retrospectively. In addition to the Brazilian isolates, 42 published sequence types (STs) from Brazil and other countries, representing all previously published VGII sequence types, maintained in the MLST database (mlst. mycologylab. org), were used for comparison, in order to place the Brazilian population in an international context. For isolate information, see S1 Table. The molecular subtypes and the genetic diversity of the Brazilian C. gattii VGII isolates were investigated using the ISHAM MLST consensus scheme for C. neoformans and C. gattii [14]. Seven unlinked genetic loci were amplified, including the genes CAP59, GPD1, LAC1, PLB1, SOD1 and URA5 and the IGS1 region, using the published PCR conditions for all seven loci [14]. The sequences were manually edited using the software Sequencher 5. 3 (Gene Codes Corporation, MI, USA) and aligned using MEGA 6. 06 [20]. The allele types and the sequence types (ST) were identified via sequence alignments against the C. gattii MLST database available at http: //mlst. mycologylab. org/. The sequences of all newly identified allele types have been submitted to the C. gattii MLST database and GenBank. In order to infer the phylogenetic relationships of the isolates, the best evolutionary model for concatenated sequences of the seven loci was selected using the software jModelTest 2. 1. 7 [21,22] applying the corrected Akaike Information Criterion (AIC) and/or Bayesian information criteria (BIC). The model K80 + I + G with Ti/Tv: 3. 4548 and gamma shape 0. 4430 [23] was the best model for the concatenated dataset, which was then used in the software MEGA 6. 06 [24] to construct an unrooted Maximum Likelihood (ML) phylogenetic tree. In addition, the dataset was submitted to Neighbour Joining (NJ) analysis based on the K80 [23] model and Maximum Parsimony (MP) based on the nucleotide substitution model and using the Subtree-Pruning-Regrafting (SPR) algorithm [11]. For the ML and MP methods, all sites were included in the analysis while for NJ, all positions containing alignment gaps were eliminated. Bootstrap analysis using 1,000 replicates was used to estimate support for the identified clades of the concatenate dataset in all analysis. The minimum spanning tree using the goeBURST algorithm in the PHILOVIZ software (http: //www. phyloviz. net/wiki/) [25] was generated from concatenated sequence regions to visualize the relatedness of the C. gattii isolates with their region of origin. The diagrams show where the ST differs in the single locus variant (SLV), double locus variant (DLV), and triple locus variant (TLV), respectively. A clonal complex (CC) concept was adopted when a SLV linkage with the founder ST was found [24,25]. In order to better understand the correct number of C. gattii VGII populations (K) that were geographically homogeneous and maximally differentiated from each other, and to evaluate the presence of immigrant individuals with respect to their geographical population, we used a Bayesian statistical model [26], which calculates the membership coefficient to each of the population using the software STRUCTURE 2. 3. 4, available at http: //pritchardlab. stanford. edu/structure. html [27]. Twenty runs were performed for each value of the number of populations (K) ranging from 1 to 10. Each run consisted of Markov-chain Monte Carlo (MCMC) simulations of 1,000,000 interactions with a burn-in period of 100,000 generations. The model selected was Admixture model that takes into account the presence of migrants in the population. The actual number of K was calculated using the average and standard deviation of each K using the ad hoc statistic of the software Structure Harvester available at http: //taylor0. biology. ucla. edu/structureHarvester/ [28]. The results of the coefficients of the optimal K were graphed using the software Clumpp version 1. 1. 2 available at https: //web. stanford. edu/group/rosenberglab/clumpp. html [29] and Structure plot [30]. The software DnaSP 5. 10 (http: //www. ub. edu/dnasp/) [31] was used to analyse the haplotype diversity (Hd) and nucleotide diversities. The presence of recombination in the dataset was checked by phylogenetic compatibilities of nearby polymorphic sites along single and concatenated sequences in the software SplitsTree v. 4. 13. 1 (http: //www. splitstree. org/) [32]. Recombination events can be visualized by the formation of parallelograms between the neighbours using the reticulated algorithm NeighborNet. The Pairwise Homoplasy Index (PHY) test implemented in SplitsTree v. 4. 13. 1 and the pairwise linkage disequilibrium (D) available in the software DnaSP v. 5. 10 were also used to detect the presence of recombination. To perform the recombination analysis, the optimal molecular evolutionary model per gene was selected in the software jModelTest 2. 1. 7 as described above for the phylogenetic analysis and applied in the software SplitsTree v. 4. 13. 1. Thus, the parameters were used as follows: CAP59: K80 + I, Ti/Tv: 31. 3584, and pinv: 0. 9740; GPD1: K80, Ti/Tv: 2. 7548; IGS1: F81; FA (0. 2655), FC (0. 1533), FG (0. 2906), FT (0. 2903); LAC1: K80, Ti/Tv: 3. 0044; SOD1: K80 + I + G, Ti/Tv: 1. 8129, pinv: 0. 9360, and alpha: 0. 7400; URA5: JC; PLB1: F81, FA (0. 2336), FC (0. 2036), FG (0. 2908), and FT (0. 2725). The standard index of association (IA) is a measure of linkage disequilibrium of genotypes and/or population [33]. This test checks the null hypothesis of linkage equilibrium and p <0. 05 indicates that the null hypotheses of linkage equilibrium should be rejected, which means that the population is under clonal reproduction. In this study we applied also the standardized IA (IAS) with 10,000 randomizations available in the program LIAN 3. 5 (http: //guanine. evolbio. mpg. de/cgi-bin/lian/lian. cgi. pl) using both the parametric method and the Monte Carlo simulation for the concatenated dataset to infer the presence of linkage disequilibrium. The mating type was characterized by PCR of the pheromone genes using primers specific for MATalpha, MFalfaU (5’TTCACTGCCATCTTCACCACC 3’) in combination with MFalfaL (5’TCTAGGCGATGACACAAAGGG 3’); and for MATa JOHE9787 (5’ ACACCGCCTGTTACAATGGAC 3’) in combination with JOHE9788 (5’ CAGCGTTTGAAGATGGACTTT 3’) [34]. Amplifications of the pheromone genes MATalpha and MATa were performed independently, in a final volume of 50μL containing 50 ng of DNA, 1X PCR buffer [200 mM Tris-HCl (pH 8. 4), 500 mM KCl—Invitrogen], 0. 2 mM each of dATP, dCTP, dGTP, and dTTP (Invitrogen), 2 mM magnesium cloride, 2. 5 U Taq DNA polymerase (Invitrogen), and 50 ng of each primer. The amplification was carried out in a thermocycler (Eppendorf mastercycler gradient, California, USA) at 95°C for 3-min initial denaturation, 30 cycles at 94°C for 1 min, annealing at 57. 5°C for 1 min, extension at 72°C for 1 min, and a final extension at 72°C for 7 min. The unique fragment corresponding to each mating type was visualized after 3% agarose gel electrophoresis at 100 V.
A total of 145 C. gattii VGII isolates, including 127 clinical and 18 environmental isolates, collected between 1989 and 2010 in 4 out of the 5 Brazilian regions: 1. Northeast (n = 39), including isolates from Piauí (PI) and Bahia (BA); 2. North (n = 38), including isolates from Pará (PA), Amazonas (AM) and Roraima (RR); 3. Southeast (n = 59), including isolates from Rio de Janeiro (RJ) and São Paulo (SP); and 4. Central-West (n = 9), including isolates from Mato Grosso do Sul (MS). No VGII isolates were collected in the South region (Rio Grande do Sul, Santa Catarina and Paraná) of the country. The individual isolate data are in S1 Table. MLST analysis identified 24 allele types for the CAP59 locus, 13 for GPD1, nine for LAC1,11 for PLB1,38 for SOD1, eight for URA5 and 34 for the IGS1 region. Based on the combined analysis of the seven loci, a total of 81 sequence types were observed (Table 1, S1 Table), with 100 polymorphic sites detected in 4,186 sites analysed. The haplotype diversity (Hd) of all strains was equal to 0. 978, revealing a high genetic variability among the Brazilian C. gattii VGII strains. All Brazilian regions showed high haplotype diversity, with the highest one found in the Northeast (NE) region (Hd = 0. 981) with 31 STs, and the lowest one in the Central-West (CW) region (Hd = 0. 889) with five STs (Table 1). The genetic relationships of the obtained MLST genotypes may be separated into two main groups, the first one with 35 STs, including the VGIIa sub-genotype (ST20, major outbreak genotype on Vancouver Island), and the second one with 46 STs, including the VGIIb sub-genotype (ST7, globally present and minor outbreak genotype on Vancouver Island). No isolates of the third North American sub-genotype VGIIc (ST6) were identified in Brazil, but a closely related sequence type ST272 (strain 438BP) from MS, the CW region of Brazil was identified (Fig 1). Among the 81 sequence types identified in Brazil, 54 are represented by a single isolate. The most frequent subtype, ST40, accounts for 13 isolates found in the Central-West (CW) and Southeast (SE) regions, followed by ST20 (VGIIa) and ST5, which contained nine and seven isolates, respectively (S1 Table). In general, the different regions harboured different genotypes. The majority of the sequence types (n = 73) are unique for each of the Brazilian regions analysed. Only six sequence types were identified in more than one region (ST20, ST28, ST40, ST133, ST185 and ST287) (S1 Table, Fig 2A). The regional distribution of the STs within the regions was also evaluated with the goeBURST analysis including all 145 C. gattii isolates from this study, and another nine different Brazilian STs obtained from previously published data (S1 Table, Fig 2A). In this analysis, 9 clonal complexes (CC) were identified (e. g. 9 groups presenting SLV). Clonal complex CC 278 is composed of the clinical sequence type ST278, isolated from a patient in Piauí, ST7 isolated from clinical and environmental samples from Amazonia, and ST124, isolated from clinical and environmental samples from Piauí state. ST278 seems to play an important role in the epidemiological distribution of C. gattii due to its link with the less virulent ST7 (VGIIb). In addition, three main groups are linked to CC 278: 1) ST301 and all its descendants, mainly present in the SE region of Brazil, which is a triple-locus variant (TLV) (IGS1, PLB1, URA5 allele) of ST124; 2) ST277 and all its descendants, mainly present in the North (N) region of Brazil, which is a double-locus variant (DLV) (GPD1, PLB1) of ST7 (VGIIb); and 3) ST281 and all its descendants, a mixed group of strains from all regions of the country, which is a TLV (IGS1, PLB1, SOD1 allele) of ST278. The important role played by ST278 isolated from the semi-arid NE region was confirmed after addition of 34 STs from other countries (Fig 2B). Within the above mentioned three main groups, some representative clonal complexes can be identified: Clonal complex 40, composed of the ancestor ST40, which is the dominant ST in the SE and CW regions of Brazil, isolated from 13 clinical samples from São Paulo, Rio de Janeiro, and Mato Grosso do Sul, and its two single-locus variants (SLV) (ST325 and ST313), both isolated from clinical samples. Clonal complex 5 is represented by the ancestor ST5 and constituted of eight clinical and environmental isolates from the North of Brazil. The other three SLVs of ST5 are ST265, ST288, and ST296, all isolated from clinical and environmental samples. Clonal complex 20 is represented by five STs (ST20, ST122, ST252, ST266 and ST267), being the ST20 (VGIIa) the founder ST of this complex and composed by isolates from the SE and N (Fig 2A). In order to better understand the number of populations and their distribution throughout the country, we applied the admixture model of Structure in our dataset and identified K = 3 populations (Fig 3A). A high proportion of admixture was observed in our sample (Fig 4A). One of the populations, here presented in green, was mainly found in those States from the N/NE part of the country while the population described in blue was mainly presented in the States of the SE/CW part of the country, such as São Paulo and Mato Grosso do Sul. The third population, presented in red was found to be distributed all over the country and seems to act as an important contributor of genetic material to the remaining populations. We then compared the 87 Brazilian isolates included in the 97 South American isolates, representing all STs obtained in Brazil, with isolates recovered from different regions of the world in order to see how the Brazilian population contributed to the global C. gattii VGII distribution, detecting K = 4 number of populations (Fig 3B). In this analysis, a high proportion of admixture was also detected within the whole population of C. gattii VGII and among the South American isolates one more population was detected (here identified in yellow), which mainly derived from North America, Asia and Australia (Fig 4B). The majority of the isolates (129/145 = 89%) were identified as mating type alpha, 10% (15/145) were mating type a, including 13 clinical and two environmental isolates, and one isolate of clinical origin was mating type alpha/a (S1 Table). Random mating can be evidenced by the linkage disequilibrium (D), converging to zero. The SNPs present in the seven loci were used to detect the evidence of recombination separately. Pairwise Linkage Disequilibrium (D) between SNPs suggested at least six recombination events responsible for the polymorphism at the SOD1 locus and four at the CAP59 locus (D’ < 0. 2) (Fig 5). The other five loci showed alleles in total disequilibrium (D’ = 1). The Brazilian VGII isolates also showed evidence of recombination with a high degree of homoplasy demonstrated by a Consistency Index (CI) of 0. 27 (p<0. 05) (Table 2). In order to confirm these results we applied two other recombination tests to our dataset. The strong reticulation in the networks and phi test implemented in the SplitsTree software for the single sequences also indicate recombination within the Brazilian VGII isolates for CAP59 and SOD1 (p<0. 05) (Fig 6). These results were confirmed in the concatenated data set with an IAS value of 0. 0407 and statistically significant for recombination (p<0. 0001) in the Brazilian population.
Since the unexpected emergence of cryptococcosis caused by the VGII subtype of C. gattii in temperate North America in 1999, it has been recognized as a major agent of severe pulmonary and neurological infections in this region [reviewed in 36]. The North American cases of human and animal cryptococcosis caused by distinct highly clonal populations (VGIIa, VGIIb and VGIIc) [11,16] point to their capacity to emerge from original habitats to adapt and colonize new environments and hosts, rapidly multiplying the new adapted populations. The current study shows a high genetic variability amongst Brazilian C. gattii VGII isolates, presenting 81 MLST STs in 145 clinical and environmental isolates. In addition, a high level of haplotype diversity was observed, while also demonstrating a high degree of homoplasy, with the Consistency Index suggesting the absence of a selective genetic pressure. The patterns of the polymorphisms identified among the Brazilian strains surveyed in this study indicated a history of recombination for the genetic loci CAP59 and SOD1 (Figs 5 and 6, Table 2), which contributed to the haplotype diversity observed. The fact that both mating types were present among the clinical and environmental Brazilian VGII isolates, with 10% of them being mating type a, emphasizes that recombination events are likely to occur in Brazil, leading to the great variability/high genetic diversity observed. These findings are also reinforced by the mosaic of multiple small chromosomal chunks presented in most of the isolates studied (Fig 4). Recombination amongst VGII genotypes has also been detected previously at global [15,37] and local [38,39] scales. Although limited number of isolates have been analysed and a very limited number of sequence types have been identified [13,39,40], they already indicated the occurrence of high molecular polymorphisms in South American Cryptococcus strains. Despite the high genetic diversity in the Brazilian C. gattii VGII population, nine clonal complexes were found. Some are represented by very common and frequently recovered STs in clinical and environmental samples (e. g. ST20-VGIIa, ST40, and ST5). The persistence of successful STs, which are stable in space and time and most significant in cases of widespread adapted clones, may follow the features of clonal evolution which is defined as strongly restrained recombination [41]. This has been described for several microorganisms, in bacteria [42], protozoa [43], and fungi [44]. Linked populations have been identified as most likely being stepping stones in the global spread of VGII. Analysis of VGII in Australia [39] identified six sequence types (ST7 (VGIIb), ST38, ST5, ST21, ST33 and ST48), suggesting an introduction into Australia, which created a possible founder effect followed by a clonal expansion of the subtypes. In Thailand, the majority of the C. gattii isolates belonged to the sub-genotype VGIIb (11 out of 12) [45], suggesting again a clonal expansion of this subtype. Despite some well-adapted clonal isolates, the herein described population is recombining. The evolutionary processes, sex crossing and consequently recombination, generates new combinations of genes, some of which may increase adaptation of the population to harsh environments to increase the chance of their survival [41]. On the other hand, DNA repair is a reasonable explanation for the high rate of recombination in diploid and haploid organisms, and could be an ancestral mechanism of general sexuality [46]. As recombination acts as ancient machinery of DNA repair, which is not only related to sexual reproduction, but also associated with a fast and simple way of propagation observed in the clonal reproduction, it is an advantage in overcoming the challenges of the environment [46,47], with some of the well-adapted cells could become more virulent pathogens to humans (e. g. outbreak strains), as the killing of the host, in the case of an opportunistic Cryptococcus infection, will not interfere with fungal cell propagation. Although the Brazilian isolates do not show a very well established population structure according to the geographic origin (Fig 4), we showed that the different Brazilian regions are dominated by different genotypes (Fig 1). The six sequence types identified (Fig 2) in more than one region may reflect the Brazilian human population migration patterns, e. g. as São Paulo and Rio de Janeiro (SE) are the biggest cities in the country, many people from other regions migrate to the these regions to find greater and better work possibilities. In order to check the influence of migration of C. gattii throughout the country, which could also be due to human migration, the clone corrected dataset was submitted to the admixture model in Structure and showed one basal population distributed all over the country (presented in red in Fig 4), one mainly found in the N and NE part of the country (presented in green in Fig 4), one more frequently found in the SE/CW region (presented in blue in Fig 4). Imported cases between these populations and within each State were also found (Fig 4). The subtypes VGIIa and VGIIb, responsible for the outbreak on Vancouver Island, Canada [13,34] and the subsequent spread to the Pacific Northwest of the USA [9,40] have been identified in the North of Brazil (Fig 2). The sequence type ST20 (VGIIa) shows a large scattered distribution pattern in the Amazon region, with eight clinical isolates from the states of Pará, Amazonas and Roraima, and one environmental isolate from the state of Amazonas. In addition, three STs (ST122, ST266, ST267) linked to the clonal complex 20, mainly represented by ST20 (VGIIa), were also found in the Amazon region. The high frequency of this complex in the North may be related to a better adaptation/and or microevolution of these isolates to the environment, although one isolate of ST20 and the only isolate of ST252 were found in the city of São Paulo, which are most likely related to human migration processes. Imported cases caused by this sequence type have also been described in patients who had visited Vancouver Island from Denmark, Germany, Switzerland and the Netherlands [48,49,50] (presented in yellow in Fig 4). The sequence type ST7 (VGIIb) has been found all over the world, including: Australia, Canada, China, Korea, Thailand and the USA [9,17,39,45,51,52]. The Brazilian isolates of the sequence type ST7 (VGIIb) were found in the state of Amazonas. Besides these two outbreak associated sequence types, three additional sequence types from other countries have now been identified amongst Brazilian VGII strains, indicating further intercontinental spread as had been previously described [52,53]. These include ST5, which had been reported from Australia [39], ST19, present in Greece [18,54], and ST182, which has been found in France and China [55]. MLST analysis provided further evidence for close relationships between many Brazilian sequence types and the sequence types globally present (Fig 1). The sequence types ST7 (VGIIb), ST20 (VGIIa) and ST5, are the three sequences types identified in the current study which were also previously detected in dwelling dust samples and clinical specimens in the Amazonas state (North of Brazil), reinforcing the possibility of indoor infection, especially in wooden houses, very common in the northern part of Brazil, which was originally suggested by Brito-Santos et al. [56]. In the North and Northeast of Brazil, C. gattii behaves as an endemic fungal pathogen that causes infection in apparently healthy individuals categorized as immunocompetent patients, and the predominant VGII genotype has been recognized for at least the last 20 years among clinical and environmental strains from those large regions [57]. The results here reinforce recent findings supported by MLST and whole-genome SNP analysis indicating that the North American outbreak lineages, including the VGIIc genotype, which has only been found in the Pacific Northwest of the USA [12] but is closely related to South American strains [21], have most likely arisen from a highly recombining C. gattii population from South America, probably from the Amazon rainforest [12,16,18]. The detection of high genetic diversity amongst Brazilian C. gattii VGII isolates in the current study strongly supports the possibility of the emergence of highly virulent strains in the N and NE regions of Brazil, associated with different biotopes, one with extremely humid forest in the North (the Amazon Forest) and the other with open and predominantly dry savanna formations in Northeast (brushwood known as “caatinga”). Between them, there is a transitional region, with overlapping areas of humid forest, less humid tropical savannah (known as “cerrado”) then the dry caatinga, best observed in the states of Piauí and Maranhão. An important finding of the current study is the central role of the ST278, which is associated with a clinical isolate (CFP 243) from the state Piauí and other closely associated STs from the same area. It shifts the global origin of C. gattii VGII, which was previously placed in the Amazon region in the state Roraima (CFP 439/LMM645 from 1998), North of Brazil, by Hagen et al. [18,57] to the transitional ecological area in the Brazilian Northeast. Another very close lineage to ST278 is the ST124 from Piauí, isolated from clinical samples and decaying wood in tree hollows. One clinical sample was isolated from a case with cryptococcal meningitis and the other was isolated from the spleen of an armadillo without any evidence of disease. Thus, VGII has a potential wide host range, behaving as a multi-host pathogen. Protected microenvironments, such as tree hollows or armadillo burrows in Piauí state, probably play an important role in the C. gattii life cycle under such variable climatic conditions. These findings show the ecological adaptability of VGII to spread to new habitats, allowing it to survive in dry and humid warm or cold climate. The Brazilian Northeast and North are large geographical regions, which have been subject to extensive deforestation, leading to enormous landscape changes and the establishment of new settlements, disturbing original communities and related habitats, and causing a large-scale biodiversity loss. These habitat changes, shifts in species composition and other stress factors may affect the profile of C. gattii populations, inducing recombination events (and/or hybridization). Human-induced land use and extensive trade of native wood from the Amazon rainforest are also possible drivers of geographical dispersal of propagules, and consequently, disease emergence events. Taking into account that the present study has, like all previous studies, a possible sampling bias, over-representing some STs while others are underrepresented, it is necessary to suggest further studies investigating other ecological niches, such as a variety of human inhabited places as well as environmental samples from other tropical countries. The results indicate that the isolates from the transitional ecological area in Northeast Brazil are the most likely ancestor lineages, translocating from caatinga/cerrado by adapting progressively throughout Amazonia in South America, and spread to the North American Pacific Northwest regions and other parts of the world on multiple occasions. This picture is intrinsically related to climatic changes and devastating human activities globally. Therefore, a multifocal origin for the outbreak lineages of cryptococcal infections must be considered. | Cryptococcus neoformans and Cryptococcus gattii are fungal agents responsible globally for almost one million cryptococcosis cases yearly, mostly in immunocompromised patients, such as those living with HIV. Cryptococcosis is a life-threatening mycosis, frequently causing meningoencephalitis. Infections due to C. gattii were originally described in tropical and subtropical regions, but its adaptation to temperate regions was highlighted by the emergence in the Pacific North West of North America by C. gattii molecular type VGII. The present work aimed to assess the genetic diversity within the Brazilian C. gattii VGII population to gain new insights into its origin and global dispersal from the South American continent using the ISHAM MLST consensus typing scheme. Our results corroborate that the Brazilian C. gattii VGII population is highly diverse, and strongly supports the emergence of highly virulent strains from ancestors in the Northern regions of Brazil. Numerous genotypes represent a link between Brazil and other parts of the world, and the isolates from the transitional ecological area in Northeast Brazil are the most likely ancestor lineages, translocating from caatinga/cerrado by adapting progressively throughout Amazonia in South America, and spread to the North American Pacific Northwest region and other parts of the world on multiple occasions. | Abstract
Introduction
Methods
Results
Discussion | sequencing techniques
cryptococcus gattii
medicine and health sciences
cryptococcus
pathology and laboratory medicine
pathogens
population genetics
geographical locations
microbiology
north america
fungi
phylogenetic analysis
molecular biology techniques
population biology
fungal pathogens
research and analysis methods
sequence analysis
mycology
south america
medical microbiology
microbial pathogens
biological databases
molecular biology
brazil
molecular biology assays and analysis techniques
people and places
sequence databases
database and informatics methods
genetics
biology and life sciences
evolutionary biology
organisms | 2016 | Population Genetic Analysis Reveals a High Genetic Diversity in the Brazilian Cryptococcus gattii VGII Population and Shifts the Global Origin from the Amazon Rainforest to the Semi-arid Desert in the Northeast of Brazil | 8,191 | 320 |
Buruli ulcer (BU), a necrotizing skin infection caused by Mycobacterium ulcerans is the third most important mycobacterial disease globally after tuberculosis and leprosy in immune competent individuals. This study reports on the retrospective analyses of microbiologically confirmed Buruli ulcer (BU) cases in seventy-five health facilities in Ghana. Pathological samples were collected from BU lesions and transported either through courier services or by car directly to the laboratory. Samples were processed and analysed by IS2404 PCR, culture and Ziehl-Neelsen staining for detection of acid-fast bacilli. From 2008 to 2016, we analysed by PCR, 2,287 samples of 2,203 cases from seventy-five health facilities in seven regions of Ghana (Ashanti, Brong Ahafo, Central, Eastern, Greater Accra, Northern and Volta). The mean annual positivity rate was 46. 2% and ranged between 14. 6% and 76. 2%. The yearly positivity rates from 2008 to 2016 were 52. 3%, 76. 2%, 56. 7%, 53. 8%, 41. 2%, 41. 5%, 22. 9%, 28. 5% and 14. 6% respectively. Of the 1,020 confirmed cases, the ratio of female to male was 518 and 502 respectively. Patients who were 15 years of age and below accounted for 39. 8% of all cases. The median age was 20 years (IQR = 10–43). Ulcerative lesions were 69. 2%, nodule (9. 6%), plaque (2. 9%), oedema (2. 5%), osteomyelitis (1. 1%), ulcer/oedema (9. 5%) and ulcer/plaque (5. 2%). Lesions frequently occurred on the lower limbs (57%) followed by the upper limbs (38%), the neck and head (3%) and the least found on the abdomen (2%). Our findings show a decline in microbiological confirmed rates over the years and therefore call for intensive education on case recognition to prevent over-diagnosis as BU cases decline.
Buruli ulcer (BU), a necrotizing skin and soft tissue disease, is caused by the environmental pathogen Mycobacterium ulcerans. BU is the third most important mycobacterial disease after tuberculosis and leprosy in immunocompetent individuals [1]. Currently, BU has been reported in 33 countries worldwide, mainly with tropical climates, and more than two thirds of the global cases reported in West African countries along the gulf of Guinea particularly Côte d′Ivoire, Ghana, Benin and Cameroon [2]. The disease has variable clinical presentation based on geography; in the pacific regions BU may start as a papule, however, in West and Central Africa it may start as a painless nodule without the involvement of subcutaneous tissues [1]. The lack of pain that characterizes the initial stage of the disease pathogenesis, cultural practices and socio-economic factors [3,4] results in delay in seeking medical care [5]; early clinical forms such as the nodule gradually erode leaving a well-demarcated ulcer with wide undermined edges due to the cytopathic action of the plasmid-encoded macrolide toxin, mycolactone [6,7]. BU is classified into three categories in terms of severity: Category I, a single lesion < 5 cm in diameter, Category II, a single lesion 5–15 cm in diameter and Category III, a single lesion >15 cm in diameter, multiple lesions, critical sites, and osteomyelitis [8]. The epidemiology of BU in endemic countries is not entirely known, due to the focal distribution of cases, late reporting of cases and lack of health facilities including laboratory expertise and infrastructure for case confirmation in endemic countries of Africa. In Ghana, the first passive surveillance system reported about 1,200 BU cases between 1993 and 1998 and more than 9,000 BU cases were also reported between 2004 and 2014 [9,10]. A nation-wide active case search that was conducted in 1999 found BU in all the 10 administrative regions of Ghana with an overall prevalence of 20. 7 per 100,000 of the population [9]. Currently, BU control in Ghana is mainly through early case detection [9,11] and clinical diagnosis at peripheral health facilities designated by the National Buruli Ulcer Control Program (NBUCP) followed by laboratory confirmation and subsequent antimycobacterial therapy. Prior to 2004, surgical debridement of infected necrotic tissues and subsequent skin grafting to correct deformities were the main treatment options [12,13]. The outcome of such an invasive procedure was not certain since the extent of excision was the sole prerogative of the clinician. Furthermore, lack of surgical facilities in the BU endemic areas, high cost of surgical procedures and prolonged hospitalization after surgery lasting often more than 3 months posed as a major socio-economic burden in the affected communities and discouraged a number of patients in seeking medical treatment [14]. Based on findings from a clinical trial initiated by the World Health Organization (WHO), the recommended treatment regimen is daily oral rifampicin and intramuscular injection of streptomycin for 56 days with surgery as an adjunct for improving wound healing and correction of deformities [1,2]. The introduction of antimycobacterial therapy made laboratory confirmation of presumptive cases very critical to avoid misdiagnosis and unnecessary antibiotics administration, albeit several studies have previously reported cases that were treated on clinical grounds only but later found not to be BU but other conditions [15–17]. Nevertheless, the infrastructure and technical expertise for the gold standard method, which is polymerase chain reaction (PCR) detection of the insertion sequence IS2404, is nonexistent within the Ghana Health Service (GHS) facilities. Thus the GHS requested the Noguchi Memorial Institute for Medical Research (NMIMR) to assist in laboratory confirmation. Here we report on findings from a retrospective analysis of samples tested in our laboratory from 2008 to 2016.
Samples were collected for analysis based on the national and World Health Organization guidelines for case confirmation. The procedures for sample handling and laboratory analysis was reviewed and approved by the institutional review board of the Noguchi Memorial Institute for Medical Research (NMIMR) (Federal-wide Assurance number FWA00001824). All adult participants provided informed written consent, and a parent or guardian of any child participant provided informed written consent on the child’s behalf. The study was a retrospective one and prior to specimen collection from seventy-five selected health facilities across the country (Fig 1) designated by the National Buruli Ulcer Control Program Ghana (NBUCP). These health facilities were chosen by the NBUCP to manage BU cases across the country. The NBUCP first organized two separate workshops to build the capacity of laboratory staffs involved in the management of BU. The first workshop was conducted in 2007 in the Eastern regional capital, Koforidua and the second was conducted in the Ga West Municipal Hospital (GWMH), Amasaman of the Greater Accra region in 2008. During both workshops, health staffs comprising clinicians, nurses, laboratory staff, and diseases control officers were trained on how to appropriately collect clinical specimens from the various clinical forms of BU using swab stick and Fine Needle Aspirate (FNA) [8]. Participants were also trained on the packaging of clinical specimen and transportation under a cold chain system [8]. Also between 2009 and 2014, the Stop BU project at NMIMR as a support to the NBUCP, conducted quarterly early case search activities in these selected facilities. In addition, the Ghana Health Service also engaged courier services for the transportation of clinical specimen to the laboratory. On the average, specimens collected from within the Greater Accra region were received within 12 hours upon collection and for specimens outside Greater Accra region, the average transport time was less than 24 hours. Weekly, a team from the reference laboratory visited the two major health facilities located in the Greater Accra region (Fig 1). Samples taken prior to the team’s visit, together with newly sampled lesions were taken along for laboratory confirmation at NMIMR. Samples received usually for ulcerative lesions were two swab specimens from the undermined edges of the lesions while one fine needle aspirate (FNA) in 500 μl phosphate buffered saline (PBS) for pre-ulcerative lesions were also received. Punch and surgical biopsies were also received in some instances. All specimens were received in a well packaged specimen collection bags and most were together with the sample collection forms (BU 04 form and NMIMR laboratory specific form). The isolates confirmed as AFB positive were harvested, killed by heating at 95°C for 30 min, and used for genomic DNA extraction as previously described [23]. A 441-bp portion of mycobacterial heat shock protein 65 (hsp65) was amplified using the Mycobacterium genus-specific TB-11 5’-ACC AAC GAT GGT GTG TCC AT-3’ and TB-12 5’-CTT GTC GAA CCG CAT ACC CT- primers [24], as described previously [25]. The PCR mixture contained 5μl of a 1: 100 dilution of template DNA, 0. 25μM concentrations of each primer, 6 μl of Q-solution, 3μl of 10X buffer, 200 μM (each) dATP, dCTP, dGTP, and dTTP (Pharmacia Biotech), 1. 5Mm MgCl2, and 0. 5 U of Fire Taq polymerase, in a total volume of 100 μl. Amplification was performed using 32 cycles of 5 min at 94°C, 30 s at 94°C, 30 s at 60°C, 1 min at 72°C, and 10 min at 72°C, in an Applied Biosystems 2720 thermal cycler [18]. Amplified products were confirmed by gel electrophoresis. The amplified PCR product (40 μl) was sequenced by outsourcing. The generated sequences were edited using Codon Code Aligner 6. 0. 2 software to remove vector sequences, and species were identified by using NCBI Microbial Nucleotide BLAST, using default settings [26]. Clinical and demographic data were retrieved from all study participants using the specimen collection form and entered into Microsoft Access with validation to correct for entry errors. All statistical analyses were carried out using the Stata statistical package version 14. 2 (Stata Corp. , College Station, TX, USA). The chi square test for trend (ptrend) was explored to assess the significance of the observed decreasing proportion of laboratory cases over the years. A P-value < 0. 05 was considered significant.
A total of 2,287 clinical specimens from 2,203 BU presumptive cases were received from 75 health facilities from seven out of the ten regions of Ghana (Fig 1). Pathological specimens received were 1,892 (82. 7%) swabs, 384 FNA (16. 8%) and 11 (0. 5%) biopsies (Tables 1 and 2). The specimens were from 1,637 (74. 2%) ulcerative lesions, 250 (8. 4%) nodules, 65 (2. 9%) plaque, 42 (1. 9%) oedema and 26 (1. 2%) with osteomyelitis (Table 1). Samples collected from multiple clinical forms include: ulcer and oedema 84 (3. 8%), ulcer and plaque 77 (3. 5%) and ulcer and osteomyelitis 22 (0. 9%) as indicated in Table 1. The numbers of males were 1117 (50. 7%) and females 1086 (49. 3%). The median age was 22 years (range: 0. 3–100 years). The summary results for samples analysed by both PCR and ZN are presented in Tables 3,4 and 5. Of the total 2,203 BU cases analysed, 1020 (46. 3%) were positive for IS2404 PCR and 491 (22. 3%) were positive for smear microscopy by ZN staining. Out of the 1020 IS2404 PCR positives, 852 (38. 7%) were swabs, 160 (7. 3%) FNA and 8 (0. 8%) were biopsies. The 491 ZN positive cases comprised of 441 (89. 9%) swabs, 46 (9. 4%) FNA and 4 (0. 8%) biopsies. As indicated in Table 3, out of the 1020 confirmed by IS2404 PCR, AFB were detected among 344 (33. 7%) of these PCR positives. Further analyses showed that 645 (63. 2%) samples that were negative by ZN staining were positive by PCR (Table 3). Among IS2404 PCR confirmed cases, 502 (49. 2%) were males and 518 (50. 8%) were females (Table 1). The age range was 1–100 years with median age 20 years. In 45 (4. 4%) of the confirmed cases, age records were not available on sample collection forms. Children below 15 years of age formed 39. 8% of all cases confirmed. Comparing this lower age category (≤15 years), confirmed BU cases in each other age category was significantly lower (p<0. 001) (Fig 2). Of the 1,020 confirmed cases, 744 (72. 9%) were ulcerative lesions, 96 (9. 4%) nodules, 30 (2. 9%) plaque, 22 (2. 2%) edema and 11 (1. 1%) were osteomyelitis. Some confirmed cases presented lesions with multiple clinical forms including ulcer and edema 63 (6. 2%), ulcer and plaque 43 (4. 2%) and ulcer and osteomyelitis 11 (1. 1%). The lesions categories presented were: 270 (26. 5%) category I, 158 (15. 5%) category II and 357 (35. 0%) category III while 235 (23. 0%) had no information on lesion category. Although BU lesions were broadly distributed in all body parts, BU lesion were mostly located on lower limbs 609/1020 (59. 7%) followed by upper limbs 160/1020 (15. 6%) (Fig 3). Lesion located on lower limbs was consistently highest; 2008 [14/22 (63. 6%) ], 2009 [22/48 (45. 8%) ], 2010 [187/281 (66. 5%), 2011 [190/327 (58. 1%) ], 2012 [83/136 (61. 0%) ], 2013 [60/117 (51. 3%) ], 2014 [31/58 (53. 4%) ], 2015 [18/24 (75. 0%) ] and 2016 [4/7 (57. 1%) ]. (Fig 3) Altogether, case samples were received from seven regions in Ghana and were confirmed as BU by PCR. Among the seven regions of Ghana, cases distribution are as follows, Ashanti 74, Brong Ahafo 67, Central 27, Eastern 163, Greater Accra 668, Northern 4 and Volta 17 as shown in Table 4. The regional BU confirmation rates by PCR from our data set were Ashanti 45. 7%, Brong Ahafo 60. 9%, Central 32. 1%, Eastern 44. 0%, Greater Accra 44. 7%, Northern 80. 0% and Volta 42. 5% as indicated in (Table 4 and Fig 4). Analyses of BU cases by ZN staining method followed a similar pattern as PCR results (Table 5). Overall, a decreasing trend of proportion of confirmed BU cases (p ≤ 0. 001) was reported over the study period from (76. 8%) by the end of 2009 reaching 7/48 (14. 6%) by the end of 2016 as shown in Table 6 and Fig 5. Of the 1020 IS2404 PCR positive specimens cultured, macroscopic growth was detected for only 316 (30. 9%) after 6 months of incubation. Although ZN staining confirmed all the 316 isolates as AFB, subsequent confirmation by hsp65 sequencing identified 244 (77. 2%) as Mycobacterium ulcerans whilst the remaining were members of nontuberculous mycobacteria; Mycobacterium avium 31 (43. 1%), Mycobacterium fortuitum 28 (38. 9%), and were Mycobacterium abscessus 13 (18. 1%).
We confirmed only 46% of the total 2,203 BU cases by IS2404 PCR. Our findings suggest that over 50% of the clinically diagnosed cases may not be BU. This finding calls for the need to confirm cases before they are put on antimicrobial treatment to avert putting individuals on needless antimicrobials. Recently, we followed up on 77 cases that were historically negative for IS2404 PCR. We observed that 86. 8% of these cases wounds were completely healed and 13. 2% were partially healed without any antimycobacterial treatment [27]. Similarly, there have been reports of lesions clinically diagnosed as BU but were later confirmed as tropical phagedenic ulcer, deep fungal infection, cellulitis and diabetic ulcer [15–17,28]. It has previously been thought that diagnosis of ulcerative lesions is very straightforward due to readily recognized indolent, undermined edges lesions. However within our analyzed samples the confirmation rates of ulcerative lesions were equally low compared to other presentations of BU. Ghana introduced the current antibiotic treatment in 2006, and the policy is that all clinically recognized cases are put of SR8 without laboratory confirmation results. This practice/policy needs to be revised considering the number of reports on misdiagnosis of BU. Moreover the main bactericidal drug rifampicin (RIF) is also one of the main anti-mycobacterial agents for the treatment of tuberculosis (TB). Ghana currently has been recognized as one of the 30 most burdened TB nation due to the high HIV-TB. At the same time due to the emergence of drug resistance TB strains the use of RIF must be reduced only to needed patients. In addition, RIF is hepatotoxic [29,30] while streptomycin [31–33] is autotoxic especially to children. Considering that a significant proportion of those affected were children below the age 16 years, we propose 1) training of clinicians involved in BU diagnosis and treatment at all levels 2) that the GHS and other health services in endemic African countries revise the initial policy of antimycobacterial treatment based on clinical diagnosis alone. As an alternative, wound care practices could be employed for all clinician-diagnosed cases as an interim arrangement till microbiological confirmation is done. We also observed that nearly 40% of the PCR-positive cases were also positive by smear microscopy after ZN staining. Currently, smear microscopy has not been included BU case management in Ghana although it is employed in routine diagnosis of TB in most peripheral laboratories in the country. We are of the opinion that smear microscopy by ZN staining has to be included as a first line diagnostic tool for BU. Of the 1020 PCR positive samples that we cultured, macroscopic growth was obtained for only 30%. Sequencing analysis confirmed 77% as M. ulcerans whilst the remaining 23% were members of the nontuberculous mycobacteria. Although culture is the full proof of viable mycobacteria the cultivation challenges such as the slow growth nature which may take about 8 to12 weeks or more, the presence of fast growing microorganism that may take over cultivation media makes culture not suitable for BU diagnosis as patients would have to wait for longer period before undergoing treatment [34]. One of the theories proposed to explain the mechanism of M. ulcerans transmission is direct contact of an exposed skin with a contaminated environmental source such as sharp leaves or through pre-existing wounds [35]. Two separate studies have shown that BU lesions mostly occur where the bones are close to the skin (shins, knees, elbows and forearms) [21,36]. We observed a distribution pattern that supports the direct contact with an exposed skin hypothesis as 74. 4% of BU lesions were restricted to the limbs; lower limb (59. 7%) and upper limb (14. 7%) corroborating with several other reported studies [37–42]. In the African BU endemic regions, the hot weather conditions may be a major contributory factor for this localized lesion restriction as farmers in particular are less likely to wear protective clothes during activities to enhance efficient work output; a behavior that is likely to enhance the exposure of the skin to M. ulcerans in the environment [42]. The annual BU case confirmation rates over the years indicate that BU cases are on the decline, for example, the rates increased from 52. 4% in 2008 to 76. 2% in 2009. The improved confirmation rates observed during this period may be attributed to the prior training activities conducted by the NBUCP to healthcare givers within the Ghana Health Service facilities on case detection and proper specimen collection. By the end of 2016, BU rates had gradually declined to 14. 6%. This downward trend may be a reflection of actual reduction in BU cases or replacement of previously trained personnel with new health staff with very little or no experience in BU diagnosis. It must be emphasized that the Stop BU project at NMIMR conducted quarterly early case search activities between 2009 and 2014 when outreach activities by the NBUCP ceased. This probably might have accounted for the improved BU rates seen within those periods. In addition, due to the focal distribution nature of BU, even in endemic countries such as Ghana not all clinicians are familiar with BU. This requires routine training of clinical staff in endemic countries. We found children 15 years and below to be a major risk group and this is in agreement with findings from other countries [43,44] although a study in Benin showed that adults between 75 and 79 years are at high risk of developing BU [45]. We observed that the number of BU cases was low in children <5 years, which agrees with our recent studies in both Ghana and Cameroun that indicated less exposure of children in this age bracket to M. ulcerans. BU is known to affect both sexes, however; we observed that males below age ≤ 15 years were more affected than females (p-0. 011). The observed differences may be due to the different recreational activities engaged in by both sexes. For instance, in African endemic countries, young males are more likely to play football shirtless and move to riverside to swim thereby having more environmental contact [42]. The limitation of the study was that all case confirmation was by gel-based PCR although many reference laboratories have switched to real-time PCR and semi-automated platforms which have a lower propensity for amplicon contamination (false-positives). However, at NMIMR, we have had no experience of contamination in our laboratory as we employed a four chamber system in the analyses of our samples. Different rooms with separate biosafety cabinets are used for sample preparation and DNA extraction whilst mastermix preparation, temperate addition and finally PCR are performed also in separate rooms. In conclusion, laboratory confirmation of presumptive BU cases still remains an essential aspect in the management of Buruli ulcer in Ghana and needs to be included in the case management of BU as done for tuberculosis to avert misdiagnosis and unnecessary antibiotic treatment. While we agree that case confirmation by PCR may be a challenge for all cases, a simple differential diagnosis chart plus microscopy could also be beneficial for case recognition in remote centres. | Buruli ulcer (BU), a necrotizing skin disease caused by Mycobacterium ulcerans, is currently reported in 33 countries, with the greatest disease burden mostly in West African countries along the gulf of Guinea. The lack of pain associated with BU disease enhances delay in seeking medical treatment that could result to complications. The current existing control strategy is early case detection. Previously BU diagnosis was based solely on clinical evidence by a healthcare worker, however, since other skin conditions present similar clinical signs as BU there is the need for further laboratory diagnosis. We microbiological confirmed all clinically diagnosed cases by IS2404 PCR, and Ziehl-Neelsen. We found that over 50% of the clinically diagnosed cases were not BU, thereby averting any unnecessary antimycobacterial treatment with the associated side effects. | Abstract
Introduction
Methods
Results
Discussion | medicine and health sciences
pathology and laboratory medicine
tropical diseases
geographical locations
surgical and invasive medical procedures
bacterial diseases
signs and symptoms
ulcers
neglected tropical diseases
molecular biology techniques
bacteria
africa
research and analysis methods
research facilities
infectious diseases
buruli ulcer
artificial gene amplification and extension
lesions
actinobacteria
molecular biology
research laboratories
people and places
mycobacterium ulcerans
ghana
diagnostic medicine
polymerase chain reaction
biology and life sciences
government laboratories
organisms | 2018 | Laboratory confirmation of Buruli ulcer cases in Ghana, 2008-2016 | 5,378 | 185 |
Echinostoma caproni (Trematoda: Echinostomatidae) is an intestinal trematode that has been extensively used as experimental model to investigate the factors determining the expulsion of intestinal helminths or, in contrast, the development of chronic infections. Herein, we analyze the changes in protein expression induced by E. caproni infection in ICR mice, a host of high compatibility in which the parasites develop chronic infections. To determine the changes in protein expression, a two-dimensional DIGE approach using protein extracts from the intestine of naïve and infected mice was employed; and spots showing significant differential expression were analyzed by mass spectrometry. A total of 37 spots were identified differentially expressed in infected mice (10 were found to be over-expressed and 27 down-regulated). These proteins were related to the restoration of the intestinal epithelium and the control of homeostatic dysregulation, concomitantly with mitochondrial and cytoskeletal proteins among others. Our results suggests that changes in these processes in the ileal epithelium of ICR mice may facilitate the establishment of the parasite and the development of chronic infections. These results may serve to explain the factors determining the development of chronicity in intestinal helminth infection.
Intestinal helminth infections are among the most prevalent parasitic diseases. Recent studies have estimated that about 1 billion people are currently infected with at least one species of intestinal helminth mainly in developing regions of Asia, Africa and Latin-America [1]. Although intestinal helminths rarely kill their human hosts, they commonly cause chronic or recurrent infections that have an important impact in health. The most common symptoms are related to effects on nutrition causing growth retardation, malabsorption syndrome and vitamin deficiencies or impaired cognitive function [2,3]. Additional abnormalities such as intestinal obstruction, chronic dysentery, rectal prolapse, respiratory complications, iron-deficiency anemia or debilitating disease can also appear [4–6]. Moreover, parasitic helminth infections in livestock are responsible for significant economic losses due to decreases in animal productivity and the cost of anthelminthic treatments of parasitized individuals [1]. About 40 million people are infected with food-borne trematodes, including members of the family Echinostomatidae, mainly in East and Southeast Asia [7]. Echinostomes are cosmopolitan parasites that infect a large number of different warm-blooded hosts, both in nature and in the laboratory. About 20 species belonging to nine genera of Echinostomatidae are known to cause human infections around the world [8,9]. They constitute an important group of food-borne trematodes of public health importance with prevalences that ranges from 3% in some areas of Asia [10,11]. Apart from their interest in human health echinostomes, and particularly Echinostoma caproni, have been used for decades as experimental models to the study of food-borne trematodes—vertebrate host relationships [12,13]. E. caproni is an intestinal trematode with no tissue phases in the definitive host [13]. After infection, the metacercariae excyst in the duodenum and the juvenile worms migrate to the ileum, where they attach to the mucosa [13]. E. caproni has a wide range of definitive hosts, although its compatibility differs considerably between rodent species on the basis of worm survival and development [12]. In mice and other hosts of high compatibility, the infection becomes chronic, while in hosts of low compatibility, (such as rats) the worms are expelled from the 2–4 weeks post-infection (wpi) [14,15]. Moreover, the consequences of the infection in each host class are markedly different. The establishment of chronic infections in CD1 mice is dependent upon a local Th1 response with elevated production of IFN-γ [16]. The infection induces important inflammatory responses, a marked epithelial injury and a rapid increase of iNOS expression [15–17]. Concomitantly with these events, chronic infections impair the processes of renewal of the intestinal epithelium inducing elevated levels of crypt-cell proliferation and tissue hyperplasia at the site of the infection [18]. In contrast, the early rejection of E. caproni is associated with the development of a local Th2/Th17 phenotype and changes in the tissue structure are not observed [16,19]. Because of these characteristics, the E. caproni-rodent model is extensively used to elucidate several aspects of the host-parasite relationships in intestinal infections, such as the induction of distinct effector mechanisms and their effectiveness in parasite clearance. Comparative proteomic studies allow to obtain a broad view of the changes induced by a particular process, as in this case, the establishment of intestinal infections. Herein, we analyze the alterations in the protein expression induced by the E. caproni infections in the ileum of a host of high compatibility in which chronic infections are developed. This information may be useful to gain a better understanding on the factors that facilitate the development of chronic infections with intestinal helminthes and the consequences of helminth infections in hosts chronically infected.
The present study was performed using male CD1 mice weighing 30–35 g. The strain of E. caproni employed and the infection procedures have been described previously [20]. Briefly, encysted metacercariae of E. caproni were removed from kidneys and periacardial cavities of experimentally infected Biomphalaria glabrata snails and used for infection. A total of 16 mice were each infected by gastric gavage with 75 metacercariae of E. caproni. Additionally, 16 mice were left uninfected and used as uninfected s. All the animals were sacrificed at 2 weeks post-infection (wpi) to obtain tissue samples. The animals were maintained under conventional conditions with food and water ad libitum. This study has been approved by the Ethical Committee of Animal Welfare and Experimentation of the University of Valencia (Ref#A18348501775). Protocols adhered to Spanish (Real Decreto 53/2013) and European (2010/63/UE) regulations. Ileal sections from uninfected and infected mice were removed at necropsy and intestinal epithelial cells (IEC) were isolated as described before [21]. In brief, the ileal sections were opened longitudinally and rinsed by gentle shaking in washing buffer: ice-cold Hank’s balance salt solution (HBSS) containing 2% of heat-inactivated fetal calf serum (FCS). Supernatant was then removed and fresh washing buffer was added to the ileal sections. This step was repeated at least 4 times, until the supernatant was clear. The tissue was then cut into small 1 cm-long segments and incubated for 20 min at 37°C in HBSS containing 10% FCS, 1nM EDTA, 1mM DTT, 100 U/ml penicillin and 100 μg/ml streptomycin (dissociation buffer). The supernatant was collected and maintained on ice and the incubation was repeated a second time with fresh dissociation buffer. Supernatants were combined and filtered through a 100 nm cell strainer before IEC were pelleted out by a centrifugation at 200 g for 10 min at 4°C and washed three times in PBS under the same centrifuge conditions to remove any residual medium. Protein extraction was performed using the M-PER Mammalian Protein Extraction Reagent (Thermo Scientific) according to the manufacturer’s instructions. Shortly, M-PER Mammalian Protein Extraction Reagent was added to the IEC pellet (20: 1, v/v), mixed by vortex and incubated at room temperature (RT) for 20 min with continuous gentle agitation. The lysate was then clarified by centrifugation at 18,000 g for 15 min at 4°C, transferred into a new tube and stored at -80°C until use. In order to increase the biological significance and avoid erroneous conclusions due to individual variations, four biological replicates were performed for each experimental group (uninfected and infected). Each biological replicate was obtained by pooling the same amount (20 μg) of protein extracted from the IEC isolated from four different mice. Then, 50 μg of protein from each biological replicate were cleaned and precipitated with the 2D Clean-up Kit (GE Healthcare), pellets were resuspended in 18 μl of a proper buffer (25 mM Tris, 7 M urea, 2 M thiourea, 4% CHAPS, pH 8,5), and proteins were fluorescently tagged with CyDye DIGE Fluor minimal dyes (GE Healthcare), following manufacturer’s instructions. One microliter of dye (400 pmol) was added to each sample and maintained on ice for 30 min in the dark. The reaction was stopped by adding 1 μl of 10 mM lysine. To minimize any dye-specific labeling artefacts, two biological replicates of each experimental group (infected and uninfected) were labeled with Cy3 and the other two were labeled with Cy5. The internal standard, prepared by mixing the same amount of protein of each sample included in the experiment, was always labeled with Cy2. Ileal protein extracts from E. caproni-infected and uninfected mice were compared across four 2D-DIGE gels to identify proteins significantly modulated by the presence of the parasite. The four pairs of Cy3- and Cy5-labeled biological replicates (50 μg of protein each) were combined with a 50 μg aliquot of the Cy2-labeled internal standard. The mixtures containing 150 μg of protein were then separated in the first dimension, i. e. isoelectric focusing as second dimension were run following previously described protocols using the isoelectric focusing protocol for 24 cm Immobiline Drystrips. The IPG strips (24 cm, nonlinear pH 3–11) where rehydrated overnight with rehydration buffer (8 M urea, 4% CHAPS, 1% ampholytes and 12 μl/ml of DeStreak™), and the labeled samples were then applied to the strips by anodic cup loading, after the addition of DTT and ampholytes up to a final concentration of 65 mM and 1%, respectively. The isoelectric focusing was carried out at 20°C in the Ettan IPGphor 3 System (GE Healthcare) as follows: (I) 300 V for 4 h; (II) gradient to 1,000 V for 6 h; (III) gradient to 8,000 V for 3 h; and (IV) 8,000 V up to 32,000 Vh. Prior to the second dimension the strips were equilibrated in two steps, 15 min each, in equilibration buffer (50 mM Tris, 6 M urea, 30% glycerol and 2% SDS) containing either 2% DTT or 2. 5% iodoacetamide, respectively. The separation of proteins in the second dimension was performed on an Ettan DALTsix system (GE Healthcare) using 12. 5% polyacrylamide gels. Electrophoresis was run at 1 W/gel for 1h followed by 5 h, approximately, at 15 W/gel. Gels were scanned in a Typhoon 9400 Variable Mode Imager (GE Healthcare) at appropriate wavelengths for each fluorophore: Cy2 (488/520 nm), Cy3 (532/580 nm) and Cy5 (633/670 nm), and at 50 μm resolution. The non-essential information was removed using ImageQuant Tools software and DeCyder v7. 0 software was employed for image analysis. The differential in gel analysis module was used for automatic spot detection and abundance measurements in each individual gel, comparing the normalized volume ratio of each spot from a Cy3- or Cy5-labeled sample to the corresponding Cy2 signal from the internal standard. The data sets were collectively analyzed using the biological variation analysis module of the same software, which allows inter-gel matching and calculation of standardized average volume ratios (AVRs) for each protein spot among the 4 gels of the study. Statistical analysis was assessed for each change in AVR using Student’s t test and false discovery rate (FDR). Statistical significance was considered when p<0. 05 and q<0. 05, respectively. Moreover, inter gels matching of statistically different spots was confirmed manually. Unsupervised principal components analysis (PCA) and hierarchical clustering analysis (HCA) (Euclidean) were performed using the DeCyder extended data analysis module, both on all protein spots that were present in the 4 gels of the experiment (100% presence) and the group of spots identified as significantly modified as a consequence of the infection. These multivariate analyses clustered the individual biological replicates based on a collective comparison of expression patterns from the set of proteins chosen, with any a priori knowledge of the biological reasons for clustering. The protein spots showing greater changes in their expression levels were manually excised from the gel and washed twice with double-distilled water. Thereafter, proteins were reduced in 100mM ammonium bicarbonate containing 10 mM DTT for 30 min at 56°C, alkylated with iodoacetamide 55 mM in 100 mM ammonium bicarbonate for 20 min at RT in the dark and, finally, digested in-gel with an excess of sequencing grade trypsin (Promega) overnight at 37°C, as described before [22]. Protein digestion was stopped with 1% trifluoroacetic acid (TFA) and peptides were dried in a vacuum centrifuge and resuspended in 7 μl of 0. 1% TFA, pH 2. One microliter of peptide mixture was spotted onto a MALDI target plate and allowed to air dry at RT before adding 1 μl of matrix, a 5 mg/ml solution of α-cyano-4-hydroxy-transcinnamic acid (Sigma) in 0. 1% TFA and 70% acetonitrile (ACN), and left to air dry again. The samples were analyzed in a 5800 MALDI TOFTOF (AB Sciex) in positive reflectron mode using 3000 laser shots per position. Previously, the plate and the acquisition methods had been calibrated with 0. 5 μl of CM5 calibration mixture (AB Sciex), in 13 positions. For the MS/MS analysis, 5 of the most intense precursors were selected for each position, according to the following threshold criteria: a minimum signal‐to‐noise of 10; a minimum cluster area of 500; a maximum precursor gap of 200 ppm and a maximum fraction gap of 4. MS/MS data was acquired using the default 1kV MS/MS method. Several spots could not be identified by MALDI TOFTOF, however, so liquid chromatography and tandem mass spectrometry (LC-MS/MS) was performed. Five microliters of each sample were loaded onto a trap column: NanoLC Column, 3 μ C18-CL, 350 μm x 0. 5 mm (Eksigen) and desalted with 0. 1% TFA at 3 μl/min for 5 min. The peptides were then loaded onto an analytical column: LC Column, 3 μ C18-CL, 75 μm x 12 cm (Nikkyo), equilibrated with 5% ACN, 0. 1% formic acid (FA). Elution was carried out with gradient of 5 to 45% B in A for 15 min (A: 0. 1% FA; B: ACN, 0. 1% FA) at a constant flow rate of 300 nl/min. Peptides were analyzed in a mass spectrometer nanoESI qQTOF (5600 TripleTOF, AB Sciex). The tripleTOF was operated in information-dependent acquisition mode, in which a 0. 25-s TOF MS scan from 350–1250 m/z was performed, followed by 0. 05-s product ion scans from 100–1500 m/z on the 50 most intense 2–5 charged ions. Both MS-MS/MS and LC-MS/MS data were sent to MASCOT 2. 5 (Matrix Science) via ProteinPilot (AB Sciex) and database search was performed on NCBInr (non-redundant) database with taxonomy set to Metazoa. Searches were done with tryptic specificity, allowing one missed cleavage and a tolerance on mass measurement of 100 ppm in MS mode and 0. 8 Da for MS/MS ions. Carbamidomethylation of cysteine was used as fixed modification and oxidation of methionine and deamination of asparagine and glutamine as variable modifications. A protein identification was considered accurate when at least three peptides were identified with an overall MASCOT score greater than 50. Functional classification and intracellular localization of the identified proteins were assessed using the KEGG Pathway (http: //www. genome. jp/kegg/pathway. html) and UniProtKB resource (http: //www. uniprot. org/). A cytoscape plugin, the Biological Networks Gene Ontology (GO) tool (BiNGO 2. 3) was used to identify overrepresented biological processes GO terms [23]. Settings for BiNGO included using a hypergeometric test with a significance threshold of 0. 05. The P-values were corrected for multiple testing by the Benjamini & Hochberg correction.
A 2D-DIGE proteomic analysis was performed on whole ileal cell extracts from eight biological replicates corresponding to E. caproni-infected and uninfected mice (4 replicates each) and 2D-gel images were then subjected to computational analysis using the DeCyder software (S1 Fig). Both univariate and multivariate statistical analysis indicated that E. caproni infection induces an intense remodeling of protein expression pattern in the ileal mucosa of mice early after the establishment of the infection. The inter-gel spot matching, carried out using biological variation analysis module, revealed a total of 1,698 well defined spots with a 100% of presence, i. e. found in each gel included in the 2D-DIGE experiment (Fig 1). The average abundance of each spot was then calculated and significant changes were evaluated. A total of 876 spots, representing a 51. 6% of the total number of spots, showed significant variations within a 95% confidence interval in both Student’s t test (p<0. 05) and FDR (q<0. 05). In view of the large number of differentially expressed proteins, different selection criteria were sequentially applied in order to select for protein identification those spots whose expression was mostly affected as a consequence of the infection. In a first step, the confidence interval in the Student’s test was reduced to 99%, yielding a total of 361 spots showing a value of p<0. 01. To guarantee the proper comparison of spots among gels, the correspondence of these 361 spots among all the gels were manually validated, and 148 were unambiguously confirmed (68 up-regulated in the ileum of infected mice and 80 down-regulated). Forty-seven of these spots showed an AVR greater than or equal to |2. 00| in the four gels analyzed (17 overexpressed and 30 downregulated at 2 wpi). Finally, 37 of these spots (11 and 26 up and downregulated, respectively) could be extracted from the gel and successfully identified by MS. Fig 1 summarizes the results of applying the consecutive selection filters from the initial set of spots with 100% of presence to those that were eventually identified by MS. In order to establish the biological significance of the infection-induced protein changes, multivariate statistical tests were performed on the proteins identified by 2D-DIGE. PCA and HCA were carried out on both the total number of spots with 100% presence in the experiment (1,698) and those displaying larger significant statistical differences between uninfected and infected mice (361 spots). As shown in Fig 2, both PCA and HCA applied to the set of 1,698 spots with 100% presence were able to separate graphically the two groups of samples. In the PCA the two groups were separated in the basis of the first principal component. Moreover, all biological replicates were within the range of normality (95% of confidence), discarding the existence of outliers among the samples (Fig 2). Similarly, HCA applied to the same set of protein spots grouped the replicates in two main categories according to their condition of infected or uninfected. The heat diagram shows that protein expression patterns displayed by uninfected and infected animals were clearly different, suggesting that E. caproni infection induces a significant change on IEC (S2 Fig). In this first analysis, however, the proteins could not be clustered according to their expression pattern indicating the existence of a wide variability when the spots are compared individually (S1 Fig). This is not strange since both, significantly and non-significantly differentially expressed spots were selected for the analysis and their correspondence among all the gels had not been manually validated. As expected, when multivariate statistical tests were performed on the set of 361 spots differentially expressed between uninfected and infected mice (p<0. 01 and q<0. 05), both PCA and HCA also separated the biological replicates into two different groups. In the PCA the samples were separated by the first principal component, indicating that this set of proteins is enough to explicate the differences between the two groups (Fig 3). Similarly, HCA grouped replicates in two categories each including infected or uninfected samples (Fig 4). In this case, the protein spots were also classified in two main categories according to their expression pattern, i. e. up- or down-regulated in one group of samples respect the other one (Fig 4). This confirms that, E. caproni significantly alters the protein expression pattern of the IEC, affecting a large number of proteins 2 weeks after the infection. In view of the large number of proteins significantly affected by the E. caproni infection, those spots displaying a greater difference between uninfected and infected animals were selected for identification by MS and database search. A total of 37 from 47 spots with p<0. 01, q<0. 05 and AVR≥|2| were accurately identified: 11 of them overexpressed in the ileum of infected mice and 26 down-regulated as a consequence of the infection (S1 Fig). These 37 spots corresponded to 31 different proteins (10 up-regulated and 21 down-regulated), since 6 redundancies were detected (Table 1). This can be attributed to different post-translational modifications, the existence of isoforms or to protein modifications during sample preparation [24]. Identified proteins are classified in Table 1 according to their function, with detailed information comprising accession number, 2D-DIGE-related data, cellular role, localization and identification parameters. An analysis of the GO biological process of the proteins presenting an up-regulated or down-regulated expression was performed using the plugin BiNGO with Cytoscape. Proteins overexpressed in the ileum of infected mice were related to three main processes: Lipid and fatty acid metabolism, lipid and fatty acid transport and digestion and intestinal absorption (Fig 5). In contrast, proteins with a down-regulated expression in the intestine of infected mice were related to different biological processes such as energy and cell respiration, regulation of inflammatory responses, oxidative stress and lipid and fatty acid metabolism among others (Fig 6).
After a manual annotation of the proteins using the Uniprot database, the group of proteins that became more altered as a consequence of the infection was related to the energy metabolism. Our proteomic data suggests that mitochondrial function (particularly energy and cell respiration processes) is markedly reduced in the ileum of E. caproni-infected mice (Fig 6). Significant down-regulation of a component of pyruvate dehydrogenase complex (PDH) and the subunit α of NAD+-dependent isocitrate dehydrogenase 3 (IDH3) was detected (AVRs: -2. 03 and -2. 15, respectively). The mitochondrial PDH complex catalyzes the conversion of pyruvate to acetyl coenzyme A (acetyl-CoA), linking glycolysis to Krebs cycle. IDH3 is a mitochondrial matrix enzyme that catalyzes the rate-limiting step of the Krebs cycle, the oxidation of isocitrate to oxalosuccinate. Alteration of these processes deteriorate mitochondrial ATP production causing energy depletion [25]. In the case of an inefficient oxidative phosphorylation, the mitochondrial fat oxidation pathway becomes important in providing an alternative source of energy. The β-oxidation of fatty acids appears to be also affected in the IEC isolated from infected mice since enoyl-CoA hydratase, a mitochondrial enzyme that catalyzes the second step of each cycle of β-oxidation, was down-regulated after infection (AVR: -2. 00). Moreover, fatty acids metabolism and, consequently, energy production were affected as a result of a reduction in the carnitine biosynthetic pathway. Carnitine is required for energy metabolism since it enables activated fatty acids to enter the mitochondrial matrix. In the ileum of E. caproni-infected mice, reduced expression of 4‐trimethylaminobutyraldehyde dehydrogenase, an enzyme involved in carnitine biosynthesis that catalyzes the conversion of 4-trimethylaminobutirate to γ-butirobetaine, was also noted (AVR: -3. 39). Although the last step of carnitine biosynthesis from γ-butirobetaine occurs in the liver, precursor metabolites are absorbed in intestine and kidneys and transformed into γ-butirobetaine that is converted into carnitine [26]. Thus, reduced intestinal biosynthesis of carnitine-precursor metabolites affects mitochondrial import of fatty acids, reducing β-oxidation and favoring their cytosolic accumulation. This is supported by the fact that fatty acid binding proteins (FABPs) and apolipoprotein (Apo) A-I are overexppressed in the ileum of infected mice. In the small intestine, both liver and intestinal FABPs (LFABP and IFABP, respectively) are expressed in villus enterocytes. In the ileum of E. caproni-infected mice both IFABP (AVR: +3. 20) and LFABP (AVR: +2. 54) overexpression occurs concomitantly with mitochondrial dysfunction and down-regulation of enoyl-CoA hydratase. In IEC, FABPs have been also proposed to have a role in the regulation of intracellular levels of unbound fatty acids [27–29], which can be toxic for the cells [30]. In our study, their overexpression may be a collateral consequence of the reduced mitochondrial metabolism and the subsequent accumulation of fatty acids in the cytosol of enterocytes. Increased expression of Apo A-I, the major protein component of high-density lipoprotein (HDL), was also detected in the ileum of infected mice (AVR: +3. 45). The intestine can also act as a source of Apo A-I [31], which is incorporated to form mature chylomicrons that transport the exceeding lipids to other tissues such as adipose, cardiac or skeletal muscle [32]. All these metabolic alterations suggest that in the ileum of infected mice, enterocytes display a limited respiration and ATP production through the oxidative phosphorylation system, which can induce a metabolic shift to obtain energy from alternative metabolic processes. Indeed, parallel to mitochondrial dysfunction, lactate dehydrogenase (LDH) overexpression was detected in the ileum of infected mice (AVR: +2. 70). LDH up-regulation associated to PDH down-regulation has been previously described [33,34] and involves a crucial shift in cell metabolism to prevent pyruvate accumulation and the consequent stop of glycolysis. This shift in the cellular energy supply is a signature of oxidative stress-induced cellular senescence [35]. Since oxidative phosphorylation is a more efficient mechanism for ATP production than anaerobic metabolism, such change in energy metabolism is normally accompanied by an increase in the glycolytic flux [33–35]. Our results revealed a simultaneous increase in the expression of two different isoforms of the glycolytic enzyme enolase 1, also named α-enolase, (AVRs: +2. 00 and +2. 74, respectively). Mitochondrial dysfunction is associated with a number of medical disorders and ageing and may be a major mechanism underlying the development of mitochondria-related diseases consisting in an increase in the intracellular oxidative stress [36]. Increased production of reactive oxygen and nitrogen species (ROS and RNS, respectively) lead to an elevation in nitroxidative stress, which can oxidatively damage mitochondrial DNA, lipids and, primarily, proteins and, ultimately, induce tissue injury and cell death [36]. The establishment of chronic E. caproni infections in mice is known to be associated with the development of early and strong local inflammatory responses and tissue damage concomitantly with elevated mRNA levels of IFN-γ and iNOS [15,16]. Increased ROS production is commonly associated with failings in the mitochondrial respiratory chain that reduce effective oxidative phosphorilation and increase the leakage of electrons and the formation of reactive species [36]. According to our proteomic data two different isoforms of both α and β subunits of the electron transfer flavoprotein (ETF) were found to be down-regulated after infection (AVRs from -2. 16 to -2. 91), which is likely to be related to the infection-induced impairment of mitochondrial metabolism. Furthermore, decreased expression of the cytochrome c oxidase (CcO) subunit IV isoform 1 (IV-1) was observed in the ileum of infected mice. Although CcO activity can be regulated at several levels, the subunit IV has been shown to be a key regulatory subunit in response to ATP and O2 levels [37]. At high ATP demand CcO IV-1 can be replaced by isoform 2 (IV-2) at the expense of ROS production [38]. Thus, the down-regulation of CcO IV-1 suggests that CcO activity is regulated through the modification of the expression of subunit IV isoforms in response to the infection. The gene expression of IV-2 is induced under hypoxic and toxic condition, and is up-regulated via hypoxia inducible factor 1 alpha (HIF-1α) [38,39]. In ischemic and inflammatory diseases of the intestine, the activation of HIF-1α in epithelial cells plays a protective role through the regulation of genes involved in the maintenance of epithelial tight barrier and mucosal immune response [40–42]. HIF-1α have been shown to correlate with PDH dysfunction and have a major role in promoting the shift of cell metabolism to anaerobic glycolysis [43], which agrees with our proteomic data. Although the mechanisms leading to inflammation-mediated hypoxia are not fully understood, most likely it involves vasculitis and edema [44]. Additionally, neutrophil migration into the intestinal mucosa is critical in depleting local O2 and activating HIF-1α [42]. Neutrophil infiltration at the site of the infection is also characteristic in the high-compatible host [15,16]. Therefore, in the ileum of infected mice, inflammation and neutrophilia may lead to overexpression of HIF-1α that, in turn, contributes to shift the enterocyte metabolism and the exhibition of a senescent phenotype. All this suggests that HIF-1α can be a key mediator in regulating the metabolic changes and controlling the intestinal pathology in response to E. caproni infection in mice, which we consider to merit further attention in future studies. Finally, ROS accumulation would be favored by the down-regulation of the antioxidant enzyme manganese superoxide dismutase (MnSOD), a mitochondrial matrix and intermembrane space protein that transforms the highly reactive O2·− into H2O2 and O2. Inactivation of MnSOD gene in mouse induces mitochondrial disease associated with ROS toxicity and apoptosis [45]. Although antioxidant enzymes are generally believed to be up-regulated in response to an oxidative stress [46], a positive role for reduced expression of MnSOD in uninfected ling the homeostatic dysregulation of the intestinal tissue during E. caproni chronic infection is discussed below. Overall, the infection-induced alterations on the IEC metabolism suggest that E. caproni infection induces a rapid and intense mitochondrial dysfunction, mainly characterized by the shift of cell metabolism to an anaerobic use of glucose. These changes seem to be consequence of the oxidative stress induced by the overexpression of IFN-γ and iNOS in the intestinal mucosa of infected mice. Intestinal tissue hyper-proliferation is a hallmark response to E. caproni chronic infection [18], and the results obtained herein strongly support this observation. E. caproni infection in mice is characterized by an intense tissue damage in the ileum caused by both the parasites and the local inflammatory response developed against the infection [15–17]. Moreover, villi tip erosion and gaps in the epithelial line are common at the site of infection [15,17]. Despite these facts, tissue necrosis is not developed suggesting that epithelial restitution mechanisms work actively in an attempt to restore the constant tissue damage. Herein, we have seen that E. caproni infections in mice induce alterations in several proteins implicated in the IEC proliferation and epithelial restitution (Fig 6). Galectin 2 (Gal2) was found to be overexpressed in the ileum of infected mice (AVR: +2. 20). Galectins are a family of lectins play a major role in re-epithelialization of wounded tissues [47–51]. In addition to Gal2, an EF-hand domain containing protein (EFhd2, also named swiprosin-1) was among the most up-regulated proteins in the ileum of infected mice (AVR: +2. 36). EFhd2 is also up-regulated under inflammatory conditions [52]. EFhd2 is found together with actin and actin-binding proteins modulating bundling and cell spreading [53] and actin remodeling [54], respectively. During epithelial restitution, an extensive reorganization of the actin cytoskeleton is needed [55], suggesting that EFhd2 may play a role in this process. Both IFN-γ and nitric oxide (NO) have been shown to impair IEC migration through different mechanisms [56–58], thus Gal2 and EFhd2 appear as potential candidates to direct epithelial restitution under inflammatory conditions in the ileum of E. caproni-infected mice. Once epithelial restitution has started, augmented cell proliferation is required to provide new enterocytes to restore the damaged area and the results obtained herein reveal that several pathways are involved in the regulation of tissue hyper-proliferation. The downregulation of the antioxidant enzyme MnSOD plays a role in this process (Fig 6). Apart from its function in controlling oxidative damage, MnSOD also plays a role on tissue renewal, since MnSOD genetic deficiency promotes cell turnover [59]. In our study, this downregulation may be one of the mechanisms responsible for infection-induced cell hyper-proliferation during the establishment of chronic infections. A striking feature is that changes in MnSOD expression are commonly accompanied by increased levels of ornithine decarboxylase (ODC). ODC was not found to be among the most altered proteins. However, three isoforms of the ornithine aminotransferase (OAT), which is involved in the catabolism of L-ornithine, the substrate of ODC, were found to be markedly down-regulated (AVR: -3. 94, -2. 64 and -2. 14). Another enzyme involved in ornithine catabolism, ornithine carmaboyltransferase (OCT), was also down-regulated (AVR: -2. 26). Ornithine is synthesized from arginine (Arg) by the cytosolic isoform of the enzyme arginase and is a necessary metabolite for the synthesis of polyamines and prolines. In addition to ornithine synthesis, Arg can be also metabolized by nitric oxide synthase (NOS) to generate NO and L-citrulline, so that arginase/NOS balance is determinant to displace Arg metabolism to one or another pathway [60]. As mentioned above, E. caproni infection in CD1 mice is characterized by increased mRNA expression of iNOS [16]. Hence, in the ileal epithelium of infected mice Arg metabolism can be expected to be displaced to the production of NO and L-citrulline at the expense of ornithine synthesis. Neither arginase nor iNOS appeared to be primarily affected during E. caproni infection in mice at proteomic level. The down-regulation of ornithine catabolic enzymes (i. e. OAT and OCT) may allow the increase in ornithine bioavailability for polyamine synthesis through ODC in milieu in which ornithine biosynthesis is diminished due to the displacement of Arg metabolism. Polyamines (putrescine, spermidine and spermine) are small, polycationic, organic molecules, synthesized from ornithine via ODC, which are mandatory to cell proliferation [61]. In inflammatory models, NO production is considered to be an early phase response, whereas the production of polyamines occur in the repair-phase response after iNOS inhibition by agmatine aldehyde [62]. However, in E. caproni chronic infections, both iNOS overexpression and crypt-cell hyper-proliferation occur early and exacerbate over the course of the infection [16,18]. Thus, the increase of ornithine bioavailability for polyamine synthesis through the down-regulation of enzymes responsible for its use in other metabolic pathways may represent a different route to guarantee tissue repair in the presence of sustained elevated levels of NO production during chronic infections. We also have found increased expression of several proliferation markers, such as keratin (K19) (AVR: +2. 89), aminoacylase 1 (ACY1) (AVR: +2. 23) and sulfotransferase (SULT) 1B1 (AVR: +2. 34) in addition to two isoforms of α-enolase (AVR: +2. 74 and +2. 00). K19 is a marker of the gut morphogenesis, as it is mainly expressed in proliferative crypts [63]. Its overexpression in the ileum of infected mice is consistent with the crypt hyperplasia developed [18]. Similarly, the activity of the cytosolic enzyme ACY1, responsible for the deacylation of α-acylated amino acid residues during intracellular protein catabolism, is greater in the crypt areas than in the villous portion of small intestine [64]. Although the biological function of SULT1B is not well defined, high levels of its mRNA expression are detected in human fetal small intestine [65]. Its overexpression suggests that it play a role in cell proliferation, differentiation and/or tissue structural organization in the small intestine. Apart from its role as a glycolytic enzyme, α-enolase serves as a plasminogen receptor on the surface of a variety of cells activing plasmin [66,67]. Plasmin-enolase interactions are involved in promoting cell migration in pathophysiological processes, such as the inflammatory response, cell invasion and cancer metastasis [68,69]. mRNA expression of α-enolase increases in growing cells, but remains almost at an undetectable level in the stationery/resting/quiescent phase [70] and, at protein level, it was found to be around 2-fold in proliferating versus differentiated human keratinocytes [71]. In the gastrointestinal tissue, α-enolase overexpression has been found in Helicobacter pylori-infected gastric mucosa, both at mRNA and protein levels [70,72], as well as in ulcerative colitis [73]. In E. caproni-infected mice, the elevated expression of α-enolase in the ileal enterocytes may well be a marker of intestinal inflammation and/or tissue overproliferation. This protein was also found to be overexpressed in the ileum of E. caproni-infected rats [21] in which increased epithelial cell renewal occurs in the absence of inflammatory responses [16–18], suggesting that α-enolase may have a key role in restoring homeostasis of injured intestine. We have also found that several proteins implicated in the regulation of cell death became altered in the ileum of mice after E. caproni infection (Fig 6). In particular, proteins related to the mitochondrial-driven apoptotic pathway were affected, suggesting that the intrinsic pathway is activated because of the infection. This is of importance since in a context with increased cell proliferation, elevated levels of cell death are required to maintain tissue homeostasis and prevent massive dysregulation [74]. Moreover, prolonged inflammation and wound healing represent a high risk for DNA damage and malignant transformation and defective cells need to be rapidly eliminated [75]. Among the down-regulated proteins implicated in cell growth and apoptosis, we found MnSOD. It has been previously shown that MnSOD deficiency increases cell turnover via AP-1- p53-mediated pathways [59]. Moreover, the down-regulation of proteasome subunit alpha type 1 (PSMA1) may also affect the levels of p53. PSMA1 plays a role in gating the entry of proteins into the proteasome and its overexpression has been involved in tumor genesis [76,77]. PSMA1 is an important regulator of proteasome-mediated proteolysis, with a key role in cancer development and/or progression through modulation of p53 and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signaling, which can play a key role in the control of intestinal tissue hyperplasia and homeostatic dysregulation after E. caproni infection. Proteasome inhibition has been associated with increased intrinsic apoptosis by different mechanisms. The availability of p53 increases since this protein is degraded through the ubiquitin-proteasome pathway [78,79]. Moreover, proteasome dysfunction also affects NF-κB through the stabilization of the inhibitory subunit IκB-a [80]. PSMA1 down-regulation has been detected in colorectal cancer cells after treatment with caffeic acid phenethyl ester [81] and this has been shown to induce apoptosis of cancerous cells through the inhibition of NF-κB signaling [82,83]. As mentioned before, the decreased expression of MnSOD in a milieu of elevated oxidative stress is surprinsing. However, it has been shown that this enzyme is transcriptionally regulated by NF-κB [84]. Suppression of NF-κB translocation results in reduction of MnSOD expression leading to ROS accumulation and cell death [85]. Therefore, the down-regulation of this antioxidant enzyme observed herein may be consequence of NF-κB repression and is likely to play a positive role in the ileum of infected mice, promoting ROS-mediated programmed cell death to counteract homeostatic dysregulation. Alterations in different structural proteins were also noted. Together with the augmented expression of crypt-specific K19, a decrease in the expression of type II cytokeratin 8 (K8) was detected (AVR: -2. 87). Keratins are structural proteins that associate to form non-covalent tissue specific heteropolymers (i. e. type I and type II keratins) that build up the intermediate filament cytoskeleton of epithelial cells. K8 is the main type II keratin present in mature enterocytes from the opening of the crypts to the villi apices [86]. K8-/- mice showed a lack of intermediate filament cytoskeleton in small IEC, with differentiated enterocytes displaying progressive loss of apical membrane-associated proteins and alterations in microtubule organization [87]. Although tissue functional deficiencies were not observed in K8-/- mice, it was noted that mature IEC lacking intermediated filament cytoskeleton displayed shortened microvilli and they seem to be unable to fully recover from tissue injury [87]. Moreover, small intestine enterocytes from K8-/- mice appear to be more predisposed to apoptosis compared to those obtained from K8+/+ mice [88]. In the ileum of E. caproni-infected mice, this cytoskeletal deficiency may convert the infected epithelium even more sensitive to both parasite- and immune-mediated tissue damage, accelerating the induction of cell death. A marked downregulation of the structural nuclear protein lamin B was also observed in mice after E. caproni infection (AVR: -2. 16). Lamins, A- and B-types, are the major components of the nuclear lamina and, in addition to their role as structural proteins, these type V intermediate filament proteins contribute to nuclear envelope integrity [89]. A number of studies have linked B-type lamins to several aspects of cell physiology such as transcription, replication, spindle assembly, chromatin organization, resistance to oxidative stress or regulation of cell senenscence [90–93]. The decrease in lamin B1 expression occurs in response to stimulation of either p53 or pRB tumor suppressor pathways and induces inhibition of proliferation and premature senescence [92,93]. However, altered lamin expression is common in gastrointestinal neoplasms and reduced expression of either lamin A/C or lamin B1 is a marker of potential malignancy in the gastrointestinal tract in humans [94], thus their role in the infection with E. caproni needs to be further characterised. Finally, the expression of two isoforms of the mitochondrial elongation factor Tu (TUFM) was also found to be down-regulated in IEC from infected mice (AVR: -2. 87 and -3. 56, respectively). TUFM is one of the major mitochondrial biogenesis regulating proteins, which has been found to be down-regulated during ageing in muscle cells [95]. Moreover, its inhibition induces mitochondrial dysfunction and increased cell death in cancer cells [96], which is fully consistent with the results observed herein. Despite the fact that proteomic data strongly support the idea of elevated levels of IEC death, tissue hyperplasia develops in the ileum of E. caproni infected mice [18], thereby suggesting that mitochondrial dysfunction and premature cellular senescence are not enough to equilibrate cell proliferation and death rates. In primary cultured hepatocytes, it has been shown that high amounts of IFN-γ-induced ROS are not sufficient to induce cell death, but a combination of ROS and proper endoplasmic reticulum (ER) stress responses is required to induce apoptosis [97]. In this sense, we have found protein disulfide isomerase A3 (PDIA3, also known as ERp57 or 1,25D3-MARRS) to be down-regulated in the ileum of infected mice (AVR: -2. 27). PDIA3 is a stress-responsive protein, which is involved in protein folding, glycoprotein quality control and the assembly of the major compatibility complex class I in the ER [98]. Therefore, the lack of proper ER stress responses may be responsible for the low rate of IEC death and the development of tissue hyperplasia, despite premature senescence is induced in mature enterocytes in response to the infection. Nevertheless, in addition to the ER, PDIA3 is present in many other subcellular locations, which makes it difficult to predict the effects of its down-regulation over the course of the intestinal infection [98]. Constant wound repair represents an elevated risk for DNA damage and genomic instability in proliferating cells, promoting the development of a tumorigenic environment, with chronic inflammation being the most important risk factor [99]. Moreover, a continuous state of chronic inflammation and wound healing have been regarded as the key events for cancer development in other chronic helminth infections [100,101]. Our proteomic data suggest that both pro-tumorigenic (i. e. inflammation-mediated oxidative stress, cell hyper-proliferation) and anti-tumorigenic mechanisms (i. e. cellular senescence, apoptosis) are activated early after infection in E. caproni-infected mice. Malignant tumors are often developed at sites of chronic injury, re-epithelialization and inflammation. Thus, according to our results, persistent damage of the intestinal epithelium in long-lasting infections could represent a risk factor for cancer development. The proteomic alterations described herein can be directly associated with the chronic establishment of the parasite in hosts of high compatibility. These changes are markedly different to those observed in the ileum of infected rats, in which the parasite is rejected a few weeks after infection. A previous study [21] showed that the effects of E. caproni infection on the IEC of rats are low in comparison with mice, mainly inducing the overexpression of proteins related with the cytoskeleton and the maintenance of the functional integrity of the epithelial barrier (e. g. actin, T-plastin, both 8 and 19 cytokeratins or annexin A4). Consequently, changes on the absorptive/secretory function of enterocytes and, especially, an increased regenerative capacity of the intestinal epithelium appear to be potentially IL-13-mediated effector mechanisms involved in the early rejection of worms in rats. In contrast to mice, a strict control of proliferation and programmed cell death seems to be essential to maintain the intestinal homeostasis in rats, hence protecting the host against the injurious effects of the infection. This is consistent with the overexpression of the intestinal proliferation marker K19 and chaperones such as a heat shock cognate 71 KDa and BiP, together with the down-regulation of peroxiredoxin 3, prohibitin and 14-3-3 zeta isoform [21]. Moreover, proteomic data indicate that cellular energy metabolism becomes differentially modified in the ileum of E. caproni-infected mice and rats. Whereas in mice the intestinal infection induces mitochondrial dysfunction and an increase in the anaerobic use of glucose to yield ATP, in rats the transition to a more aerobic and oxidative metabolism is suggested, leading to a reduced glycolytic flux and overall ATP production [21]. These alterations in energy metabolism could be of relevance for a better comprehension of the mechanisms involved in the control of infections on mucosal surfaces. In summary, our results indicate that the presence of the parasite induces a rapid and profound remodeling in the protein expression pattern of IEC, associated with the development of inflammation and oxidative stress. The identification of those proteins whose expression was mainly altered indicates that the cellular processes that become primarily affected after E. caproni infection in CD1 mice are related to the restoration of the damaged intestinal epithelium and the control of homeostatic dysregulation. Wound healing and crypt-cell hyper-proliferation appear to be constitutively active processes from the early stages of the infection. Concomitantly, mitochondrial dysfunction and cytoskeletal changes indicate that cellular senescence is induced on mature enterocytes. These facts, together with the pro-apoptotic changes observed, suggest that programmed cell death is augmented in the ileal mucosa of infected mice. Although previous studies have shown that proliferation and cell death are not well balanced in the ileum of infected mice, and the IEC turnover is diminished after infection, augmented cell death may be essential to control the level of homeostatic dysregulation in the gut and eliminate potentially damaged cells, which may conduct to malignant transformation. | Intestinal helminth infections are among the most prevalent parasitic diseases and about 1 billion people are currently infected with intestinal helminths. Incidence of intestinal helminth infections is high due to both socio-economic factors that facilitates continuous re-infections and the lack of effective vaccines. In this context, further knowledge on the host-parasite relationships is required to elucidate the factors that determine the expulsion of the intestinal helminths or, in contrast, the chronic establishment of the infections. Echinostoma caproni (Trematoda) is an intestinal trematode that has been extensively used as experimental model to investigate these factors. Depending on the host species. E. caproni is rapidly rejected or develops chronic infections. Herein, we analyze the changes in protein expression induced by E. caproni infection in a host in which the parasites develop chronic infections. These data may serve to get a better understanding of the factors determining the development of chronic intestinal infections. A total of 37 spots were identified differentially expressed. These proteins were related to the restoration of the intestinal epithelium and the control of homeostatic dysregulation, mitochondrial and cytoskeletal proteins among others. This suggests that the changes in these processes in the intestinal mucosa may facilitate the development of chronic infections. | Abstract
Introduction
Material and Methods
Results
Discussion | 2015 | Altered Protein Expression in the Ileum of Mice Associated with the Development of Chronic Infections with Echinostoma caproni (Trematoda) | 12,417 | 298 |
|
The conserved internal trimeric coiled-coil of the N-heptad repeat (N-HR) of HIV-1 gp41 is transiently exposed during the fusion process by forming a pre-hairpin intermediate, thus representing an attractive target for the design of fusion inhibitors and neutralizing antibodies. In previous studies we reported a series of broadly neutralizing mini-antibodies derived from a synthetic naïve human combinatorial antibody library by panning against a mimetic of the trimeric N-HR coiled coil, followed by affinity maturation using targeted diversification of the CDR-H2 loop. Here we report crystal structures of the N-HR mimetic 5-Helix with two Fabs that represent the extremes of this series: Fab 8066 is broadly neutralizing across a wide panel of B and C type HIV-1 viruses, whereas Fab 8062 is non-neutralizing. The crystal structures reveal important differences in the conformations of the CDR-H2 loops in the complexes that propagate into other regions of the antigen-antibody interface, and suggest that both neutralization properties and affinity for the target can be attributed, at least in part, to the differences in the interactions of the CDR-H2 loops with the antigen. Furthermore, modeling of the complex of an N-HR trimer with three Fabs suggests that the CDR-H2 loop may be involved in close intermolecular contacts between neighboring antibody molecules, and that such contacts may hinder the formation of complexes between the N-HR trimer and more than one antibody molecule depending on the conformation of the bound CDR-H2 loop which is defined by its interactions with antigen. Comparison with the crystal structure of the complex of 5-Helix with another neutralizing monoclonal antibody known as D5, derived using an entirely different antibody library and panning procedure, reveals remarkable convergence in the optimal sequence and conformation of the CDR-H2 loop.
The initial steps of fusion of HIV-1 virus to host cells involve binding of the HIV-1 surface envelope (Env) glycoprotein gp120 to the primary receptor CD4 and the chemokine co-receptor CXCR4 or CCR5 [1], [2]. These binding events trigger a series of conformational changes in both gp120 and the associated Env glycoprotein gp41 that lead to the formation of a so-called pre-hairpin intermediate (PHI) of the ectodomain of gp41 [3]. In the PHI, the C-heptad repeat (C-HR; residues 623–663) and the helical coiled-coil trimer of the N-heptad repeat (N-HR, residues 542–591) do not interact with one another, but rather bridge the viral and target cell membranes. The C-terminal transmembrane region of gp41 remains inserted into the viral membrane and the N-terminal fusion peptide of gp41 is inserted into the target cell membrane [3]–[5], [2], [6]. Subsequent apposition of the trimeric N-HR coiled-coil with three C-HR' s results in the formation of a six-helix bundle (6-HB) that brings the viral and cell membranes into close proximity, eventually leading to their fusion [7]–[10]. The PHI constitutes an attractive target site for fusion inhibitors since both the N-HR and C-HR are accessible [11]–[31]. Moreover, the N-HR is highly conserved across a wide range of HIV-1 strains, and it has recently been shown that neutralizing antisera can be elicited by vaccination with a disulfide stabilized, trimeric peptide mimetic of the N-HR [32]. Recently, a number of monoclonal antibodies directed against the N-HR of gp41, many of them shown to neutralize HIV-1 to varying degrees, have been reported [33]–[40]. One such antibody, D5 [34], [41], was derived from a naïve human scFv library selected by panning against an inner core mimetic of gp41, known as 5-Helix. The 5-Helix construct comprises a single chain in which the N-HR trimeric coiled-coil is surrounded by only two C-HR helices, thereby exposing one face (comprising two N-HR helices) of the internal trimeric N-HR coiled-coil [15]. A crystal structure of D5 complexed to 5-Helix (PDB code 2CMR) reveals that one of the predominant interactions involves the complementarity determining region CDR-H2 loop of D5 protruding into the conserved hydrophobic pocket of 5-Helix [41]. In previous studies [35], [39] we reported a series of broadly neutralizing mini-antibodies derived from a synthetic human combinatorial antibody library (HuCAL GOLD [42]), comprising more than 1010 human specificities, by panning against the chimeric gp41-derived construct NCCG-gp41 [16]. The latter molecule exposes, in a stable manner, the complete N-HR internal trimeric coiled coil in the form of a disulfide-linked trimer. The parental Fab 3674 [35] was subjected to affinity maturation against the NCCG-gp41 antigen using targeted diversification of the CDR-H2 loop which resulted in significant enhancement of HIV-1 neutralization properties, both in terms of IC50 and neutralization breadth, over a standard panel of Envs from contemporary primary isolates of HIV-1 subtypes B and C [39]. Indeed, the best affinity-matured Fab in a monovalent format was comparable in its neutralization potency to the parental Fab 3674 in a bivalent format. The structural basis of such properties was, however, not previously characterized. Here we report the crystal structures of two affinity matured Fabs from this series (8066 and 8062), each complexed to 5-Helix. These Fabs represent the extremes of this series, since Fab 8066 was the most potent of the affinity matured Fabs, whereas Fab 8062 did not exhibit any detectable neutralization activity. The affinity of both Fabs for NCCG-gp41, determined by solution equilibrium titration using an electrochemiluminescence-based affinity measurement, was comparable (KD ∼15 nM) [39], although the affinity of Fab 8066 for 5-Helix measured by isothermal calorimetry (this work) is more than two orders of magnitude higher than that of Fab 8062. The crystal structures reveal important differences in the conformations of the CDR-H2 loops of the Fabs in the two complexes that provide a structural basis for the differences in neutralization properties and affinity for 5-Helix of the two Fabs.
To assess complex formation between the Fabs and 5-Helix, we carried out size exclusion chromatography with detection by multi-angle light scattering and refractive index (SMR) (Fig. 1). Stable complex formation was demonstrated by the appearance of a single peak that is retained with a mass of 72440±1014 Da (Fig. 1A, red). The peaks corresponding to Fab 8066 (Fig. 1A, black) and 5-Helix (Fig 1A, red) have masses of 49480±594 and 24680±592, respectively. The calculated masses of Fab 8066 and 5-Helix are 48896 and 24459, respectively. The affinity for 5-Helix of Fab 8066 and 8062, as well as the D5 antibody, was assessed by isothermal titration calorimetry (Fig. 1B and Table 1). The stoichiometry of binding determined by ITC is 1∶1 for the complexes of the two Fabs (8066 and 8062) with 5-Helix, in agreement with the SMR data. The D5 monoclonal antibody, however, is a bivalent full-length IgG, and hence the stoichiometry of 5-Helix to D5 is 2∶1. The KD values for the binding of Fab 8066, D5 (Fab and IgG) and Fab 8062 to 5-Helix are <<10 nM, 10–20 nM and ∼200 nM, respectively. For these three antibodies, the binding data with 5-Helix correlate with neutralization activity. The HIV-1 neutralization potency of Fab 8066 (in monovalent format) is 1. 5 to 4-fold higher than that of both the D5 IgG and Fab in an Env-pseudotyped neutralization assay using Envs from four laboratory-adapted strains of HIV-1 (HXB2, SF162 JR CSF, and 89. 6). Fab 8062, on the other hand, displays no neutralization activity towards these four strains (Table 1). It should also be noted that the bivalent format of Fab 8066 is 2-7 fold more potent than the monovalent format in terms of neutralization activity over a large panel of subtype B and C HIV-1 strains [39]. The crystal structures of Fab 8066 and 8062 complexed to 5-Helix were determined at 2. 05 and 2. 5 Å resolution, respectively (Table 2). Representative electron density maps for the CDR-H2 loops of both complexes are shown in Fig. 2. The structures of these two Fabs were compared to the previously published 2 Å resolution crystal structure of a complex of the D5 antibody with 5-Helix [41]. The 5-Helix construct can be denoted as Na-L-Ca-L-Nb-L-Cb-L-Nc, where Na, Nb and Nc are the three N-HR helices that form the internal trimeric coiled-coil of gp41, Ca and Cb are the first two C-HR helices of the 6-HB of gp41, and the L segments are 5-residue linkers; the Cc helix of the 6-HB is absent in 5-Helix [15]. In the context of the 6-HB trimer, Na and Ca, Nb and Cb, and Nc and Cc belong to three separate subunits. It should be noted that identification of the helices is different here than in the D5 complex, where the numbering of residues was incorrect by not depicting the actual connectivity of helices in 5-Helix (for details, see Figure S1 in Supporting Information S1). Since the amino acid sequences are identical among the three N helices, as well as in the two C helices, such a numbering change does not affect the nature of the intermolecular interactions between the Fabs and 5-Helix. However, reinterpretation of the connectivity of helices does change the assignment of intra- and intersubunit interactions when interpreted in the context of the 6-HB trimer. The overlapping epitope, which is recognized by both Fabs on the surface of 5-Helix (Fig. 3A), is similar to the one described in the structure of the D5/5-Helix complex. This epitope comprises three helices, Na, Nc and Ca, out of the five present in 5-Helix (Figs. 2B, C). The Na helix is located in the middle of the epitope and thus contributes residues to the interactions with five CDRs of the antibodies. Two CDRs, H1 and H2, interact with residues in the groove located between helices Nc and Na, while the other two CDRs, H3 and L3, interact with residues in the groove located between helices Na and Ca. CDR-H2 covers the groove between Na and Nc and is positioned primarily over the Na helix (Figs. 3B, C). The total accessible surface area buried at the antigen/antibody interface is comparable for the Fab 8066 and 8062 complexes with 5-Helix (∼2300 Å2) and divided approximately equally between antigen and antibody. This area is ∼250 Å2 greater than for the D5/5-Helix complex. The interactions with the heavy chain CDRs bury about twice as much accessible surface as those with the light chain CDR' s. The number of residues at the interface (defined as residues whose accessible surface area decreases by >1 Å2 upon complexation [43]) is also similar for the Fab 8066 and 8062 complexes (∼60 residues subdivided equally between antibody and antigen) and somewhat larger than for the D5 complex (53 residues, 28 from the antibody and 25 from 5-Helix). Despite the overall similarity of the three antibody or Fab complexes with 5-Helix, there are significant differences in the conformations of the corresponding CDRs, as well as in the conformation of the interacting residues within the antibody-antigen interface. These differences are strongly correlated with the variations in the sequences of the CDRs. The amino acid sequences of all CDRs other than CDR-H2 differ markedly between D5 and Fab 8066, leading to obvious structural differences in these segments. Fortuitously, the CDR-H2 of Fab 8066 and D5 converged to highly homologous sequences in the selection procedures used to generate these antibodies (Fig. 3E), leading to similarities in the conformation of this part of their structures. By way of contrast, the CDR-H2 sequences of Fabs 8066 and 8062 diverged during affinity maturation (Fig. 3E), resulting in significant differences in their conformations. These conformational differences, as well as the altered interaction patterns, may be reflected in different biological properties of these three antibodies, as manifested by the binding and neutralization data shown in Table 1. The conformations of CDRs H1, H3, and L3 of Fabs 8066 and 8062 in the complexes with 5-Helix and their interactions with the antigen are most similar. CDRs H3 and L3 interact with residues from Na and Ca helices, and therefore are positioned in a shallow groove between these two helices on the surface of 6-HB (Fig. 3B). Detailed analysis of the interactions between the CDRs and the antigen (Table S1 in Supporting Information S1) indicates that, prior to affinity maturation of CDR-H2, CDR-L3 and CDR-H3 carried most of the interface interactions. The loops in both CDR-L3 and CRD-H3 are shorter in the D5 antibody (by 2 residues each), accounting for the rigid body movement of 5-Helix towards the antibody, as well as for the reduced number of the antigen-antibody interactions in the D5 complex, compared to the two Fab complexes. The effect of the differences in length of CDR-L3 and CDR-H3 is illustrated in Fig. 3D, where all three complexes are superimposed based on the Cα coordinates of the antibodies. The conformation adopted by CDR-H1 in the complexes of Fabs 8066 and 8062 with 5-Helix is very similar. Two residues of CDR-H1, Ser-31 and Ala-33, interact with His-23 and Leu-27 of helix Na of 5-Helix, and their mode of interactions is essentially the same in both complexes (Fig. 4) (Italics are used throughout for residues of 5-Helix.) Although the side chain of Ser-31 is found in two orientations in the Fab 8066 complex, compared to a single conformation in the Fab 8062 complex, a long hydrogen bond between its carbonyl oxygen and the Nε2 atom of His-23 is present in both complexes. The other contacts involving CDR-H1 are hydrophobic. When compared to the D5 complex, an additional hydrophobic contact is made between Ala-33 and Trp-30, due to the different orientation of the latter' s side chain (Fig. 4 and Table S1 in Supporting Information S1). Similarly to CDR-H1, CDR-H2 interacts with residues located in a groove between helices Nc and Na of 5-Helix (Fig. 3C). CDR-H2 in the antibody/5-Helix complexes effectively substitutes for the N-terminal end of the Cc helix that is present in the six-helix bundle of complete gp41, but is absent in 5-Helix. The significance of the interactions made by CDR-H2 residues might be attributed to the unique orientation of this loop, which lies parallel to the surface of the epitope and covers the groove like a lid (Fig. 3C). This mode of interaction distinguishes CDR-H2 from the other CDRs which are all oriented approximately perpendicular to the interface. Because of the unique position of CDR-H2, every residue of this loop is involved in the interactions with the antigen. The amino acid sequence of the CDR-H2 loop converged during selection to be almost identical in Fab 8066 and the D5 antibody, with the exception of a single residue (Thr-57 in Fab 8066 is an Ala in D5), and therefore the interactions of CDR-H2 with 5-Helix are very similar in both complexes (Fig. 5A). The interactions involving CDR-H2 are predominantly hydrophobic, and the few polar interactions that are present vary between the two complexes (Table S1 in Supporting Information S1). Two such polar interactions are unique to the Fab 8066 complex, and involve a short hydrogen bond between the side chains of Thr-57 of CDR-H2 and Arg-38 of the Na helix, as well as a hydrogen bond between the side chains of Asn-58 and Gln-34 (Fig. 5A). In the D5 complex, Asn-58 is forced to adopt a different conformation due to the close proximity of the bulky side chain of Tyr-94 of CDR-L3. The structural equivalent of the latter residue in Fab 8066 is the much smaller Val-95 (Fig. 5A). In the case of Fabs 8066 and 8062, the sequence of CDR-H2 diverged during affinity maturation (Fig. 3E), leading to a different conformation of this loop, as well as to variations of the interaction pattern in the two complexes (Fig. 5B). Nevertheless, a majority of the hydrophobic interactions are preserved for those residues of CDR-H2 that are either identical or of a similar type (for example, Ile/Leu-53) in both Fabs. Some of the polar interactions in the Fab 8062 complex are lost, and compensated by additional hydrophobic interactions (Table S1 in Supporting Information S1). For example, a hydrogen bond between Asn-58 of CDR-H2 and Gln-34 of the Na helix, present in the Fab 8066 complex, is substituted by hydrophobic interactions between the side chain of Val-58 of Fab 8062 and the same residue of 5-Helix (Gln-34). However, the hydrogen bond between Thr-57 of CDR-H2 and Arg-38 of the Na helix in the Fab 8066 complex has no equivalent in the Fab 8062 complex due to the Thr-57Ala substitution in the latter. A hydrogen bond between the carbonyl oxygen of Thr/Ala-57 and the side chain of Arg-38 is present in both complexes (Fig. 5B). A hydrogen bond formed between the side chains of Thr-56 and Gln-34 of the Na helix in Fab 8066 complex is not found in the Fab 8062 complex, since the latter has a Phe in the corresponding position. In turn, the large Phe side chain is involved in extensive hydrophobic interactions between Fab 8062 and 5-Helix. As a result, the number of contacts between the two Fabs and 5-Helix is comparable. Nevertheless, when the topography of the CDR-H2 loops within two Fab complexes is reviewed, a significant difference can be noticed (Fig. 5C). Substitution of Thr-56 in Fab 8066 by Phe-56 in Fab-8062 results in shifting of the whole segment of the CDR-H2 loop out of the groove on the surface of the bundle, where it was docked in the Fab 8066 complex. Thus, while the hydrophobic interactions between Phe-56 and 5-Helix are preserved, including an unfavorably short one (3. 2 Å) with Gln-212 of the Nc helix, this substitution results in displacement of the main chain surrounding Phe-56 away from 5-Helix. The CDR-H3 sequences are identical in Fabs 8066 and 8062, and these loops maintain very similar conformations in the complexes with 5-Helix. As noted above, residues of CDR-H3 interact with the residues in the groove between the Na and Ca helices of 5-Helix, maintaining very similar contacts, with only a slight variation in the distances between the same pairs of atoms (Fig. 6). The interactions are predominantly of a hydrophobic nature, with only a few polar interactions. One strong hydrogen bond between the hydroxyl of Tyr-102 and His-65 (2. 7 Å) in the Fab 8066 complex is lost in the Fab 8062 complex, due to different orientation of the side chain of His-65 in the two complexes. His-65 has two alternate conformations in the Fab-8066 complex, but only a single one in the Fab 8062 complex, most likely due to the differences in the conformation of CDR-L1 in the two complexes. The hydrogen bond between Tyr-102 and His-65 in the Fab 8066 complex is formed with the histidine side chain adopting the conformation that is absent in the Fab 8062 complex. The number of contacts involving interactions of CDR-L1 and CDR-L2 with 5-Helix is much lower than for the other CDRs discussed above. The conformation of CDR-L1 is different in each of the three antibody complexes compared here. However, as the CDR-L1 sequences are identical in Fabs 8066 and 8062, their interactions with 5-Helix are similar, with differences attributed to crystal contacts present in the Fab 8066 complex (Fig. 7A). CDR-L1 maintains the interactions on the periphery of the epitope, mostly with residues from helix Ca (Fig. 3B). Only two residues with bulky side chains, Trp-30 and Lys-33 from helix Na, extend to CRD-L1 and interact with the side chain of Tyr-31, and these intermolecular contacts are shorter in the Fab 8066 complex (Fig. 7A). The side chains of Glu-30 adopt different conformations in the two complexes. In the Fab 8066 complex Glu-30 is hydrogen bonded to Ser-62, whereas in the Fab 8062 complex this residue forms an ion pair with His-65. One additional hydrophobic interaction in the Fab 8066 complex that is not present in the Fab 8062 complex is found between Pro-28 and the Cγ atom of Asn-58; in the Fab 8062 complex Pro-28 is oriented away from the bundle. Although CDR-L1 in the D5 complex recognizes the same part of the 5-Helix epitope as the corresponding CDRs in the other two complexes, its conformation differs from both of them. Such different conformation of CDR-L1 in the D5 complex is due not only to the unique sequence of this loop, but also to the presence in the interface of a glycerol molecule from the cryoprotectant (Fig. 7B), which mediates antigen-antibody interactions [41]. The conformation of CDR-L2 is similar in all three complexes, except that in the Fab 8066 complex the tip of the CDR-L2 loop is oriented more towards 5-Helix then in the other two complexes. This allows formation of two solvent-mediated interactions with His-65 and Glu-69 of helix Ca (Fig. 8). In the Fab 8062 complex, CDR-L2 is not involved in any contacts with the antigen. The conformation of CDR-L3 is very similar in the Fab 8066 and 8062 complexes. The tip of this loop is oriented into the groove between helices Na and Ca (Fig. 8). In the Fab 8066 complex, two residues on the tip of CDR-L3, Ser-92 and Met-93, interact with two residues from the helix Ca, Met-51 and Asn-58, respectively. No equivalent interactions are seen in the Fab 8062 complex, but several other hydrophobic contacts are similar in both complexes. When the first 6-HB structures of the gp41 ectodomain emerged [7]–[10], key hydrophobic interactions between residues located on the N-terminus of a C-HR helix, and the C-termini of two adjacent N-HR helices in the trimer, were described. Several residues, such as Trp-628, Trp-631 and Ile-635 on the C-HR helix and Ile-573 and Val-570 (Ile-208 and Val-205 in the Fab 8066 and 8062 complexes, respectively) on the N-HR helix from the same molecule of gp41, maintain numerous intramolecular hydrophobic interactions (numbering for the Cc residues is taken from the 1AIK structure [7] and corresponds to the numbering of the native gp160 sequence). These interactions were identified to play a crucial role for the stability of the 6-HB and hence for successful initiation and completion of the fusion process [7]–[10]. When complexed to 5-Helix, the CDR-H2 and CDR-H1 of Fabs 8066 and 8062, as well as D5, occupy the space taken by N-terminal end of the C-HR helix (Cc) in the 6-HB structure. Indeed, the segment of the CDR-H2 in Fab 8066 comprising residues 51–58 literally takes the place of the position of the N-terminus of the third C-HR helix in the gp41 trimer. In particular, the side chains of Phe-54, Ile-53 and Thr-56 of CDR-H2 preserve the interactions with Ile-208 and Val-205 of the Nc helix, substituting for the intrasubunit interactions between the Nc and Cc helices of the 6-HB (Fig. 9A). Interestingly, the conformations of the hydrophobic residues on the Nc and Na helices are not significantly affected by the binding of the antibody, and the topography of the surface between the two adjacent N-HR helices, Nc and Na, required for successful docking of the Cc helix seems to be unperturbed with only two exceptions, namely the side chains of His-23 and Trp-30 (His-564 and Trp-571 in gp41). In the case of His-23, both the χ1 and χ2 angles adopt different rotamers, with the result that the hydrogen bond between the Nδ atom of His-23 of the Na helix and the hydroxyl of Tyr-638 of the Cc helix in the 6-HB is replaced by a hydrogen bond between the Nε atom of His-23 and the carbonyl of Ser-31 of CDR-H1 in the Fab 8066/5-Helix complex (Fig. 9A). Likewise, the side chain of Trp-30 in the Fab/antibody complexes with 5-Helix also changes its conformation relative to that found in the 6-HB: for the Fab 8066 and 8062 complexes there is a ∼180° flip in the χ2 angle, while for the D5 complex both the χ1 and χ2 rotamers are altered. In the absence of Fab/antibody the intersubunit hydrophobic interactions between Trp-30 of Na (Trp-571 of a N-HR helix of gp41) and two tryptophans (Trp-628 and Trp-631) of a C-HR helix are strictly conserved within all 6-HB and 5-Helix structures solved to date (Fig. 9B). (In 5-Helix these interactions occur between helices Nb and Ca and between Nc and Cb). It seems reasonable to suggest that the formation of such an intersubunit “tryptophan lock” serves to optimally align the hydrophobic moieties on a given C-HR helix and two adjacent N-HR helices to form a hydrophobic zipper along the axis of the bundle. In the context of HIV-1 neutralization, binding of Fabs 8066 and 8062 and the D5 antibody compete with the formation of the intersubunit “tryptophan lock”. Hence, even subtle differences in the stability of the complexes with various antibodies under conditions of ongoing competition can determine the success in blocking productive 6-Helix bundle formation vital for viral fusion. The crystal structures of the Fab complexes with 5-Helix presented here involve a single antibody molecule substituting for one C-HR helix. Apposition of the viral and target membranes requires the conversion of the PHP in which the N-HR trimer is exposed to a 6-HB in which the C-HR helices are bound to the outside of the N-HR trimer. Success in the competition between antibody and C-HR helices for binding to the exposed N-HR trimer of the PHP may depend on the ability of the antibody to form a complex with more than one antibody molecule bound. In principle, up to three antibodies could bind to three equivalent sites on the N-HR trimer. Indeed, maximum antagonism between Fab 8066 and the N-HR trimer mimetic NCCG-gp41 in the HIV-1 neutralization assay is observed at a molar ratio of 3∶1 [39]: at this ratio the concentrations of free NCCG-gp41 (which has 3 Fab binding sites) available to bind the C-HR of gp41 and free Fab 8066 available to bind the N-HR trimer of gp41 are at a minimum. To evaluate the impact of the observed structural differences in the interactions of the CDR-H2′s of the different Fabs with 5-Helix, we modeled complexes of the three Fabs bound to a N-HR trimer. Modeling is complicated by the fact that whereas the gp41 trimer is formed from three identical N-HR helices, only two of the three N-HR helices, Na and Nc, interact with the Fab in the crystal structures of the complexes with 5-Helix. As a result, the symmetry of the N-HR trimer is broken in the crystal structures. To generate models of the N-HR trimer with three Fabs we used as reference structures two coordinate sets comprising three N-HR helices (Na, Nb, and Nc) and the Fab, which were extracted from the two crystal structures of 5-Helix bound to single Fab molecule, 8066 or 8062. The fragment of the each reference structure, containing helices Na and Nc, that interact with the Fab in the 5-Helix complex, plus the Fab molecule itself, was treated as a rigid body unit to create a model of a trimer with three Fabs bound. We initially modeled the binding of two additional Fab molecules by superimposing two N-HR helices of the rigid body unit (Na and Nc) with two other helices of the trimer (first, with Nc and Nb, and, subsequently with Nb and Na). The aim of this approach was to preserve the intrinsic structure of the experimentally observed gp41 trimer. The results are shown in Fig. 10. The two modeled Fab 8066 molecules are reasonably well positioned in a pseudo-symmetric fashion between pairs of N-HR helices in the trimer, while two areas of steric clashes are observed for Fab 8062. The first involves slightly short contacts between CDR-L3 of one Fab and the CDR-H2 of another (2. 2 Å between the side chain of Met-93 and the carbonyl oxygen of Phe-54, respectively). The second area displays short contacts for the interactions between CDR-H2 and helix Nb of the trimer (1. 7 Å between the side chains of CDR-H2 Phe-54 and Gln-124 of Nb). A closer look at the results of the modeling based on the superposition of two pairs of helices explains the latter as follows. The three N-HR helices are involved in very different types of interactions in the 5-Helix complexes. The Nb helix interacts solely with C-HR helices on both sides, while the Na and Nc helices each interact on one side with a C-HR helix and on the other side with a Fab molecule. When the Na and Nc helices of the rigid body unit are superimposed on two other pairs of helices from a trimer, a shift in the resulting positions of that pair of helices is observed in the modeled structure, compared to their position in the original structure of the 5-Helix complex. This shift originates from the changes in the environment of each helix in the model, since all three N-HR helices are involved in a similar type of interactions with two Fabs on both sides. However, since helix Nb was used twice during model building to create two Fabs on either side of it, it has two distinct positions in the modeled structure, as can be clearly seen in Fig. 10, panel Nb. Thus, to accommodate a Fab on the left side of helix Nb, helix Nb has to move to the right compared to its position in the gp41 trimer, and vice versa. The ambiguity of the positioning of helix Nb in the model demonstrates the shortcomings of this approach. We conclude that to build a model with three Fabs we should not constrain the structure of the helical trimer. Nevertheless, results of such modeling consistently indicate that the shifts in all three helices are larger when bound to Fab 8062 than to Fab 8066, suggesting that that formation of the complex with three Fab 8062 molecules requires more extensive structural rearrangements. To address problems with the first model, we used an alternative approach in which only one helix out of two comprising the rigid body unit is used for sequential superpositions. In this manner, the second helix will be placed where it falls, and for the next round its new position will be used in the superposition for the purpose of creating the second modeled Fab. We fully realize the imperfections of such an approach, which will allow the trimer structure to relax without any restrictions applied. The advantage, however, is that the positions of each pair of helices is consistent with the presence of the Fab molecule bound between them. The resulting models are then only used for comparisons, although they are not considered to be accurate in absolute terms. Fig. 11A shows the superposition of models of N-HR trimer- (Fab) 3 complexes with Fabs 8066,8062, and D5. Since we started to build the models for each Fab from the superimposed crystal structures, for clarity they are not shown. Similarly to the first model, the shift in the position of the third Fab is most pronounced for Fab 8062, less so for D5, and least for Fab 8066. This reflects the extent of rearrangements in the trimer upon increasing the ratio of Fabs to gp41 in the complexes. This is also confirmed by comparing the shifts of the helices for individual Fab complexes between the original and modeled structures (Fig. 11 G, H, I). The similarities of the first and the second models are extended to the areas with steric clashes at the interface between neighboring Fabs (Fig. 11B). The first area of steric clash involves the CDR-H2 of one Fab and the CDR-L3 of another (Fig. 11C), and the second clash area involves loop 71-77 of one Fab and the CDR-L1 of the other (Fig. 11D). Close contacts are only observed in the first area for Fab 8062 (between Gly-55 and Met-93, Fig 11C), whereas no bad intermolecular contacts are present for Fab 8066, owing to the different conformation of the CDR-H2. Since CRD-L3 is shorter in D5 than in the two others Fabs, it is also not involved in any collisions. Interestingly, comparison with the neutralizing antibodies DN9 (from a human B-cell library from an HIV infected patient, selected using the engineered disulfide stabilized N-HR trimer, N35CCG-N13), and 8K8 (raised in rabbits using N35CCG-N13) reveals a high level of sequence similarity of their CDR-H2′s with that of Fab 8066 [38]. In DN9, a residue with a longer side chain (Glu) occupies the position of Gly-55 (Table S2 in Supporting Information S1), but position 56 is occupied, as in Fab 8066, by Thr, which should allow avoidance of the collision described above. In contrast, position 56 in 8K8 is occupied by a long Arg side chain, but position 55 is still a Gly. In the second area, short contacts are present in the trimer models of both Fab 8062 and D5, involving Ser-74 in both, as well as Asp-25 and Glu-27 of CDR-L1 in Fab 8062 and D5, respectively (Fig. 11D). The contacts between the neighboring Fabs in the model of the N-HR trimer- (Fab 8066) 3 complex seem to be much better. There are no bad contacts between the Fabs and the individual helices in the N-HR trimer of the second model, because the positions of the helices were shifting as the model was being built to accommodate the Fabs.
Although Fab 8066 and the D5 monoclonal antibody were developed using completely different naïve antibody libraries with different panning and selection procedures [34], [35], [39], the interactions between their CDR-H2 loops and the hydrophobic pocket of 5-Helix converged to similar solutions (Figs. 2E and 5A). Affinity maturation of the CDR-H2 conveyed significant improvements in neutralization breadth and potency of Fab 8066 relative to its parental Fab 3674 [39], indicating that interactions with this region of gp41 are crucial for efficient neutralization of HIV-1. Examination of the CDR-H2 sequences of a series of 10 affinity matured Fabs generated from the parental Fab 3674 [39], as well as the CDR-H2 sequences of D5 [41], DN9 and 8K8 [38] reveal that the key determinants relating to neutralization activity are correlated with the nature of the residues at positions 53,54 and 56. In our series of affinity matured Fabs, 3 Fabs were broadly neutralizing (Fabs 8060,8066 and 8068), 3 Fabs (8059,8064 and 8069) as well as the parental Fab (3674) displayed weak neutralizing activity, and 4 Fabs (8061,8062,8063 and 8065) had no detectable neutralization activity. The neutralizing Fabs were all characterized by a Leu or Ile at position 53 with the exception of Fab 8064 which had a Trp; a Phe at position 54; and a polar residue at position 56. The presence of some of these key residues in CDR-H2 sequences, when reviewed in the framework of structural and modeling data presented in this study, seem to correlate with their affinity and biological activity, as shown by a few examples below. Fabs 8064 and 8065 are of interest since they display very different biological activity (neutralizing versus non-neutralizing, respectively), their affinities for 5-Helix differ by a factor of ∼30 (KD ∼5–10 nM versus ∼200 nM; unpublished data), and their CDR-H2 sequences differ at only two positions (Phe versus His at position 54, and Ser versus Gly at position 56). Both Fabs 8064 and 8065 have a bulky Trp at position 53 which likely necessitates some conformational changes, probably propagating to CDR-H1 and CDR-L1, to circumvent steric clash (Fig. 12A). However, the presence of a His at position 53 in Fab 8065 creates an unfavorable environment by introducing a polar residue into a hydrophobic pocket formed by Leu-27 and Thr-28 of helix Na and Val-205 and Ile-208 of helix Nc. Fab 8069 also exhibits some unusual features since it differs from Fab 8062 by a single residue with a larger Trp substituting for a Phe at position 56. The presence of the bulky Phe-56 in Fab 8062 induced the changes in the conformation of CDR-H2 that were correlated in this study to the lack of biological activity of Fab 8062. Nevertheless, Fab 8069 has two-fold higher affinity for the N-HR mimetic NCCG-gp41 than Fab 8066 (∼7 versus ∼15 nM; [39]), a KD of 10-20 nM for binding to 5-Helix (unpublished data), and weak neutralization activity, roughly comparable to that of the parental Fab 3674 [39]. Modeling of the CDR-H2 of Fab 8069 based on that of the Fab 8062 complex shows that simple replacement of Phe-56 by a Trp results in numerous short contacts with residues of the Nc helix (Fig. 12B). Thus accommodation of a Trp at position 56 requires some structural rearrangements to occur within the gp41 bundle. However, since plasticity of the isolated N-HR trimer is likely to be greater than that of 5-Helix, such rearrangement could be readily achieved. While affecting the overall structure of a trimer, modeling suggests that, when bound, Trp-56 makes multiple tight contacts with an N-HR helix. This is reflected in the high affinity of Fab 8069 for the N-HR mimetic NCCG-gp41 [39]. However, shifts in the N-HR helices required to accommodate Trp-56 may prevent binding of more than one antibody molecule. This is in agreement with neutralization studies carried out in the presence of NCCG-gp41 which failed to demonstrate maximum antagonism at a molar ratio of Fab 8069 to NCCG-gp41 of 3∶1 (unpublished data), in contrast to the results with Fab 8066 [39] and even Fab 8064 (unpublished data) where maximum antagonism was observed at a 3∶1 molar ratio. Based on the HIV-1/SIVcpz protein alignment from the 2009 Los Alamos Sequence Compendium [44], the N-HR epitope recognized by the two Fabs and the D5 antibody is highly conserved. The same or a closely overlapping epitope is also recognized by there other neutralizing antibodies directed against the N-HR trimer of gp41, namely DN9 and 8K8 [38]and HK20 [33]. Using the native gp160 numbering, the epitope comprises residues 560,563–565,567,568,570–577 and 579 of the N-HR (For the corresponding numbers of the residues in 5-helix in the structures of the complex, see Figure S1 in Supporting Information S1). Of these 15 residues, eight are absolutely conserved, and all substitutions involving the remaining seven residues are highly conservative (E560N/Q, Q563R, H564Q/R, L565M, I573V, Q577R and R579Q). Ten N-HR residues (656,570–577 and 579) interact with the CDR-H2, of which seven are absolutely conserved. In contrast, much more variability is seen for the seven C-HR residues within the epitope: only one residue is absolutely conserved (E647), three are subject to conservative substitutions (D632E, H643Y and I646L), and three (residues 629,636 and 640) are not conserved at all. Only two C-HR residues, both of which are subject to conservative substitutions, interact with the heavy chain (CDR-H3). None of the C-HR residues that contact CDR-L1 are conserved, and two out of three that contact CDR-L2 are also not conserved. The two residues (E647 and H643Y) contacting CDR-L2, however, are highly preserved. Thus, one can conclude that the broadly neutralizing properties of Fab 8066 largely originate from the conservation of the N-HR epitope across different strains and clades of HIV-1. The remarkable difference in the neutralization activities of Fabs 8066 and 8062 [39], despite relatively small differences in the interactions with the antigen when compared with the D5 complex, would also suggest that adapting a productive conformation while binding to the pre-hairpin intermediate of gp41 is accompanied by structural rearrangements in the CDR-H2 region. It seems likely that the presence of two Phe residues at positions 54 and 56 of the CDR-H2 of Fab 8062 may result in a reduction in conformational plasticity of the CDR-H2 loop of Fab 8062 relative to that of Fab 8066, which results in the two Fabs adopting distinct CDR-H2 conformations upon binding 5-Helix. In both complexes every residue of the CDR-H2 loop is involved in interactions with 5-Helix, but topologically they are located very differently in relation to the body of the bundle (Figs. 2C, 5B and 5C). In particular, the CDR-H2 of Fab 8066 is buried in the body of the bundle, whereas in the case of Fab 8062 it is barely touching its surface (Figs. 5B, C). This difference also impacts the local conformation of 5-Helix, reflected by small but significant backbone atomic displacements in the region contacting the CDR-H2, as well as in immediately adjacent areas. The structural differences involving interactions of CDR-H2 with 5-Helix in the Fab 8066 and 8062 complexes are likely to perturb the lifetimes of the two complexes (as reflected in dissociation rate constants and hence affinity since the association rate constants are similar for the two complexes; unpublished data). Since the antibodies take the place of the Cc helix needed for completion of the structure of the 6-HB conformation of gp41 (Fig. 9), the lifetime of the complex will correlate with the ability of the antibody to prevent creation of a proper 6-HB needed for viral entry. The different conformation and mode of interaction of the CDR-H2 of Fab 8066 and 8062 with 5-Helix also result in smaller propagated structural changes throughout the complex, which are likely to further contribute to the different binding affinities with 5-Helix. For example, there are a few additional interactions involving CDR-H3 (Fig. 6), CDR-L1 (Fig. 7A), CRD-L2 (Fig. 8), and CDR-L3 (Fig. 8) in the Fab 8066 complex compared to the Fab 8062 complex. Further, some of the interactions that are present in both complexes are stronger (i. e. shorter intermolecular contacts) in the Fab 8066 complex. Indeed, the gap volume index (which reflects the tightness of the interfacial packing [43]) is significantly smaller in the Fab 8066 complex (2. 3 Å) than in the Fab 8062 complex (2. 6 Å). Although Fabs 8066 and 8062 are both able to bind to the exposed N-HR trimer with similarly high affinity [39], their ability to bind to the 5-Helix construct differs by about 2 orders of magnitude, and that difference is reflected in their neutralization activity (Table 1). This observation leads us to two conclusions. First, the interactions of the Ca helix of 5-Helix with the CDR-H3 (Fig. 6), CDR-L2 (Fig. 8) and CDR-L3 (Fig. 8) loops may modulate the affinity of Fabs 8066 and 8062. Second, the assay (involving a solution equilibrium titration in conjunction with an electrochemiluminescence-based affinity measurement) used to determine KD values for the binding of Fabs to the N-HR trimer mimetic NCCG-gp41 reported by Gustchina et al. [39] reflect the binding of only a single Fab to NCCG-gp41, and hence do not report on any possible negative cooperativity arising from steric clash between neighboring Fabs bound to the N-HR trimer (cf. Fig. 11). A key feature of fusion inhibitors that target the N-HR trimer, including presumably antibodies directed against the N-HR trimer, is that deactivation of gp41 in vivo is a slow reversible process that is dependent on chemokine receptor binding to Env, and that the exposed N-HR trimer remains accessible to inhibitors until the final conformational changes in gp41 that lead to the formation of the 6-HB have taken place [31]. This suggests that inhibition of fusion will be most effective when two or more Fabs are bound to the exposed N-HR trimer of the pre-hairpin intermediate of gp41. Since multiple Fabs/antibodies bound to the N-HR trimer are unlikely to dissociate simultaneously, the probability that at least one antibody is bound to the N-HR trimer at all times will be increased. Our modeling suggests that three molecules of the most potent neutralizing Fab in our series, Fab 8066, can readily bind to the N-HR trimer without any steric clashes between adjacent Fab molecules (Figs. 10 and 11). In contrast, as a consequence of the different mode of binding of the CDR-H2 to the hydrophobic pocket on the surface of the N-HR trimer, binding of three molecules of Fab 8062 to the N-HR trimer may result in steric clash between adjacent Fab molecules involving the CDR-H2, CDR-L1, CDR-L3 loops and loop 71–78 (Fig. 11). It seems likely in the light of the current structural data, modeling results and neutralization properties of our Fab series [39], that neutralization is dependent not only on tight binding of a single Fab to the N-HR trimer but also on the ability to bind multiple Fabs to a single N-HR trimer at a preliminary step of the fusion process.
The plasmid construct to express 5-Helix fused to a C-terminal His-tag was generously provided by Michael Root (Thomas Jefferson University). A stop codon was engineered preceding the His-tag using the Quik-Change mutagenesis protocol (Stratagene, La Jolla, CA) to express a tag-less version of 5-Helix. 5-Helix was expressed in E. coli, purified from the insoluble fraction under denaturing conditions on a Superdex-200 column, and then subjected to reverse-phase HPLC. The protein was dialyzed against 50 mM sodium formate, pH 3, concentrated to ∼10 mg/ml and stored at 4°C. Fabs 8066 and 8062 were initially expressed and purified as described previously. Further purification was carried out by size-exclusion chromatography on a Superose-12 column (16×60 cm, GE HealthCare) equilibrated in 10 mM Tris-HCl, pH 7. 5,150 mM NaCl. Peak fractions were pooled and concentrated to ∼2 mg/ml and stored frozen. The D5 monoclonal antibody in IgG and Fab formats was generously provided by Joseph Joyce (Merck Research Laboratories). Liquid chromatography-mass spectrometry/mass spectrometry analysis [45] of endoprotease (trypsin, Lys-C, Glu-C, chymotrypsin) digested Fabs either as liquid samples [46] or after separation by SDS/PAGE [47] was used to provide partial sequence information of regions of the engineered Fabs prior to obtaining the full sequences from AbD Serotec. Size exclusion chromatography with detection by Multiangle light scattering (DAWN EOS, Wyatt Technology Inc. , Santa Barbara, CA) and Refractive index (OPTILAB DSP, Wyatt Technology) (SMR) was used for mass analysis of 5-Helix in complex with Fab 8066 and to optimize conditions for the large-scale preparation of complexes for crystallization. 5-Helix (103 µg) and Fab 8066 (106 µg) before and after mixing with each other (∼50 µg each giving a molar ratio of 1∶2 Fab 8066 to 5-Helix) in a final volume of 150 µl were loaded on Superdex-75 column (1×30 cm) equilibrated in 1x PBS at a flow rate of 0. 5 ml at room temperature. Molecular masses were calculated using the Astra software provided with the instrument. Preparation of the Fab 8066/5-Helix complexes for crystallization was carried out by slowly mixing 64 µl of 5-Helix (575 µg corresponding to a 1. 5-fold molar excess over Fab) with 1. 5 ml of Fab (∼ 0. 8 mg) solution kept in 10 mM Tris-HCl, pH 7. 5 and 150 mM NaCl and separating the complex on a Superose-12 column (16×60 cm) in the same buffer. Because of the lower affinity of Fab 8062 for 5-Helix, a 3-4 fold higher protein concentration of the Fab 8062/5-Helix complex was maintained prior to subjecting the complex for fractionation on the column. The complexes were concentrated to a final concentration of ∼10 mg/ml and stored at 4°C. ITC measurements were performed using an ITC200 titration calorimeter (Microcal Inc.) at 28°C. The stock solution (∼10 mg/ml) of 5-Helix in 50 mM sodium formate, pH 3, was diluted to about 0. 5 mg/ml in 25 mM Tris-HCl, pH 7. 5,150 mM NaCl, dialyzed against 10 mM Tris, pH 7. 5,150 mM NaCl and concentrated to ∼5 mg/ml. Antibody solutions (5–15 µM) in 10 mM Tris, pH 7. 5,150 mM NaCl kept in the calorimetric cell were titrated against 10 and 20 fold higher concentration of 5-Helix for monovalent Fabs 8062,8066 and D5, respectively. The data were analyzed using the Origin software provided with the instrument. The cell lines and molecular clones employed were identical to those described in our previous publication [39]. Preparation of Env-pseudotyped HIV-1 and Env-pseudotyped HIV neutralization assays (employing TZM-b1 indicator cells that constitutively express CD4, CCR4 and CXCR4) were carried out exactly as described previously [39]. Crystallization of the Fab 8066/5-Helix and Fab 8062/5-Helix complexes was carried out by the hanging drop, vapor diffusion method. The Fab 8066/5-Helix complex was concentrated to 7 mg/ml in 10 mM Tris buffer at pH 8. 0, also containing 0. 15 M NaCl. The reservoir solution contained 1. 2 M dibasic ammonium phosphate in 0. 1 M HEPES buffer at pH 8. 0. Each drop contained 2 µl of protein samples and 2 µl of reservoir solution. The crystals grew to the maximum size of ∼0. 2 mm in 4 to 5 days. The Fab 8062/5-Helix complex was treated in a similar way, with the exception that the well solution contained 14% polyethylene glycol 10K, 0. 1 M ammonium sulfate, and 5% ethylene glycol. Diffraction data for both complexes were collected using synchrotron radiation at the SER-CAT ID-22 beamline at the APS, Argonne National Laboratory, from one crystal each. The resolution of the data for the Fab 8066/5-Helix complex and Fab 8062/5-Helix complex was 2. 05 and 2. 5 Å, respectively. The structure of the Fab 8066/5-Helix complex was solved by molecular replacement with the program PHASER [48], using the coordinates of the D5/5-Helix complex ([41] PDB ID 2CMR) as the search model. The structure was completed through a number of cycles of refinement with REFMAC5 [49] and rebuilding with COOT [50]. The refined structure of the Fab 8066/5-Helix complex was used as the search model to solve the structure of the Fab 8062/5-Helix complex, with similar procedures used for subsequent refinement and model building. The statistics for data collection and structure refinement are listed in Table 2. The coordinates and structure factors for the Fab 8066/5-Helix and Fab 8062/5-Helix complexes have been deposited in the Protein Data Bank with IDs 3MA9 and 3MAC, respectively. | Membrane fusion of HIV-1 with its target cells represents the first step in viral infection. This process involves a series of conformational changes in two viral envelope glycoproteins, gp120 and gp41, subsequent to binding of gp120 to the CD4 receptor and the chemokine coreceptor on the target cell membrane. During the fusion process, the conserved N-heptad repeat (N-HR) of gp41 in the form of a trimeric coiled-coil is accessible and presents an attractive target for the generation of broadly neutralizing antibodies. Here we present the crystal structures of two monoclonal Fabs complexed to a mimetic of the N-HR trimer. These Fabs were derived from a synthetic human combinatorial antibody library comprising more than 1010 human specificities by first panning against an N-HR mimetic, followed by affinity maturation through targeted diversification of the CDR-H2 complementarity determining region. One of the Fabs is broadly neutralizing across a wide range of primary isolates from subtype B and C HIV-1, whereas the other one is non-neutralizing. Our structures reveal the key role of the CDR-H2 loop in antigen recognition and how this correlates with HIV-1 neutralization properties. | Abstract
Introduction
Results
Discussion
Materials and Methods | infectious diseases/hiv infection and aids | 2010 | Structural Basis of HIV-1 Neutralization by Affinity Matured Fabs Directed against the Internal Trimeric Coiled-Coil of gp41 | 13,499 | 299 |
The branch point (BP) is one of the three obligatory signals required for pre-mRNA splicing. In mammals, the degeneracy of the motif combined with the lack of a large set of experimentally verified BPs complicates the task of modeling it in silico, and therefore of predicting the location of natural BPs. Consequently, BPs have been disregarded in a considerable fraction of the genome-wide studies on the regulation of splicing in mammals. We present a new computational approach for mammalian BP prediction. Using sequence conservation and positional bias we obtained a set of motifs with good agreement with U2 snRNA binding stability. Using a Support Vector Machine algorithm, we created a model complemented with polypyrimidine tract features, which considerably improves the prediction accuracy over previously published methods. Applying our algorithm to human introns, we show that BP position is highly dependent on the presence of AG dinucleotides in the 3′ end of introns, with distance to the 3′ splice site and BP strength strongly correlating with alternative splicing. Furthermore, experimental BP mapping for five exons preceded by long AG-dinucleotide exclusion zones revealed that, for a given intron, more than one BP can be chosen throughout the course of splicing. Finally, the comparison between exons of different evolutionary ages and pseudo exons suggests a key role of the BP in the pathway of exon creation in human. Our computational and experimental analyses suggest that BP recognition is more flexible than previously assumed, and it appears highly dependent on the presence of downstream polypyrimidine tracts. The reported association between BP features and the splicing outcome suggests that this, so far disregarded but yet crucial, element buries information that can complement current acceptor site models.
Pre-mRNA splicing, which is essential for the production of functional mRNAs, is a co-transcriptional set of reactions catalyzed by a large ribonucleoprotein complex – the spliceosome – composed by five small nuclear RNAs (snRNAs) and more than hundred proteins [1], [2]. In addition to these core factors, splicing is often dependent on other proteins that can either activate or repress signal recognition, therefore playing a very important role in the regulation of specific events [3], [4]. In a process called Alternative Splicing (AS) introns can be differentially removed, generating multiple isoforms from the same pre-mRNA molecule [5], which is key for the increased protein diversity observed in metazoans [6], [7]. The importance of splicing in the regulation of gene expression is also underlined by the fact that mutations affecting it are frequently associated with, or directly responsible for, severe genetic diseases [8], [9]. Splicing requires the presence of three main signals that directly participate in the reaction and that are present in every intron: the 5′ splice site (5SS); the 3′ splice site (3SS); and the branch point (BP) [10]. These signals, along with the polypyrimidine tract (PPT), are critical for correct spliceosome assembly [10], [11]. Additionally, there are also cis-acting splicing regulatory sequences that can function as enhancers or silencers of splicing [12]–[17]. These are not only important in the regulation of splicing in a context dependent manner, but are also crucial for splice site recognition in general [18]. While the 5SS is located at the start of the intron, the other three core elements – the BP, the PPT and the 3SS – are normally arranged in this order within the last 40 nucleotides (nts). However, this arrangement is not mandatory. There are introns in which the BP can be located up to 400 nts away from the 3SS [19]–[24]. These are referred to as distant BPs (dBPs) and account for approximately 1% of all human introns. dBPs are rarely found by computational methods, since most BP prediction methods use, as condition, the proximity to the 3SS. dBPs have typically an adjacent long PPT downstream and have been associated with AS, in particular with mutually exclusive exons [21], [23]. For both distant and proximal BPs, the region between the BP and the 3SS is usually devoid of AG dinucleotides [22] (Figure 1). However, the AG dinucleotides can either occur at downstream locations close to the BP (distance<12–15 nts – region r3 in Figure 1), where they can be bypassed, or close to the 3SS (distance<12 nts – region r1 in Figure 1), where they would be competing with the actual 3SS [25]. This extended region is named AG Exclusion Zone (AGEZ). In mammals, while both 5SS and 3SS signals have been mapped precisely by aligning transcriptional evidence to the genome, allowing the development of robust statistical models from large datasets [26], [27], BP characterization has been a far more complicated task. Firstly, the lack of a sufficiently large “gold standard” set of mammalian BPs – only a few dozens have been mapped [28], [29] – makes difficult the task of building statistical models. Additionally, unlike for some fungal species, which present very strict BP consensus [30], [31], the mammalian BP is an extremely degenerate motif [28]. This is attested to by recent studies that had focused on the BP signal and on its relation with splicing factors across a wide range of species, including mammals [32], [33]. These studies based their BP predictions on the Hamming distance to the U2 complementary sequence TACTAACAC [33]. While such approach has been used successfully in fungal species, it proves insufficient for mammals where the resulting motif consensus reflects, above all, the background nucleotide frequencies and the consensus used to search it. In this paper we present a new strategy for predicting BPs in human. Using conservation and positional bias we first built a set of high-confidence putative BPs. This set was then used as a positive training set for a Support Vector Machine (SVM) learning algorithm that combined both BP and PPT information into a predictive model. We show that this method outperforms previously published methods for both proximal and distant BPs. Applying our predictive algorithm to human introns, we are able to characterize the localization of the BP within the intron and describe how BP signal features may contribute to the final splicing outcome. Moreover, we experimentally determined the BP location for some introns containing long AGEZs, which are characterized by the presence of dBPs, and in which we observed alternative BP usage. Our computational and experimental analyses suggest that BP recognition is highly plastic, partially dependent on downstream PPT features, and of critical importance to the final outcome of the splicing reaction.
To circumvent the lack of a large set of experimentally verified BPs from which a predictive model could be derived, we decided to build a set of high-confidence putative BPs and use it as positive training set. Rather than using as starting hypothesis a strict consensus and finding sequences that are similar to it, we tried to capture an unbiased BP sequence signal using positional and conservation principles. The mammalian BP is a quite degenerate motif with only two highly constrained positions, the branch point A and a T two bases upstream, which we denote as TNA. If we consider this motif alone, we observe that it is strongly conserved towards the last 50 nt of the introns (Figure 2A), with a peak around position −23 relative to the 3SS (see Figure 1 in Text S1 for all the other trinucleotide combinations). Surprisingly, a simple motif overrepresentation approach does not allow the identification of words that can potentially be associated with the TNA distribution profile. Indeed, if one computes pentamer frequencies in the region spanning from position −55 to −15 relative to the 3SS in human, the large majority of the most abundant pentamers do not contain the TNA motif and appear to be associated with the PPT signal due to their high pyrimidine content (see Table 1 in Text S1). Moreover, very few TNA-containing pentamers present a non-uniform distribution profile over the last 300 nts of human introns, hardly being representative of the expected BP signal variability (Figures 2–4 in Text S1). Thus, we considered a comparative approach. Under the assumption that functional sites are potentially more conserved than non-functional ones, we expect BP-related words to be more conserved in the region for which we observe a peak in the TNA motif distribution. Accordingly, by considering only TNA-containing pentamer instances that were perfectly conserved across 7 mammalian species (Homo sapiens, Pan troglodytes, Macaca mulatta, Mus musculus, Rattus norvegicus, Canis familiaris and Bos taurus), we were able to find a clear distribution profile for each pentamer. Next, by performing a set of statistical tests to these profiles (see Methods), we were able to separate the set of all 184 TNA-containing pentamers into 3 categories based on their positional bias: 1) No association with any positionally biased signal (N = 37), 2) PPT-associated (N = 23) and 3) BP-associated (N = 124) (see Methods). Example pentamers for each category are shown in Figures 2B, 2C and 2D (see Figures 2–4 in Text S1 for all 184 TNA-containing pentamers). Finally, to build the final set of putative BPs, we selected all 9-mers including conserved TNA instances in their central position that were unique in the last 300bp of an intron, falling between positions −55 and −15 relative to the 3SS (consTNA set – for further reference). This set was subsequently filtered, forcing the overlap of at least one BP-associated pentamer in each species. In this manner, we allow sequence variability maintaining functional conservation (Figure 2E). After this step we were left with a set of 8156 conserved putative BPs, which we denote as consTNA-BP5 (provided in Dataset S1). Using the consTNA-BP5 set, we derived the sequence logo corresponding to the BP signal in human (Figure 3B). In accordance with previously published studies, the human BP signal is quite variable, presenting very low information content (IC = 4. 739, including the fixed T and A positions) (Figure 3A). Moreover, our results corroborate the YTNAY consensus determined experimentally in [28]. Nevertheless, there seem to be some constraints on the central position, where T appears at a very low frequency, while G, A and C are found with much higher probability (Figure 3B). Positions +1 and −3 relative to the BP A seem also to contain more frequently purines than previously assumed. In order to address whether specific nucleotide combinations are more or less frequent than expected under a model of independence between positions, we used the Mutual Information (MI) measure (see Methods). We found dependencies between adjacent positions in the BP signal (discarding the 2 fixed positions) (Figure 3C). Moreover, we also found weak second-order dependencies involving the central positions of the signal. Although MI values are low in general, the relatively large set size allows us to capture these small dependencies, which we use to describe the sequence signal with a position-dependent Markov model of order 1, which we denote as MM1 (see Methods). To test whether the frequencies of different BP-associated words correlate with the U2 binding stability, we grouped all the consTNA-BP5 9-mers by their central pentamer sequence, which are the five positions with higher IC in the BP signal, and calculated the mean U2 binding energy for each group (see Methods). In Figure 4 we can observe that there is a direct correlation between the stability of the binding to the U2 and the occurrence of these words in the consTNA-BP5 set (Spearman' s rank correlation, rho = −0. 65, p = 6. 53×10−9, see Figure 5 in Text S1), which validates the captured BP signal. Interestingly, if we compare this set with the consTNA set, we observe for the consTNA-BP5 a drastic reduction in the frequency of some words for which the U2 binding energy is low (Figure 4). In fact, these are mainly PPT-associated words which were filtered out from the consTNA set. Thus, we can consider the consTNA-BP5 set as a good representative of putative functional BPs. The low IC observed for the BP signal motivates the addition of other sequence features in a predictive model. We therefore considered a model that incorporates the relation between the BP and the properties of downstream proximal PPTs. We took as positive training set the consTNA-BP5 set and as negative a random set of intronic 9-mers containing TNA in the same position (see Methods). For every candidate BP, from both positive and control sets, we computed the sequence score with a position-dependent Markov model of order 1 (MM1) and three PPT-related features (see Methods). These features were used as input for an SVM learning algorithm, which produces a score that reflects the distance in feature space between the candidate and the decision boundary. Elements for which the SVM score is above the threshold (zero, in this case) are labeled as positive, while elements scoring below the threshold are considered negative. The higher the score is (in absolute terms), the more reliable the prediction. This method shows good discriminative power between positive and negative BP candidates. Indeed, using a 10-fold cross-validation, the average accuracy for our method is 0. 794±0. 011 and the receiver operator characteristic (ROC) analysis shows an area under the curve (AUC) of 0. 878±0. 011. Additionally, in order to understand the effect of incorporating PPT information in the model, the SVM model was compared with the position-dependent Markov model MM1 and a position weight matrix (PWM) model. We show in Figures 6A and 6B in Text S1 that the SVM model outperforms the other two methods (AUC, MM1: 0. 778±0. 013 and PWM: 0. 764±0. 014) not only in accuracy but also in precision. This difference reflects the importance of additional sequence features in BP recognition. Relative to the comparison between MM1 and PWM, Figure 6 in Text S1 shows that incorporating dependencies between positions yields extra, though marginal, predictive power. In order to test our SVM model in a more realistic situation, we compared the predictions on a set of experimentally verified BPs (provided in Dataset S1) with other previously published methods for BP prediction. Thus, we collected a set of 35 human introns that were not part of the training set and for which at least one TNA-containing BP had been experimentally determined. Additionally, 7 introns containing experimentally verified dBPs were added to this test set. Out of the 42 introns, 3 contained more than one mapped BP. For these cases, we considered a prediction as correct if any of them were detected. Using our SVM model, we considered as positive the highest scoring hit falling in the AGEZ. BP search was also performed using 3 other methods from recent publications. Two of these methods, Schwartz [32] and Plass [33] methods are based on the complementarity to the U2. In these two cases, candidates are ranked according to their Hamming distance to a strict consensus. The third method, Gooding method [22] scores candidates using a PWM trained from human data and it has been successfully applied to find dBPs. Additionally, while Plass and Schwartz methods search over a region of length 100 nts and 200 nts respectively, ours and Gooding' s methods search candidates in the AGEZ only. Benchmarking results are shown in Figure 5 in terms of sensitivity, computed as the number of true positives (TP) over the total number of introns. Our SVM model shows the best performance, determining correctly the BP for 76% of the introns tested (Figure 5C). Gooding method, also trained on human data, comes up as second best, predicting correctly the BP for approximately 60% of the introns. There is a big overlap between these two methods, with about 84% of Gooding predictions being also predicted by our SVM model (Figure 5B). Hamming distance based methods are the least accurate, in part due to their high strictness. In Figure 5A, blank spaces are BPs that, in each method, either did not match the initial sequence requirements or are outside the search region. Moreover, in Schwartz method, some BPs that are ranked 1st are ultimately discarded because they are not the closest candidate to the 3SS. Consequently, Schwartz method has the highest positive predictive value (PPV) of all four (0. 83%). However, sensitivity (0. 24%) is negatively affected. To further assess the accuracy of our predictive algorithm in the particular case of long AGEZ-containing introns, where potential dBPs might be present, we decided to experimentally determine the BP position for several exons preceded by long AGEZs by means of in vitro splicing reactions followed by primer extension. We subsequently compared the mapped BPs to our ranking of candidates given by our SVM algorithm (Figures 7–11 in Text S1). For this analysis, we considered 3 exons from the muscleblind-like 1 (MBNL1) gene, which are preceded by long AGEZs (exons 6,8 and 9, with AGEZs of length 173,134 and 224 nts, respectively) and exon 4 from two members of the CDC-like kinase protein family (CLK1 and CLK3), both preceded by long AGEZs of 235 and 207 nts in length, respectively. Additionally, 4 exons with long AGEZs from the serotonin receptor 4 (HTR4) gene (exons 3,4, 5 and g, with AGEZs of length 149,291,221 and 101 nts, respectively), for which the BP location has been recently determined [34], were added to complement the analysis. Results are summarized in Table 1. With the exception of MBNL1 exon 8, which is nevertheless characterized by two BPs located at non-canonical positions (−51 and −64), all exons use dBPs, further underlying the association between long AGEZs and the usage of dBPs. Remarkably, in most cases there is not a unique BP, but more complex arrangements where several different BPs, some of which located at a more canonical position, can be alternatively used (Figures 7B, 8B, 9B, and 10B in Text S1). Interestingly, for MBNL1 exon 9 the predominant BP is at −229. The two very weak signals, located at more canonical positions (−31 and −41), appear to be associated with a slower migrating lariat species that was only observed at the last time point of the in vitro splicing reaction (Figure 9A in Text S1), whereas the more rapidly migrating lariat, corresponding to the −229 BP, appeared much earlier. In 5 out of the 9 introns considered here, our SVM classifier was able to rank as top prediction at least one of the used BPs (Table 1). In three of the remaining cases, at least one of the used BPs in each of the introns ranked second according to the SVM score. The modest performance of the SVM classifier in these cases is in part explained by the generally weak sequence score observed for most of the BPs. However, these weak signals are usually compensated by the presence of a downstream PPT (see Figures 7C, 8C, 10C, and 11C in Text S1), which results in a ranking improvement. This is the case for the MBNL1 and CLK exons, where it is always possible to observe an increase in ranking from the MM1 sequence score to the final SVM score (Table 1). One example is the BP located at position −141 relative to MBNL1 exon 6. Considering signal strength only, this BP ranks 5th among all the 11 candidates present in the AGEZ. Interestingly, our SVM classifier places it as the second best prediction for that intron, with a resulting score slightly lower than the obtained for a BP candidate located at position −84, also preceding a PPT and for which the sequence signal is considerably stronger (Figure 7C in Text S1). Similar situation can be observed for MBNL1 exon 8, where a very strong BP signal (tgctgAcag) at position −138 followed by a PPT of considerable length leads to misprediction (Figure 8C in Text S1). Here again, the ranking according to the SVM classifier for both mapped BPs (at positions −64 and −51) is better when compared to the MM1 score (raising from 3rd and 6th to 2nd and 4th, respectively). Concerning CLK1 and CLK3 exons, in both cases the BPs are located towards the 5′ end of the AGEZ. Interestingly, despite the high pyrimidine content in these regions, there is not a continuous PPT stretch due to frequent purine interruptions (Figures 10C and 11C in Text S1). Prediction in this situation is additionally hindered by the presence of much stronger signals located towards the middle of the AGEZ, which are also associated with PPTs of considerable length. In respect to HTR4 predictions, our SVM classifier identified as top prediction for exon 3 a BP candidate located at position −143. Despite the fact that primer extension experiments point to the usage of a BP located at position −151 (ranking second according to the SVM classifier), further mutagenesis analyses suggested the possible usage of the first predicted site [34]. Regarding the remaining HTR4 exons, our SVM algorithm was capable of top ranking each BP mapped by primer extension. It is worth mentioning that all HTR4 exons considered here are characterized by having the BP localized towards the 5′ end of the AGEZ and by the presence of long PPTs covering almost the totality of the AGEZ. Using the SVM classifier, all introns in our human dataset (N = 183187) were scanned for BPs. In order to study the relation between the AGEZ and BP position in more detail, all BP candidates falling in the last 500nt of every intron were scored, regardless of being in the AGEZ or not. For introns shorter than 500nt, the entire intron was scanned. In Figure 6A, we plot the distribution of the BP A position of the best hits per intron relative to the AGEZ-defining AG-dinucleotide (a3 in Figure 1). We observe that the most frequent location of the BP is inside and towards the 5′ end of the AGEZ. The left-most tail in the distribution reflects the background probability of finding a high scoring BP candidate in all the intron. Interestingly, from 5′ to 3′, the frequency of occurrences increases, starting at a distance of 7–8 nucleotides upstream the AGEZ-defining AG-dinucleotide. This distance is shorter than the 12nt considered when defining the AGEZ (see region r3 in Figure 1). These results suggest that the BP can be most frequently found within the AGEZ and that there is no need to search beyond that. In effect, only in approximately 5% of the introns no candidate was found in the AGEZ. For the remaining 95% we were able to retrieve candidates within the AGEZ, of which approximately 89% score positively (Figure 6B). This percentage drastically drops when considering the next AGEZ upstream of this one (Figure 12A in Text S1), where only in less than 25% of the cases there is a positive hit. When considering the top scoring candidates in the AGEZ (our set of predicted human BPs from this point on), we can observe a distribution bias with approximately 96% of the cases falling between −15 (downstream limit) and −55nt relative to the 3SS with a peak at position −24. However, the distribution extends up to almost the maximum of 500 nt, with ever-diminishing frequencies (Figure 6C). Considering dBPs as predicted BPs that lie beyond 100bp from the 3SS, i. e. 4 times the average 3SS-BP distance, these account for a very small percentage (0. 4%, n = 688) of the total predicted BPs (n = 173284). Comparing this set with BPs predicted in the standard range (−55, −15) (Figure 13 in Text S1), we found that dBPs have stronger motif sequences (Mann-Whitney, p = 1. 34×10−29). Interestingly, the pyrimidine content between dBP and the 3SS is similar to closely located BPs (Mann-Whitney, p = 0. 24), which is surprising considering the large distance. Consequently, PPTs nearby the dBPs are longer and thus have higher score (Mann-Whitney, p≈0). In summary, this leads to higher SVM scores for dBPs (Mann-Whitney, p≈0). Interestingly, BP-3SS distance positively correlates with AS. Skipped exons tend to be more frequently preceded by introns containing distant BPs than constitutive exons (Mann-Whitney, p = 1. 97×10−8) (Figure 7A). As the BP-3SS distance increases, so does the percentage of exons for which there is skipping evidence. It is possible to observe an almost linear correlation between BP distance and frequency of skipped exons. We found skipping evidence for approximately 43% of the exons in which the BP is located at more than 100 nts upstream, whereas for exons preceded by proximal BPs (3SS-BP distance<50 nts), only 28. 6% of them were skipped (Chi-square, p = 1. 61×10−6). Remarkably, this association also holds for the exon inclusion level. For the fraction of skipped exons, inclusion was calculated based on expressed sequence tag (EST) data (see Methods) and is plotted in Figure 7B. Exon inclusion decreases with BP distance. While skipped exons preceded by proximal BPs (distance<50 nts) are included in average in 85% of the transcripts, this value drops down to 65% for exons with a distal BP (3SS-BP distance>100 nts) (Mann-Whitney, p = 2. 87×10−9). Additionally, BP sequence score also correlates with AS. In Figure 7C, we observe that skipping of the downstream exon is more frequent for introns with lower BP sequence score. This increase in skipping is fairly gradual. Even though the sequence score distribution is skewed towards high values (not shown), the difference in skipping percentage between lower and upper sequence score quartiles (defined by scores lower than −0. 338 and higher 1. 838, respectively) is strongly significant (Chi-square, p = 1. 83×10−10). Moreover, there is small, but statistical significant, difference in BP sequence score between skipped (mean = 0. 706) and constitutive (mean = 0. 797) exons (Mann-Whitney, p = 1. 78×10−9), further validating that observation. We also found clear differences in the BP signal between short and long introns. Shorter introns have in general strong BPs (Figure 7D). Indeed, introns of length up to 100 nt, contain BP sequences scoring almost 1. 4 in average. This value tends to decrease in a gradual manner as intron length increases up to approximately 1000nt (Spearman' s rank correlation, rho = 0. 12, p≈0), where from that point on it stabilizes (Spearman' s rank correlation, rho≈0, p = 0. 154). Interestingly, even though pyrimidine content between the BP and the 3SS is higher in shorter introns, the overall PPT score is lower, possibly due to slightly shorter PPTs (see Figure 14 in Text S1). Nevertheless, the final SVM score for shorter introns is higher (Figure 15A in Text S1). Another interesting observation is that shorter introns have, in average, lower BP candidate density, both in the AGEZ, or when considering the last 100 nts (Figure 15B in Text S1). This fact cannot be explained by the shorter AGEZs in short introns, as differences are small. Recent studies have reported a strong relation between exon age and AS [18], [35], [36]. It has been suggested that the low inclusion observed for young exons is due to weaker splicing signals in general [18]. In order to investigate whether BP features are also related to the differences observed between exons with different evolutionary age, we predicted BPs in three exon sets: primate specific (PS) exons; mammalian conserved (MC) exons; and, as control, a set of pseudo exons, which have no inclusion evidence. In Figure 8A we show that BPs preceding real exons have in average higher SVM scores than those preceding pseudo exons (Mann-Whitney, p≈0). This difference is even greater when comparing to BP candidates preceding random AG dinucleotides (Figures 12B, 12C in Text S1), since pseudo exons are preceded by a PPT signal, contributing to a higher BP SVM score (see [18] for details on the pseudo exon set construction). Interestingly, BP preceding PS exons, which have low inclusion, have intermediate values between pseudo exons (Mann-Whitney, p = 6. 18×10−5) and MC exons (Mann-Whitney, p = 0. 022). As we show in Figure 8B, these differences are mainly explained by differences in sequence score (Mann-Whitney, pseudo vs. PS p = 7. 58×10−5; PS vs. MC p = 0. 039). Regarding intronic position, BPs preceding pseudo exons tend to be located closer to the 3SS compared to real exons (Mann-Whitney, p≈0). BPs preceding real exons show a distribution peak between positions 20 to 25 nts upstream of the 3SS, whereas in pseudo exons this peak is located at the smallest distance considered (15 nts). Finally, no differences were found between PS and MC exons (Figure 8C) regarding BP-3SS distance. This feature strongly correlates with AGEZ length, which does not differ significantly between sets (data not shown).
Perhaps the major challenge in mammalian BP prediction consists in circumventing the lack of a large gold standard set from which models can be trained. Here we present a strategy based on positional bias and conservation, in order to improve signal detection. We were able to select word motifs with a common sequence conservation distribution profile regardless of their frequency of occurrence along introns. This approach resembles other motif discovery algorithms, such as PEAKS [37], which are also based on positional bias. However, the use of conserved instances alone allows for the production of differentiated distribution profiles, even for words occurring at very low frequency. Using BP-associated pentamers we built a set of high-confidence BP candidates by selecting functionally conserved instances. Remarkably, the abundance of the different motif variants correlates with U2 snRNA binding stability. There are, however, some exceptions, such as the CTCAN-containing nonamers that, despite their low binding stability, are quite frequent and highly conserved motif variants (see e. g. CTCAC pentamer in Figure 3 in Text S1). Nevertheless, the observed correlation, besides being a good indicator of the set quality, also suggests that the probabilistic score resulting from the statistical modeling of the signal might be representative of the signal strength, and therefore related to splicing efficiency. To build a BP predictive model, at least three main issues had to be tackled and improved over previously published methods. First, an adequate signal modeling: some positions in the signal might play a more prominent role in U2 binding stability compared to others, something not taken into account by Hamming distance-based scoring methods, and which appears to be partially responsible for their lower performance compared to methods based on a probabilistic description of the signal. Additionally, it was possible to detect subtle dependencies between adjacent positions within the signal, which we have exploited in a position-dependent 1st order Markov model, yielding additional discriminative power over the use of a simple PWM. Moreover, the BP signal may be affected by the sequence biases of the intronic context. Indeed, the consensus BP motif in the training set differs between GC-rich and GC-poor introns. However, our model can recapitulate these properties, i. e. we predict the same consensus as expected for GC-rich introns and, likewise, for GC-poor introns (see Figure 16 in Text S1). A second issue in the prediction of the BP is the association with other signals. The low IC of this signal in mammals suggests that BP selection may depend on additional signals in the pre-mRNA sequence. Accordingly, we considered features of the downstream PPT and incorporated them in the model using a machine-learning algorithm. Our benchmarking analysis demonstrates that additional PPT information improves accuracy over the probabilistic modeling of the signal alone, and over previously published methods. This improvement is particularly clear for cases in which the actual BP sequence is not a frequent motif variant. In fact, our results suggest that PPT information accounts for the majority of the accuracy difference between Gooding' s and our SVM model. This reinforces the importance of modeling the relation between these two signals, as stronger PPTs might compensate for weaker BPs, but the lack thereof might even impair strong BP candidate recognition. Finally, a third issue in BP prediction is the localization – even though BPs generally localize towards the region from 20 to 40 nucleotides upstream the 3SS, we show that BP localization is more dynamic than normally assumed. It appears to be highly dependent on the presence of AG dinucleotides in the 3′ end of the intron. In this work, and like in Gooding method [22], we search BPs in the AGEZ only. This contrasts with the other two methods (Plass [33] and Schwartz [32]) that, for every intron, scan a fixed region of 100 nts and 200 nts respectively, preferentially selecting hits that are closer to the 3SS. This fundamental difference might account for their lower prediction accuracy not only in the test set of mapped dBPs, but also in the set of proximal BPs, where searching over an unnecessarily long region can lead to the appearance of false positives. This may be particularly relevant for ab initio gene predictors that use BP information in their acceptor site models (see for instance [38]–[40]) by scanning over a fixed window size and do not consider more distant BPs as a possible configuration, which could lead to mispredictions. In this work we devoted special attention to the specific case of long AGEZ-containing introns. These are not only interesting from a computational point of view, as they represent an atypical kind of acceptor arrangement and potentially harbor a high number of BP candidates, but also from a biological perspective due to their association with regulated alternative splicing events. We experimentally mapped the BP for 5 introns characterized by the presence of long AGEZ in the 5′ terminus from MBNL1, CLK1 and CLK3 genes. One of the most striking and intriguing observations, for these introns and for the additional set of 4 introns from the HTR4 gene, is the fact that more than one BP can be used. Remarkably, BPs often appear as doublets, with the BP adenosines in closely spaced positions, probably in association with the same PPT (see Figures 6,7, 8, and 9 in Text S1). For MBNL1 exon 9, the presence of a second population of lariats occurring at the later time point (180 minutes) of the in vitro splicing reaction also reflects the additional usage of two closely spaced BPs located at canonical positions. Interestingly, the splicing kinetics for this second population, which is associated to very weak BPs, appears to be much slower compared with the one in which the dBP is selected. Regardless of any mechanistic interpretation, the evidence presented here strongly supports that BP recognition in human introns can be more plastic than previously assumed, which probably ensures a greater resistance to BP disruptive mutations and/or allows for greater control over specific alternative splicing events. This hypothesis is in agreement with the observations in [34], where splicing of HTR4 exons 3 and 5 is very resilient to mutations of the mapped BPs, being only impaired upon mutation of every surrounding adenosine, suggesting the use of additional cryptic BPs. Indeed it agrees with some of the earliest observations upon the effects of mutations upon mammalian BPs (reviewed in [41]). In this set of experimentally validated introns, our SVM classifier had a modest performance compared to the previous benchmark. This is mainly explained by the fact that the mapped BPs for these introns are significantly different from the U2 complementary sequence (TACTAACAC). Additionally, for CLK introns the prediction is further complicated by the fact that PPTs downstream of the mapped BPs contain many purine interruptions. However, with the exception of MBNL1 exon 9, in which the mapped dBP already ranked first according to the MM1, for all MBNL1 and CLK mapped BPs, it is possible to observe a raise in prediction ranking from pure motif score (MM1) to SVM score. This adds extra evidence suggesting the importance of the PPT in BP recognition, as its associated features account for all the difference between MM1 and the final SVM model. Finally, considering the limitations of our training and the large numbers of candidates in long AGEZs, our results show that the SVM classifier is capable of delivering a good set of predictions for introns with long AGEZs. In order to refine our understanding on the relation between the BP and the AGEZ, we extraordinarily extended our search region to the last 500 nts of each intron. For approximately 5% of the introns, no candidate was found in the AGEZ, and in a fraction of those (0. 44% of our human intron dataset) no TNA-containing 9-mers were found over the search region. These cases indicate the presence of a BP signal without the canonical TNA that, like for U12 type intron signals, will require independent modeling. On the remaining introns most of the best hits are located within the AGEZ towards the 5′ end of it. Interestingly, this distribution extends significantly up to 7/8 nucleotides upstream the AGEZ-defining AG-dinucleotide (a3 in Figure 1). Beyond this distance, a previous study in yeast has shown that the BP proximal AG can, though at a low rate, be chosen, therefore affecting the recognition of the distal 3SS [42]. The distribution profile shown in Figure 6A strongly suggests that the region r3 from Figure 1 might be shorter than it was previously assumed (12nts) and at the same time supports the initial assumption that the BP should be searched exclusively in the AGEZ. Even though inter-AG dinucleotide distance appears to be determinant for the packed arrangement of BP, PPT and 3SS at the end of introns, our results suggest that large BP-3SS distances (within the AGEZ) might be related with a decrease in splicing efficiency, reflected by the higher prevalence of exon skipping and lower inclusion levels observed for exons preceded by more distant BPs. In these cases, stronger BP sequences and longer PPTs do not appear to have any compensatory effect in acceptor site recognition. Additionally, it has been demonstrated that distant BPs provide the opportunity for regulated alternative splicing through the binding of repressive regulatory factors in the extended region between the BP and the 3SS [23], [24], [43], [44], which could serve as further explanation to why such exons are more frequently skipped. Considering that long AGEZs are indicative of distant BPs, it is interesting to observe (see Figure 17 in Text S1) that their number has been increasing throughout the mammalian lineage at a similar rate for almost all the branches considered. This suggests, not only that newly formed distant BPs might provide an opportunity for new regulated alternative splicing events, but also that this process might be of evolutionary relevance in mammals. Another striking observation from our data is the inverse relation between BP sequence score and exon skipping, suggesting that BP-U2 binding stability might be of considerable importance for the overall splicing efficiency. Related to this, we have found that long introns tend to have weaker BP signals, whereas small introns show the opposite behaviour with the BP signal appearing more clearly defined, i. e. they contain fewer putative candidates and these have a stronger motif. This might be intrinsically associated to differences in the contribution of the different pre-mRNA signals responsible for exon definition, which is considered to be the prevailing mechanism of spliceosome assembly in mammals. With introns accounting for the majority of the primary transcript length, exons are early recognized through exon-spanning interactions between factors and corresponding signals, resulting in the combined recognition of the 5′SS and the upstream 3′SS [45]. In the context of long introns, sequence features other than the BP might be playing a more prominent role in exon recognition, which can potentially alleviate some of the contribution of the BP to early spliceosome assembly and splice site recognition in these cases. Finally, previous studies have shown that AS is associated with exon creation [18], [35], [36]. It has been proposed that new exons are born with reduced splicing efficiency due to weaker splice sites and smaller differences between exonic and adjacent intronic content of splicing regulatory elements [18]. Here we explored the possibility that BP features might also be contributing to the low inclusion observed in recently created exons. In effect, our results suggest that the high rate of skipping observed for primate specific exons compared to mammalian ones results from a combination of poorly defined signals in the pre-mRNA, including the BP. Moreover, the weaker BP signals found in pseudo exons, with no inclusion evidence, underline the importance of the BP signal for accurate intron excision. Splicing is a remarkably complex mechanism. The final configuration of mature mRNAs depends on an elaborate crosstalk between splicing factors and a myriad of potentially competing signals in the pre-mRNA molecule. The accurate identification of splicing signals, specially those that directly participate in the splicing reaction, may prove useful in the context of large scale analyses focusing on the characterization of disease-associated genomic mutations, as many might be directly related with alterations in the normal splicing patterns. In this paper we present a new and more accurate method for BP prediction in mammalian introns and provide new insights on acceptor site architecture. Our data strongly suggest that the BP conceals information relevant for acceptor site recognition and, therefore, it should be integrated in future splicing models.
The genome sequences for 7 mammalian species (Homo sapiens – hg18; Pan troglodytes – PanTro2; Macaca mulatta – RheMac2; Mus musculus – mm9; Rattus norvegicus –RN4; Canis familiaris – CanFam2; and Bos taurus – BosTau4) and Refseq annotations for Homo sapiens were retrieved from the UCSC Genome Browser Database [46]. All introns preceding an internal exon and containing canonical splice sites were extracted from the annotation. After duplicate removal, there were 183187 unique introns in the human intron dataset. To obtain the corresponding orthologous introns in the other 6 species, the LiftOver tool [47] was used. Removing hit pairs that did not contain canonical splice sites or for which the flanks were in different strands and/or chromosomes, we obtained a set of 128790 orthologous introns in all 7 mammalian species. Additionally, we used three sets of introns preceding pseudo exons, primate specific exons and mammalian conserved exons, obtained from [18]. Pseudo exons were defined as sequence stretches of length comparable to real exons, intronic, located between apparently viable splice sites, not containing any termination codon in frame, and for which there is no evidence of inclusion. Additionally, if included in the mature transcript, they would not alter the reading frame. Out of the set of mammalian orthologous introns, the last 300 nt were aligned between all species using PRANK+F [48] with default parameters. Only introns of length greater or equal to 300 nt in all species were considered (N = 98996). By scanning the alignments we were able to retrieve all pentamer instances that were exactly conserved in the 7 mammalian species, which we refer to as conserved instances. Their positions in the human sequence were recorded. Next we proceeded to the identification of pentamers that had a distribution of their conserved instances similar to the one expected for BPs, imposing the presence of an A preceded by a T 2 bases upstream: TNANN, NTNAN and NNTNA, which account for 184 unique pentamers. Thus, we selected pentamers according to three tests: Using the above criteria, we were able to separate the initial set of 184 pentamers into three distinct groups based on their positional bias or lack of it: BP-associated, PPT-associated, and no association with any positionally biased signal. We first applied the two statistical tests 1) and 2) together. We discarded 37 pentamers that did not pass both, i. e. they do not show any positional bias in occurrence or conservation. We applied both tests simultaneously since some pentamers, like TCACG, TTACG or TAACG, would pass test 2) but their total count is very low and their occurrence in the range −55nt to −15nt might be just due to chance. These cases with low counts got discarded because they failed the test 1). Out of the remaining 147,23 had a peak outside the region of interest. Visual inspection of these 23 shows that they' re Py-rich and their bias region lies between −15 and 0, thus we label them as PPT-associated. We were thus left with 124 pentamers that do not present a uniform distribution of conservation in the last 300 bp of the intron, their occurrences are more frequent than expected in the region between −55 to −15, and their distribution peak is inside that same region. These were considered BP-associated. In order to build a set of putative branch points imposing minimal sequence bias, we devised a 2 step strategy based on positional bias and conservation. First, we identified all introns for which there is only one TNA conserved in all 7 species, as they will more likely include a BP candidate. Moreover, each unique instance must be positioned between 15 and 55 nts upstream the 3SS. For every instance, the human nonamer containing the TNA motif in the central position was collected – consTNA set. Second, from those instances, we only kept the ones overlapped by at least one of the previously determined BP-associated pentamers in every one of the species considered – consTNA-BP5 set. For every possible nonamer, containing a T and an A in positions 4 and 6, respectively, we used the program RNAcofold from the Vienna RNA package [49] to calculate its binding energy to the U2 snRNA. We forced the complete pairing of all nucleotides between the two sequences, with the exception of the BP adenosine, which was forced not to pair with any nucleotide from the U2 snRNA sequence. The energy of the base pairing depends on the complementarity between both sequences, the length of the sequence, and the sequence composition. If the energy is high (negative but close to zero), the base pairing is very unstable, because the complementarity of the sequences is poor. Conversely, if the energy is very negative, the base pairing is much more stable. Nonamers containing the same core region (5 central nucleotides) were grouped together and mean energy was computed for each cluster. For every column in the consTNA-BP5 set, the information content (IC) a measure of conservation was computed according to the formula: where is the probability of finding the nucleotide in position In order to test possible association between different BP positions we computed the Mutual Information (MI), using all human putative BPs in the consTNA-BP5 set, between all possible position pairs according to the formula: where is the probability of finding the nucleotide in position, is the probability of finding the nucleotide in position, and is the joint probability of simultaneously finding a particular combination of nucleotides in positions, respectively. According to the 1st order dependencies detected by MI, we modeled the BP signal making use of a position-dependent Markov model. Due to the fact that positions 4 and 6 are fixed as T and A respectively, to compute the conditional probabilities of their downstream positions, 2nd order dependencies were considered. Accordingly, the probability of occurrence of the nonamer in a model, omitting positions 4 and 6, can be represented as: where is the probability of finding the nucleotide in the first position (considered independent) of the model and is the conditional probability of finding the nucleotide in position of the model assuming nucleotide, in position. This was computed for every BP candidate, taking as reference a positive set (consTNA-BP5 set) and a negative set composed of randomly selected intronic nonamers that contained T and A in positions 4 and 6 respectively. The final motif score is given by the formula: and it reflects how likely a given sequence belongs to the positive set relative the negative set. A heuristic method was used in order to identify potential polypyrimidine tracts. Sequences were scanned by a Python program that finds all subsequences with the following characteristics, maximizing for length: For every predicted PPT a score was calculated based on the sequence length and content according to the following formula: Where is the absolute frequency of nucleotide in the PPT and, , and. This scoring scheme has been previously used in [50]. Due to the low information content, BP prediction cannot simply rely on the statistical modeling/scoring of the signal. In fact, there are other additional factors responsible for the recognition of the BP in mammalian introns. Accordingly, to build a predictive BP model, the sequence score was combined with PPT associated features using a Support Vector Machine (SVM) algorithm. The aim was to score the candidates based on the SVM score which is the distance in feature space between the candidate and the decision boundary. The consTNA-BP5 set was used as positive set for training. As negative set we picked all other nonamers, in this intron set, containing TNA in the central positions. For every candidate, 4 features were collected: These features and the classification (as positive or negative) served as input to SVMLIGHT [51]. For balanced learning, an equal number of positive and negative cases were used. The resulting predictive model was used to systematically score BP candidates. BP predictions for 183187 human introns can be found in http: //regulatorygenomics. upf. edu/SVM_BP/BP_predictions. tar. gz. Additionally, we have developed a web-tool where the algorithm can be run for multiple intronic sequences. The web-tool and a stand-alone version of the software are available at the URL http: //regulatorygenomics. upf. edu/SVM_BP/. The BP and the PY-tract of the PY7 reporter containing exon 2 and 3 of α-tropomyosin [52] were replaced with individual AGEZs by cloning them using following primers via XhoI and PvuII or AluI (restriction sites are underlined): MBNL1 Exon 6 forward 5′-GTGCTCGAGCCAATAACAACTCAGTAGTGCC; MBNL1 Exon 6 reverse 5′-TTATTAGCTTAATTAGCAGGCAGCGAGCAC; MBNL1 Exon 8 forward 5′- GTGCTCGAGGGCTTTTATTCTTCACTTGAGAC; MBNL1 Exon 8 reverse 5′- TTATTCAGCTGCCCATCATGCATTGCAAC; MBNL1 Exon 9 forward 5′- GTGCTCGAGTTTTTGACTTAGCATATTAAGCCTG; MBNL1 Exon 9 reverse 5′- CTTTCGGAGGGAAAATCATATAAGC (used for blunt end cloning to preserve suboptimal 3′ splice site); CLK1 Exon 4 forward 5′- GTGCTCGAGTTCAGTGAATGCTACAACTAAGC; CLK1 Exon 4 reverse 5′-TTATTCAGCTGGAAACGTCAAGTGGGCG CLK3 Exon 4 forward 5′- GTGCTCGAGGTTTTCTTTACATACCTGTAGCTG CLK3 Exon 4 reverse 5′- TTATTCAGCTGCATGCACCGCCCCCC Py7 constructs were linearized with XbaI prior to in vitro transcription with SP6 polymerase. In vitro transcription and splicing were carried out as previously described [23], [53]–[55]. 100 fmol of 32P-5′-labelled primer were hybridized to 100 fmol of spliced, debranched or control RNA template at the most 3′end of the intron and annealing was allowed for 30 minutes at 42°C. Lariat branch points were mapped by extending with 10 units of AMV reverse transcriptase (Promega) for 45 minutes at 42°C and by comparing the resulting terminations in the RT to the ones of debranched and control RNA. Primer extension reactions were loaded on 8% denaturing polyacrylamide gels side by side with sequencing reactions with the same primers and appropriate plasmid templates using T7 DNA polymerase. EST alignments were retrieved from UCSC Genome Browser Database [46] and compared with the annotations. For each exon, the percentage of EST inclusion level is defined aswhere is the number of ESTs including the exon and the number of ESTs that cover the genomic region of the exon but skip it. This measure was calculated for all the exons preceded by introns in the human intron dataset. Only exons with were considered, accounting for a total of 78186. Some exons have zero EST inclusion, as all the corresponding ESTs show exon skipping, but their existence is supported by mRNA evidence. | From transcription to translation, the events underlying protein production from DNA sequence are paramount to all aspects of cellular function. Pre-mRNAs in eukaryotes undergo several processing steps prior to their export to the cytoplasm. Among these, splicing – the process of intron removal and exon ligation – has been shown to play a central role in the regulation of gene expression. It has been estimated that more than half of the disease-causing mutations in humans do so by interfering with splicing. The difficulty in describing these disease mechanisms often lies in the low accuracy of the methods for prediction of functional splicing signals in the pre-mRNA. This is especially the case of the branch point, mainly due to its high sequence variability. We have developed a methodology for mammalian branch point prediction based on a machine-learning algorithm, which shows improved accuracy over previous published methods. Moreover, using a combination of experimental and bioinformatics approaches, we uncovered important positional properties of the branch point and shed new light on how some of its features may contribute to the final splicing outcome. These findings might prove useful for a better understanding of how splicing-associated mutations can lead to disease. | Abstract
Introduction
Results
Discussion
Methods | computational biology/alternative splicing
computational biology/genomics | 2010 | Genome-Wide Association between Branch Point Properties and Alternative Splicing | 13,228 | 279 |
Leptospirosis has emerged as an urban health problem as slum settlements have rapidly spread worldwide and created conditions for rat-borne transmission. Prospective studies have not been performed to determine the disease burden, identify risk factors for infection and provide information needed to guide interventions in these marginalized communities. We enrolled and followed a cohort of 2,003 residents from a slum community in the city of Salvador, Brazil. Baseline and one-year serosurveys were performed to identify primary and secondary Leptospira infections, defined as respectively, seroconversion and four-fold rise in microscopic agglutination titers. We used multinomial logistic regression models to evaluate risk exposures for acquiring primary and secondary infection. A total of 51 Leptospira infections were identified among 1,585 (79%) participants who completed the one-year follow-up protocol. The crude infection rate was 37. 8 per 1,000 person-years. The secondary infection rate was 2. 3 times higher than that of primary infection rate (71. 7 and 31. 1 infections per 1,000 person-years, respectively). Male gender (OR 2. 88; 95% CI 1. 40–5. 91) and lower per capita household income (OR 0. 54; 95% CI, 0. 30–0. 98 for an increase of $1 per person per day) were independent risk factors for primary infection. In contrast, the 15–34 year age group (OR 10. 82,95% CI 1. 38–85. 08), and proximity of residence to an open sewer (OR 0. 95; 0. 91–0. 99 for an increase of 1 m distance) were significant risk factors for secondary infection. This study found that slum residents had high risk (>3% per year) for acquiring a Leptospira infection. Re-infection is a frequent event and occurs in regions of slum settlements that are in proximity to open sewers. Effective prevention of leptospirosis will therefore require interventions that address the infrastructure deficiencies that contribute to repeated exposures among slum inhabitants.
Leptospirosis is a bacterial disease that has emerged as a major health problem in the developing world [1]. The disease is caused by a spirochete from the genus Leptospira, which colonizes the kidney of a wide range of mammals [2]. Human infection occurs after direct contact with an infected animal reservoir, or water and soil contaminated with their urine [1]. Infection produces a broad spectrum of clinical manifestations, which may lead from an asymptomatic and mild self-limiting febrile illness to severe disease forms with high case fatality [3]. Leptospirosis has traditionally been a sporadic rural-based disease associated with occupational risk groups such as subsistence farmers [3]. However, changes in human demography during the last 50 years have raised awareness of the emergence of leptospirosis as an urban health problem [4]. Rapid urbanization and urban poverty have led to the dramatic growth of slum settlements throughout low and middle-income countries [5]. To date one billion of the world' s population reside in urban slums; this population continues to expand at rates of 10% per year [5]. As a consequence of poor sanitation in these communities, slum residents are increasingly exposed and are at risk of acquiring water and animal-borne diseases [6], [7] such as leptospirosis [1], [4], [8]. In slum settings, endemic transmission of leptospirosis is largely due to circulation of a single serogroup, L. interrogans serogroup Icterohaemorrhagiae, [4], [7], [9], [10] for which the domestic rat is the maintenance host [4], [11]. In tropical urban environments, increased transmission and seasonal outbreaks occur during periods of heavy rainfall and flooding [4], [12], [13]. Furthermore extreme climatic events such as monsoons, typhoons and hurricanes have precipitated urban epidemics, as exemplified by the Mumbai outbreak in 2005 [14] and more recently, in the Philippines in 2009 [15] and Australia in 2011 [16]. Moreover, leptospirosis imparts a large disease burden as the cause of life-threatening infection among slum dwellers. In Brazil, more than 10,000 cases of leptospirosis are reported each year [17], the large majority of whom are residents of urban slums and require hospitalization [18] for severe complications of Weil' s disease and leptospirosis-associated pulmonary hemorrhage syndrome (LPHS) [1], [7], [19]. Overall case fatality is >10% among reported cases from Brazil [4] and >50–70% for cases that develop LPHS [7], [19]. However, severe disease represents a small fraction of the overall disease burden [20], [21] and to date, prospective studies have not been performed to identify the risk of leptospirosis among slum dwellers. Investigations of urban leptospirosis, which have used ecological [4], [12], [22], cross-sectional [9] and case-control study designs [23], [24], have identified infrastructure deficiencies in the environment where slum dwellers reside as risk factors for acquiring leptospirosis and anti-Leptospira antibodies. For example, high risk of Leptospira transmission has been found to be associated with proximity of residence to open sewers and accumulated refuse, flood-risk areas, and areas with high rat infestation [9], [12], [22]–[26]. In addition to environmental features, low socioeconomic status among slum residents contributes to the risk of leptospirosis [12], [22] and anti-Leptospira antibodies [9]. However these investigations are limited by the ecological study design, or in the case of cross-sectional surveys, the use of anti-Leptospira antibodies, which are detected in individuals up to four or more years after infection [27], [28]. To date, there are no studies that have attempted to evaluate prospectively the risk factors for leptospirosis among urban slum populations. We previously reported the findings of a large seroprevalence survey [9] of a slum community in Salvador, a city in Northeast Brazil where 33% of the population reside in slum settlements [29] and seasonal rainfall-associated epidemics of leptospirosis occur each year [4], [23]. This study found that a large proportion (15. 4%) of slum inhabitants had anti-Leptospira antibodies, suggesting that in addition to high rates of infection, repeated exposures with the Leptospira agent may be occurring in this high-risk population. Herein, we report findings from a prospective investigation of this urban slum population to determine the risk of Leptospira infection and identify risk associations for infection.
Participants were enrolled according to written informed consent procedures approved by the Institutional Review Boards of the Oswaldo Cruz Foundation and Brazilian National Commission for Ethics in Research, Brazilian Ministry of Health, Weill Medical College of Cornell University, and Yale School of Public Health. The cohort study was conducted in the Pau da Lima slum, a community situated in the periphery of Salvador (population, 2,675,656 inhabitants) [29], Brazil. The study site has been previously described [9]. Briefly, it comprised a four-valley area of 0. 46 Km2 with poor sanitation infrastructure. In 2003, the study team performed a census in the study site and identified 14,122 inhabitants residing in 3,689 households. The median household per capita income was US$ 1. 30 per day, and most (85%) of the studied population were squatters without legal title to their domiciles. A sample of 684 (18. 5%) households from a database of all enumerated households identified within the study site during the 2003 census was selected using a random number generator. Household sampling was chosen to facilitate follow-up evaluations and avoid excluding family members from a part of the study protocol in which other members are participating. The sample size of this sub-cohort was selected to detect a risk ratio of at least 2. 0 for exposure risk factors, and was guided by seroprevalence surveys in this community, which identified a seroprevalence of 15% [9], and case-control investigations that determined that the frequency of risk exposures for leptospirosis is between 20–40% in community individuals [23]. All participants who slept three or more nights per week in the sampled households and had five or more years of age were eligible for enrollment in the cohort study. Participants were enrolled between February 2003 and May 2004. In household visits during baseline cohort enrollment, the study team of nurse technicians, physicians and nurses administered a standardized questionnaire to obtain information on demographic and socioeconomic indicators, employment and occupation, exposures to sources of environmental contamination, and presence of potential reservoirs and domestic animals, including rats, chickens, dogs, and cats, in the household and workplace. Information on race was self-reported, and interpreted as a marker of socioeconomic status. The study team evaluated literacy according to the ability to read standardized sentences and interpret their meaning. Informal work was defined as work-related activities for which the participant did not have legal working documents. Frequent exposure to contaminated environment was defined by contact with mud, floodwater, garbage or sewage in the one-month period preceding data collection. Participants were asked to report the highest number of rats sighted within the household property and workplace site in the preceding one-month period. The head-of-household, defined as the member who earned the highest monthly income, was interviewed to determine sources and amounts of income for the household. The study team surveyed the area within <10 meters of the household to determine the presence of vegetation. Between September and October, 2004, the study team surveyed the study site to record the location of open sewage and rainwater drainage systems. We also mapped the sites of accumulated refuse and measured the area of these deposits. Geographic Information Systems (GIS) was used to obtain three-dimensional distance from the household to the nearest drainage systems and accumulated refuse, and to the lowest point in the valley (height) [9]. The study team collected blood samples from participants during the baseline survey and a follow-up survey conducted between October 2004 and January 2005. Sera were evaluated using the microscopic agglutination test (MAT) as previously described [9] to determine titers of agglutinating antibodies against a panel of five reference strains (WHO Collaborative Laboratory for Leptospirosis, Royal Tropical Institute, Holland) and two clinical isolates [4] The use of this reduced panel of strains, which represent five Leptospira serovars, Autumnalis, Canicola, Copenhageni, Ballum, and Grippotyphosa, demonstrated similar performance during laboratory confirmation of leptospirosis cases [4] and seroprevalence surveys [23], [30] in studies performed in Salvador, Brazil, as the use of the WHO-recommended panel of 16 reference serovars [31]. Screening was performed with serum dilutions of 1∶25,1∶50 and 1∶100. When agglutination was observed at a dilution of 1∶100, the sample was titrated to determine the highest agglutination titer. The absence and presence of a positive agglutinating antibody titer during the baseline survey was used to differentiate primary and secondary Leptospira infections which occurred during follow-up of the cohort. A primary infection was defined as seroconversion during which the MAT titer increased from negative during the baseline survey to a titer ≥1∶50 during the follow-up survey. A secondary infection was defined as a four-fold rise in MAT titer in a participant who had a titer of ≥1∶25 during the baseline survey. The MAT was repeated for samples of participants who were defined as having primary and secondary infections, in order to confirm their status. Epidemiological and laboratory data were double-entered using the Epi-Info for Windows software (Centers for Disease Control and Prevention, Atlanta, GA). There were no missing values for any of the analyzed variables. Data for individual participants were linked by location of residence to spatially coded information for households and environmental attributes within the study site. We used Chi-square and Wilcoxon rank sum tests to compare categorical and continuous data, respectively, between participants who were and were not selected to participate in the cohort, between participants who consented and did not consent to be enrolled in the cohort and between cohort participants who completed and who did not complete the study follow-up. A P-value of 0. 05 or less in two sided testing was used as criteria for a statistically significant difference. We calculated infection rates and 95% confidence intervals according to the Poisson distribution for primary, secondary and overall Leptospira infections, adjusting for the design effect of the household-based cluster sampling strategy. Only participants who completed follow up were included in the analysis. Rates were expressed as infections per 1,000 person-years of follow-up. We applied multinomial logistic regression models in both univariate and multivariate analysis to assess the relationship between explanatory variables and the occurrence of primary and secondary infections as compared to participants without evidence for incident serological infection. Interpretation of results was based on the odds ratio and 95% confidence intervals. Confounding and interaction between independent variables were evaluated by subgroup analyses prior to performing the logistic model. The results obtained in the univariate multinomial logistic regression models were confirmed using binomial logistic regression models, for which the outcome was independently primary infection versus no infection, and secondary infection versus no infection. Variables that had significant association at a P≤0. 10 in the univariate multinomial logistic model were selected to be incorporated into a hierarchical multinomial multivariate model [32] that accounted for hierarchical inter-relationships between variables and the potential underestimation of the effects of distal determinants. The hierarchical model grouped variables into three blocks; the first block contained socioeconomic variables, such as illiteracy, educational attainment, and per capita daily household income. The second block contained household variables, such as number of residents per household, time residing at the same household, household flooding, household distances to the lowest point in valley and to nearest open sewer, presence of vegetation in <10 meters from the household, and presence of potential reservoirs in the household. The third block comprised of the individual-level variables: gender, age, contact with floodwater, sewage water or trash, excavating or cleaning an open sewer, and risk related occupations. A multinomial backward elimination strategy was then performed for each block. Variables that reached a P value ≤0. 10 in each of the three blocks were then selected and grouped into a final block. Multinomial backward elimination was pursued on the final block of variables and those reaching a P value ≤0. 10 associated with one of the two types of infection were included in the final model. A P value <0. 05 was considered statistically significant.
Of the 14,122 inhabitants within the study site, 12,651 (90%) were eligible to participate in the cohort and 2,419 (19%) were randomly selected by household for study enrollment. Participants who were selected for cohort recruitment were similar to participants who were not selected in regards to median age (23 versus 24 years, respectively; P: 0. 38) and proportion of males (47% versus 48%, respectively; P: 0. 18). Among the 2,419 selected participants, 2,003 (83%) consented for the cohort study. Those who agreed to be a cohort member were younger than those who refused to participate (median age in years, 23 vs. 25, P: 0. 02) and were less likely to be male (44% versus 61%; P: <0. 001). Of the 2,003 enrolled participants, 1,585 (79%) completed the one-year follow-up study protocol. Participants were followed for a median of 306 days (minimum of 140 and maximum of 657 days). The major reason for loss to follow-up was moving to a household outside the study site (60% of the loss-to-follow-up participants). Participants who completed follow up differed from participants who did not in that they had a lower proportion of males (42% versus 51%; P: 0. 002), had a lower educational level (23% completed primary school versus 30%; P: <0. 01), and had a lower income (median household monthly income per capita US$ 39 versus 42, respectively, P: 0. 03). Overall, 51 (3. 2%) among the 1,585 participants who completed the follow-up had serological evidence for acquisition of Leptospira infection. None of the participants who had Leptospira infection reported having been diagnosed with leptospirosis in a health care facility, hospitalized for an acute febrile illness, or identified as a case of leptospirosis during active city-wide hospital-based surveillance for leptospirosis. Highest MAT titres were observed in agglutination reactions against serovars Copenhageni and Autumnalis for samples from 50 (98%) and 1 (2%), respectively, of the 51 individuals with confirmed infection. The overall crude Leptospira infection rate was 37. 8 infections per 1,000 person-years (95% CI: 26. 3–51. 9) (Table 1). The infection rate adjusted for age and gender of the eligible population did not significantly differ from the crude infection rates (data not shown). Infection rates were higher among the group with 15 to 24 years of age (47. 9 infections per 1,000 person-years; 95% CI: 24. 5–81. 3) and with 25 to 34 years of age (58. 1 infections per 1,000 person-years; 95% CI: 27. 4–103. 6). Males had 2. 12 (95% CI: 1. 22–3. 69) times greater risk of infection than females (54. 5 infections per 1,000 person-years [95% CI: 33. 8–81. 4] versus 25. 6 infections per 1,000 person-years [95% CI: 13. 9–41. 9], respectively) (Table S1). The gender difference in infection risk was most prominent in the group with 15–24 years of age (RR 3. 55,95% CI: 1. 28–9. 88, Table S1, Figure 1A). During the two-year period from the initiation of cohort enrollment to the end of the follow-up protocol, active hospital-based surveillance in Salvador identified five suspected cases of leptospirosis among the 12,651 inhabitants of the study site who were identified during the baseline census and were eligible to participate in the cohort study. Among suspected cases, four and one had a laboratory-confirmed and probable, respectively, diagnosis of leptospirosis. All cases had highest MAT titres directed against L. interrogans serovar Copenhageni. None of the leptospirosis cases was a member of the study cohort. Based on the eligible population at the site, the annual incidence for severe leptospirosis was 19. 8 (7. 2–43. 8) cases per 100,000 population. Of the 51 participants who had serological evidence for Leptospira infection, 35 had a baseline MAT titer equal to zero (1,126 person-years of follow-up) and 16 had a baseline MAT titer ≥1∶25 (223 person-years of follow-up), which was defined as a marker for a previous infection. The primary and secondary infection rates were 31. 1 (95% CI: 19. 9–45. 4) and 71. 7 (95% CI: 35. 8–123. 6) per 1,000 person-years, respectively. The risk for secondary infections was significantly higher than primary infection during follow-up of participants who were seropositive and seronegative, respectively, at enrollment (RR: 2. 31; 95% CI: 1. 30–4. 10). The age groups of 15–24 and 25–34 years had the highest secondary infection rates of 95. 2 and 134. 6 cases per 1,000 person-years, respectively, whereas the primary infection rates in these age groups was 37. 7 and 37. 2 infections per 1,000 person-years, respectively (p = 0. 094 and p = 0. 046 respectively, Figure 1B). Although secondary infection rates were similar for males and females (RR: 1. 39; 95% CI: 0. 54–3. 61), primary infection rates were significantly higher among males (RR: 2. 43; 95% CI: 1. 24–4. 78) (Table S1). The gender difference in primary infection rates was greatest for the group with 15–24 years of age; males had a more than four-fold greater risk (RR: 4. 22; 95% CI: 1. 14–15. 59) of acquiring primary infection more than females in this age group. Due to the difference in age and gender-specific rates for primary and secondary infections, the univariate multinomial models found young adults with 15–34 years of age to have significantly increased risk for secondary infection (OR: 8. 64; 95% CI: 1. 12–66. 65), but not for primary infection (Table 2). Male gender was associated with primary infection (OR: 2. 37; 95% CI: 1. 19–4. 74), but not with secondary infection. Socioeconomic variables, including illiteracy, low per capita household income, and lack of a CPF card (financial identification number) were found to be risk factors for primary infection, but not for secondary infection. Each one-dollar increase in the daily per capita household income decreased the odds of primary infection by 50% (95% CI: 0. 28–0. 89). In contrast to primary infection, we found that environmental attributes were significant risk factors for secondary infection in univariate analyses. Household flooding during rainy periods, proximity to open waste sewer, and three-dimensional distance of residence to the lowest point in the valley and open waste sewers had a stronger relationship with risk of secondary infection than primary infection. In contrast, risk behaviors that place participants in exposure to potentially contaminated environment near the household, such as contact with mud, sewage water, or garbage, and cleaning an open sewer, had a significant or near-significant association with both primary and secondary infections. Presence of rats at the place of residence, evaluated by the maximum number of rats seen, and sighting rats during the daytime, was not found as risk factor for primary nor secondary infection. Occupational factors, such as work that involved garbage collection and the maximum number of rats sighted at the workplace, were associated with increased risk for secondary infection in univariate analyses. The multivariate multinomial model identified male gender (OR 2. 88,95% CI, 1. 40–5. 91) and per capita household income (OR 0. 54 for an increase of $1 per person per day, 95% CI, 0. 30–0. 98) as independent risk factors for primary infection (Table 3). The model also identified age of 15–34 years (OR 10. 82,95% CI, 1. 38–85. 08) and proximity of the place of residence to the nearest open sewer (OR 0. 95; 0. 91–1. 00 for an increase of 1 m distance) as risk factors for secondary infection. Contact with mud in the place of residence was found to have a non-significant association with both primary (OR 1. 99,95% CI 0. 96–4. 12) and secondary (OR 2. 51,95% CI 0. 87–7. 23) infection. Occupation-related exposures were not found to be significant risk factors for primary and secondary infection in the multivariate analyses.
The findings of this large prospective investigation identified high rates of Leptospira infection among the study urban slum population, with more than 3% of the residents demonstrating serologic evidence of infection over a mean follow-up period of approximately one year. A single L. interrogans serogroup, serogroup Icterohaemorrhagiae, was the presumptive infecting agent, since highest agglutinating antibody titers were observed in 98% of participants with serologically-confirmed infection. Slum residents had an overall high risk for a repeat exposure and infection with the same agent. Furthermore, we identified that there are distinct risk factors for acquiring primary and secondary infection, suggesting that there exist sub-populations among slum residents who are repeatedly infected with leptospirosis. These findings highlight the potential large and unrecognized burden of leptospirosis in urban slum settlements. There have been no comparable studies performed in slum settings that have followed large numbers of community-based participants and prospectively ascertained outcomes with standard serologic methods. A longitudinal study had been performed in Iquitos, Peru and found that annual incidence of Leptospira seroconversion was 288 per 1,000 persons based on IgM ELISA seroconversion during follow-up of 158 urban slum residents [33]. The much higher incidence observed in Iquitos may be due to the use of the IgM ELISA rather than the standard MAT method [33], which is more specific in detecting exposure to pathogenic leptospires [1], [3]. Alternatively, the difference may reflect differences in the frequency of infection among slum settlements that have distinct ecological and socio-economic characteristics. A key knowledge gap in leptospirosis centers on the natural history of the disease and specifically, the proportion of infections that progress to develop disease and severe outcomes in the setting of high endemic transmission [1], [3]. Leptospirosis cases were not identified among the sample of 2,003 participants who participated in the cohort study to obtain direct estimates of the infection-to-severe-disease ratio. However, active hospital-based surveillance of the 12,651 community members that were identified during the baseline census found that the annual incidence of severe leptospirosis was 19. 8 cases per 100,000 population at the study site during the cohort follow-up period. Comparison of this estimate of the severe disease incidence and the infection rate among cohort participants suggests that the infection-to-disease ratio may be as high as 191∶1 (95% CI, 82–542∶1). Although cohort participants with documented seroconversion did not report being hospitalized or visiting an ambulatory clinic for leptospirosis during follow-up, it is plausible that a significant proportion developed sub-clinical illness or clinical disease, which would not be identified and diagnosed as leptospirosis unless cases developed classic severe manifestations [1]. The high infection-to-disease ratio suggests that like dengue and other causes of acute fever in tropical urban environments, the burden of leptospirosis is under-recognized and significantly greater than reflected by reporting of severe cases. The study findings indicate that the same L. interrogans serovar causes asymptomatic and sub-clinical infections as well as severe disease in this urban slum population. Among cohort participants with documented seroconversion, 98% had highest MAT titers that were directed against L. interrogans serovar Copenhageni, indicating that the Icterohaemorrhagiae serogroup was the infecting agent. We have observed that during long-term hospital-based surveillance in Salvador [4], [34], [35] and the study site community, >95% of the severe leptospirosis cases had highest MAT titers directed exclusively against the same serovar. L. interrogans serovar Copenhageni has been the sole serovar from serogroup Icterohaemorrhagiae to be isolated from this patient population during long-term surveillance [4], [7], as well as rat populations from the study community [36] and the city of Salvador [10], [37]. Together these findings indicate that transmission of leptospirosis is due to a single circulating serovar and provides additional evidence that Rattus norvegicus, the most common host of L. interrogans serovar Copenhageni [3], [10], is the principal reservoir in this urban slum setting. Moreover, our findings raise an important question with respect to what specific factors influence disease progression and the diverse range of clinical outcomes after infection with a single serovar agent. These factors may relate to strain-specific differences within the serovar that contribute to the strain' s virulence, or alternatively, host-specific susceptibility or resistance factors and types of environmental exposures that contribute to the inoculum dose during infection [1], [3]. Re-infection was a frequent event among cohort participants during follow-up and raises the issue of whether natural infection confers immunity to a subsequent infection with a homologous serovar. Although there is clear evidence that immunization with live attenuated [38] and killed leptospires protects experimental animals against lethal infection [3], naturally-acquired immunity to re-infection in humans is poorly understood due to the limited number of prospective studies in well-characterized populations. A study from rural Andaman Islands used similar serologic criteria as employed in this study to prospectively identify Leptospira infection among school children and found that that primary infection rates were higher than secondary re-infection rates [39]. Additionally, the study observed a non-significant association between increased morbidity during follow-up and seronegative status during the baseline survey [39]. Although those findings are suggestive that a previous infection may protect against a subsequent infection, the conclusions were limited by the small numbers of clinical cases identified and potential confounding due to multiple circulating serovars. Our study found that in an urban setting of transmission of a single serovar agent, prior exposure and infection did not confer complete protection against a subsequent serologically-ascertained, asymptomatic or sub-clinical infection with the same serovar. However, we could not determine whether prior infection protects an individual against developing clinical disease during subsequent re-infection since leptospirosis cases were not identified among cohort participants, nor could we evaluate the temporal relationship between initial infection and subsequent re-infection events due to the short follow-up period. Further prospective investigation is therefore needed to elucidate this question, which has major implications for development of an effective vaccine for leptospirosis in humans. Urban slum residents who acquired a primary and secondary infection had some similar risk associations in common, yet also had important differences in infection rates and type of risk factors. Male gender and low socioeconomic status were independent risk factors for primary and secondary infections, although the associations with secondary infection were non-significant. Although male gender has not been identified as a risk factor for Leptospira infection in cohort studies performed in rural or mixed settings [33], [40], [41], males have a significantly higher risk for leptospirosis and anti-Leptospira antibodies in population-based surveillance [4], [23], [24] and seroprevalence surveys [9], [30], respectively, in Salvador and other urban settings [42]–[44]. We found that for each one-dollar increase in the daily per capita household income, the odds of primary and secondary infection decreased by 46% and 48%, respectively. Poverty and low socioeconomic status may contribute to infection risk through diverse mechanisms that include psychosocial processes that promote risky behaviors and exposures with a contaminated environment, the limited use of protective clothing against abrasions that facilitate entry of the Leptospira spirochete [21], lack of access to amenities and social support [45] and inadequate household sanitation conditions. A previous study in the same area demonstrated that adolescents and individuals who did not complete primary school had lower levels of knowledge and practices regarding leptospirosis [46]. Our prospective study confirms the findings of a seroprevalence survey performed in the same community [9] that relatively small differences in socioeconomic level, independent of poor environment, influence the risk for leptospirosis in slum populations characterized by overall high levels of absolute poverty. The secondary infection rate was more than twice that of the primary infection rate (RR: 2. 31; 95% CI: 1. 30–4. 10) among cohort participants, indicating that seropositive status at the baseline survey was a marker for increased risk for infection. In contrast to primary infection, individuals who had 15–34 years of age had significantly higher risk for acquiring a secondary infection (Table 3 and Figure 1), suggesting that a sub-population of young adults are repeatedly infected with pathogenic Leptospira. Furthermore we found that residence in proximity of an open sewer was significantly associated with an increased risk of primary infection and not secondary infection. It seems plausible that open sewers are a risk factor for primary infection, but the magnitude of this risk association was lower than that for secondary infection and not detected in this study. Together these finding suggest that there are distinct environmental settings and behavioral factors that contribute to repeat exposures. Seasonal flooding is a frequent occurrence in slum communities in Salvador and especially for households situated on the poorest land quality at the bottom of valleys. Ganoza and colleagues found that environmental surface water from urban slums in Peru contained high concentrations of L. interrogans serovar Icterohaemorrhagiae [47]. Previous cross-sectional and case-control studies found that residence in flood-prone areas and in proximity to open sewer and rainwater drainage systems were associated with increased risk for anti-Leptospira antibodies and leptospirosis [9], [12], [23], [26]. The findings of this prospective study suggest that these infrastructure deficiencies of slum settlements also serve as transmission sources for repeated exposures to the Leptospira pathogen. Furthermore the findings demonstrate that adolescents and young adults are the primary risk group for repeated exposures and re-infection and indicate the need for specific interventions that target this high risk group. This study has several limitations which need to be considered. The proportion of females and younger participants among study participants was greater than among non-participants. Infection rates that were adjusted for the age and gender distribution of the eligible population did not differ from crude rates, indicating that differences between enrolled and non-enrolled participants may have not introduced a significant bias in rate estimates. Among enrolled participants, 21% did not complete follow up due primarily to out-migration. This sub-group had a higher proportion of males, was better educated and had a higher income in comparison to those who completed follow up. Although we could not fully address the potential for bias, predictive factors for loss-to-follow-up were included in our modeling approach and the estimates for the risk associations may therefore be valid approximations. In addition, our findings may not be broadly generalizable since the study was performed in a single slum community in Brazil. The incidence of Leptospira infection and risk associations identified in our study is expected to vary given the differences in underlying conditions of social deprivation, environmental degradation and climate where urban slum communities are situated. However the study site is a typical slum community in the city of Salvador and Brazil, where respectively, 33% [29] and 28% [5] of the population inhabit in such settlements. Furthermore, a large proportion of the one billion inhabitants of urban slums, as defined by the UN-HABITAT [5], reside in communities that have the same features of poverty, climate, poor environment and inadequate access to sanitation as found in our study site. The findings of this study may therefore be highly relevant to the situation of leptospirosis in urban slum settlements in developing countries and across tropical regions. Furthermore, the study offers insights on the approaches needed to effectively address this neglected disease as the world' s population of slum dwellers doubles to two billion by 2020 [48]. The study provides the first prospective evidence to support the assertion that defined infrastructure deficiencies in slum communities serve as transmission sources for leptospirosis. Removal of these sources, through implementation of adequate closed sewage and drainage systems, should therefore be a public health priority. Our findings also highlight the importance of adolescents and young adults as a risk group for spill-over infections. Efforts need to be made to identify and target through intervention the risky behaviors in this age group that promote recurring exposures with environmental contamination. Similarly, further work is required to identify the processes by which the social gradient of status influences unequal health outcomes within slum populations living in absolute poverty. By elucidating such mechanisms, we may not only identify effective prevention for leptospirosis, but may also identify common processes and interventions for the large range of communicable and non-communicable diseases that affect marginalized urban communities. | Leptospirosis is a disease that is transmitted by human contact with an environment contaminated with urine from animals, such as rodents, infected by the Leptospira bacteria. Human illness due to these bacteria can be mild, or can have very severe complications. Residents of urban slum settlements are at high risk for this disease, but the specific risk factors for transmission in these settlements are not understood because of the lack of prospective studies in this epidemiological setting. We performed a prospective study in a Brazilian slum community to measure the risk of infection, identify the environmental and social factors that place slum residents at risk for infection, and determine whether some individuals are at risk of repeated infections. We identified a burden of infection with leptospirosis among slum residents, and found that male gender and low income both increase the risk for infection. In addition, a significant proportion of slum residents had a second exposure to leptospirosis and re-infection occurred most frequently among young adults and the poorest members of the slum community who reside in proximity of open sewers. These risk factors are amenable to interventions aimed to reduce the burden that leptospirosis imparts in this high-risk setting. | Abstract
Introduction
Methods
Results
Discussion | leptospirosis
bacterial diseases
infectious diseases
veterinary diseases
zoonoses
medicine and health sciences
environmental epidemiology
epidemiology
infectious disease epidemiology
neglected tropical diseases
biology and life sciences
population biology
spatial epidemiology
tropical diseases
veterinary science | 2014 | Prospective Study of Leptospirosis Transmission in an Urban Slum Community: Role of Poor Environment in Repeated Exposures to the Leptospira Agent | 8,373 | 278 |
Multispecificity–the ability of a single receptor protein molecule to interact with multiple substrates–is a hallmark of molecular recognition at protein-protein and protein-peptide interfaces, including enzyme-substrate complexes. The ability to perform structure-based prediction of multispecificity would aid in the identification of novel enzyme substrates, protein interaction partners, and enable design of novel enzymes targeted towards alternative substrates. The relatively slow speed of current biophysical, structure-based methods limits their use for prediction and, especially, design of multispecificity. Here, we develop a rapid, flexible-backbone self-consistent mean field theory-based technique, MFPred, for multispecificity modeling at protein-peptide interfaces. We benchmark our method by predicting experimentally determined peptide specificity profiles for a range of receptors: protease and kinase enzymes, and protein recognition modules including SH2, SH3, MHC Class I and PDZ domains. We observe robust recapitulation of known specificities for all receptor-peptide complexes, and comparison with other methods shows that MFPred results in equivalent or better prediction accuracy with a ~10-1000-fold decrease in computational expense. We find that modeling bound peptide backbone flexibility is key to the observed accuracy of the method. We used MFPred for predicting with high accuracy the impact of receptor-side mutations on experimentally determined multispecificity of a protease enzyme. Our approach should enable the design of a wide range of altered receptor proteins with programmed multispecificities.
Many natural proteins, including signal transduction hubs and enzymes that process biological information, have evolved to be multispecific–they participate in specific interactions with several interaction partners [1,2]. Evolution of multispecificity includes selection for both positive and negative specificity, involving recognition and non-recognition, respectively, of sets of interaction partners [3]. Most multispecific interactions arise when the active site of a single receptor protein interacts with multiple binding partners of differing sequence [4]. Nature uses structurally conserved protein-recognition domains (PRDs), e. g. , SH2, SH3 and PDZ domains, to mediate many multispecific interactions [5–10]. Thus, it is crucial that methods that model and modulate PRD specificity are able to accurately recapitulate their multispecific nature. Similar to cascades composed of multispecific PRDs like SH3, SH2 and PDZ domains that mediate signal transduction, proteolytic cascades are ubiquitous in the post-translational transduction of biological information [11]. Protease activity and selectivity is involved in a diverse range of biological processes including digestion, blood clotting, apoptosis and cancer [12–15]. Proteases are inherently multispecific such that they recognize and proteolyze (or cleave) a range of substrates (positive specificity) while not recognizing others (negative specificity) [3]. For example, viral proteases such as HCV protease that are involved in viral maturation cleave only specific sites in the viral polyprotein but do not cleave others [16]. These proteases may also have evolved the ability to cleave specific host proteins [17]. Prediction of protease multispecificity is, therefore, key for identifying their substrates under healthy and disease conditions. Additionally, designed proteases with programmed multispecificity have the potential to be used as therapeutics and protein-level knockout reagents in cell culture [18]. The ability to manipulate protease specificity computationally would enable the creation of such designer proteases with dialed-in recognition specificity, thereby providing tools to interrogate and intervene in biological processes. Rational modulation of protein-protein or protein-peptide interaction multispecificity has met with limited success, except in a few notable cases, such as coiled-coil interfaces [19,20]. In principle, computational structure-based modeling methods should be able to recapitulate and modulate multispecificity. In fact, several methods relying on, among others, Monte-Carlo (MC) simulations in sequence and conformation space, and genetic algorithms (GA) have been developed to predict PRD multispecificity [21–25]. However, these methods are limited by the time required to enumerate a sufficiently large number of sequences to sample the substrate/peptide sequence space. As multispecific design entails additional sampling of (thousands) of receptor variants and modeling the multispecificity of each variant separately, using current methods to design receptors for and against specificity profiles is not computationally feasible. We have developed a structure-based method that eliminates the expense of explicit sequence enumeration in multispecificity modeling. The method uses a self-consistent Mean-Field theory-based Prediction (MFPred) approach that expresses specificity as a sitewise probability distribution function that can be calculated relatively rapidly. We have benchmarked MFPred on four diverse proteases and compared the results to MC- and GA-based methods. MFPred has comparable accuracy to MC-based and GA-based methods and provides a tens- to thousands-fold speedup. We demonstrate the generality of MFPred by obtaining significant multispecificity predictions for five diverse classes of protein-recognition domains (PRDs). Finally, as a proof-of-concept for design, we demonstrate that MFPred can recapitulate experimentally determined changes in specificity profiles due to receptor-side mutations.
To predict the specificity profile, we consider an ensemble of peptide backbone conformations bound to a receptor. For each peptide backbone conformation, we simultaneously sample all rotameric conformations of all amino acids at all peptide residue positions while keeping the receptor backbone and sidechains in their crystallographic conformations. The sidechain conformations at a given peptide position are sampled in the “mean field” of all other sidechain conformations at all other positions and (fixed) receptor residues, as described in Methods. Next, the contribution of each peptide backbone conformation at each peptide position is accounted for by Boltzmann averaging the mean-field specificity profile solution obtained in the previous step. The final specificity profile is constructed by combining these individual predictions. While the sequence specificity prediction described here can be performed using any (pairwise decomposable) energy function, we implemented our prediction method in the context of the Rosetta modeling suite, thus combining its sophisticated energy function with the speed of mean-field sampling (Fig 1). To test our MFPred method, we sought to first recapitulate experimentally determined specificity profiles of a variety of PRDs. We chose PRDs where both structural as well as specificity information has been experimentally determined. We focused primarily on protease enzymes for methodology development, and tested the generality of our approach with previously developed benchmarks for multispecificity prediction on PRDs such as a kinase enzyme, and SH3, SH2, MHC, and PDZ domains. We evaluated the performance of MFPred by quantifying the differences between predicted and experimentally determined specificity profiles using several metrics (see S1 Note for detailed descriptions of these metrics). Four of these metrics, the cosine similarity, Frobenius norm, average absolute distance (AAD) and Jensen-Shannon divergence (JSD) are correlated, as shown in S3 Fig. The Frobenius norm and AAD are distance-based metrics that have been used previously to compare profiles [21,22]. The Frobenius norm is more sensitive to flatness in the specificity profile than the AAD (S4 Fig). Additionally, we evaluated the profiles by their cosine similarity, which is another distance-based metric that is less sensitive to flatness than either AAD or Frobenius norm. It falls between 0 and 1, where 0 denotes a random prediction and 1 denotes a perfect prediction. The Jensen-Shannon divergence (JSD) has also been used in the past to evaluate profiles [21] and is less distance-based. We used cosine distance as the general score of a profile, as it is easy to visualize and interpret. It falls between 0 and 1, where 0 denotes a random prediction and 1 denotes a perfect prediction. For each position, we evaluated the significance of its JSD score by scoring 100,000 random profiles against the experimental profile and thus determining the p-value of the JSD score (see S1 Note for details). We also used a second metric as a general score for each profile: area under the ROC (receiver operating characteristic) curve (AUC) is a non-distance-based metric that evaluates predictions based on their ranking more tolerated amino acids correctly [22]. It is relatively unaffected by flatness (S4 Fig) but will not evaluate well if either the experimental or predicted profile is close to uniform. It is not correlated with the above metrics. Additionally, we developed a new metric, Score Sequence AUC Loss (SSAL), which encapsulates the efficacy of the predicted specificity profile in differentiating between substrates which are recognized and cleaved by a given protease (cleaved sequences) and substrates which are not cleaved by that protease (uncleaved sequences). A perfect prediction scores an SSAL of zero. It does not correlate well with any other metric (S3 Fig). Proteolysis is a multi-step reaction which involves substrate peptide binding, the formation of a tetrahedral intermediate (acylation) and hydrolytic cleavage of the tetrahedral intermediate (deacylation). We have previously found that modeling a near-attack conformation for the acylation step was successful in discriminating between known cleaved and uncleaved peptides [31]. Therefore, starting from structures of protease-substrate complexes in a near-attack conformation, we performed MFPred-based specificity prediction. We found that MFPred robustly recapitulates protease specificity profiles (Fig 2B) in our benchmark set. The cosine similarities of the entire profiles range from 0. 66 to 0. 89, AUC ranges from 0. 73 to 0. 86, and SSAL ranges from 0. 21 to 0. 002. Out of 31 substrate positions across the protease dataset, 20 were predicted with a significant JSD p-value. The best prediction is obtained for the common biotechnologically used protease TEV-PR. The predicted profile has a high cosine similarity of 0. 89 (1 would be a perfectly accurate prediction). The primarily steric and hydrogen-bonding-based nature of molecular recognition at TEV-PR-substrate interfaces is well suited to the strengths of the Rosetta energy function underlying MFPred. Similarly, the profiles of HCV protease and granzyme B (GrB) protease are also generally recapitulated with a high degree of accuracy, except for positions with no marked preference for specific amino acids (flat positions) –positions P5 and P2 in HCV protease and positions P4, P1’, and P2’ in granzyme B protease. We attribute the lack of correlation at these flat positions to small errors in energy evaluations being equivalent to the size of the energy gaps being modeled, thus leading to erroneous ranking. Challenges in measuring prediction accuracy at flat positions have indeed been noted before [22]. The worst performance among the proteases in the benchmark set is observed for the prediction of HIV protease-1 (HIVPR1) specificity. This protease is known to have a relaxed specificity profile, with preference for small hydrophobic residues at P1 and P1’ positions. The cavity of HIV protease-1 is large and peptides may adopt large variations in backbone conformation depending on their sidechains. Additionally, substrate binding involves flexibility on the protease side, with two loops (“flaps”) that are mobile and close over the binding pocket. Incorporation of greater backbone flexibility on both the receptor and peptide parts of the HIVPR1-peptide interface may help improve predictions, as previously observed by us and others [31–33]. To determine the contribution of modeling backbone flexibility to the accuracy of prediction and to investigate if backbone sampling could be optimized for specificity prediction, we generated MFPred profiles with different levels of backbone flexibility. First, we found that predictions generated by starting from a single crystallographically-determined backbone structure for the peptide led to poor accuracy for HCV and HIV proteases (panels f, h in S6 Fig), indicating that incorporating peptide backbone diversity is a key requirement for the observed accuracy of prediction. Second, we generated peptide backbone ensembles by threading on a varying number of known substrate (cleaved) peptides using three different Rosetta-based backbone sampling protocols (FastRelax [34], FlexPepDock [35], and Backrub [36]) separately to further diversify the peptide backbone ensemble. In each case, geometric constraints [31] were used to limit the scissile peptide bond to a near-attack conformation and the catalytic residues to an active conformation. The MFPred simulations were then performed on all backbone ensembles and their results were compared to each other (Fig 2). While the algorithm is relatively robust to the method of backbone generation as long as scissile bond geometry is maintained, the FastRelax (FR) protocol has a small improvement in overall performance over the FlexPepDock (FPD) protocol, with 20 significant p-values (out of 31) for FR vs. 19 for FPD, and FPD has a minor increase in overall performance over Backrub (BR), with 19 significant p-values for FPD vs. 18 for BR. The profile for TEV-PR is predicted best by FR, due to better prediction of Q at P1 and S at P1’. In the case of HIV protease-1, FR recapitulates the profile better than FPD and BR do. However, the performance of FPD is marginally better than that of FR and significantly more accurate than that of BR in the cases of HCV protease and granzyme B protease. To determine how MFPred accuracy depends on the number and sequences of known cleaved substrates used to generate the backbone ensemble, we generated a peptide backbone conformational ensemble that was independent of peptide sequence. For all positions on the peptide backbone, we enumerated every combination of phi/psi dihedral angles that were x-15, x, and x+15, where x is the dihedral angle of the relaxed crystal structure peptide backbone. The resulting structures were filtered to remove those with clashes and to preserve hydrogen-bond interactions. The remaining structures were further clustered by all-heavy-atom RMSD of the peptide residues (see S2 Note for details) and MFPred was performed on the cluster centers. The resulting predictions are significantly less accurate than those of FR, FPD, or BR (S5 Fig), indicating that successful prediction requires a backbone ensemble that is optimally positioned in the binding site for cleavage. As a second test of the dependence of MFPred on the cleaved sequence information, we threaded five known uncleaved (i. e. , not bound by the protease in a productive conformation) sequences on the peptide backbone and then performed FastRelax on the resulting structures. The prediction accuracy of MFPred decreased on these structures (S5 Fig), to the extent that the specificity profiles are almost uniform. Therefore, diversifying the peptide structure in suboptimal sequence space led to worse predictions than those obtained while diversifying it without any sequence information. Next, to determine the impact of starting from bound complexes to generate MFPred predictions, we performed MFPred simulations on apo structures of two proteases: HCV NS3 protease and HIV protease-1 (S12 Fig). As HIV protease-1 has two flaps that can assume either a closed or open form [37], we used both a ‘closed apo’ structure and an ‘open apo’ structure for our simulations. In each case the protease all-atom RMSD between bound and open states, as determined by PyMol [38], were 1. 04 Å, 1. 85 Å, and 2. 00 Å. In all three cases, MFPred accuracy was higher when starting from the bound complex compared to the apo state. While the number of significant p-values remains similar, the overall cosine similarities, AUC, and SSAL decreased for the apo structure-based simulations. Additionally, the information content decreased significantly for the apo structures of HIV (0. 72–0. 74 bits) as opposed to the bound complex (1. 18 bits). Overall, the prediction accuracies between apo and bound states were more similar for the HCV protease where small backbone changes in the protease are incurred upon binding, compared to HIV protease where larger differences in prediction accuracy were apparent. These results suggest that especially in cases where there is significant backbone conformational change in the receptor upon peptide binding, such as the HIV protease, the incorporation of receptor flexibility may be needed for maintaining MFPred accuracy. Finally, to investigate the dependence of performance accuracy on the number of known cleaved (recognized) sequences, we executed MFPred simulations on backbone ensembles generated from differing numbers of starting peptide sequences threaded on to the crystallographic backbone conformation. We varied the number of sequences used to generate the backbone ensemble from one sequence to five sequences to ten sequences to all known sequences in the benchmark set. We found that MFPred is highly dependent on N, the number of cleaved sequences used, when N is small (panels e-h in S6 Fig). However, as N increases, this effect is decreased. For TEV-PR and HCV protease, which have relatively few sequences (68 and 198 respectively), the prediction accuracy plateaus after ten sequences, although in some cases it may fluctuate slightly from five to ten to all sequences. However, for granzyme B and HIV proteases (356 and 374 cleaved sequences respectively), the accuracy of MFPred has a minor increase from ten to all sequences. Thus, there is a near-maximum of accuracy for each system; once that point of diminishing returns has been reached, incorporating more cleaved sequences does not lead to significant increases in the accuracy. Besides determining that the level of backbone sampling was optimal for prediction, we also optimized sidechain sampling (S3 Table). Using an older version of the rotamer library (2002) [39] decreased scores for all systems. Increasing the fineness of rotamer chi-angle sampling or removing the starting sidechain conformation from the rotamer sampling had little impact on the results. Packing protease sidechains around the peptide (between distances of 4–8 Angstroms) decreased the accuracy of the results. This may be explained by the finding that hot spot residues at protein-protein interfaces often adopt strained rotamer configurations [40]; packing protease interface sidechains while designing peptide residues within MFPred may force protease sidechains to adopt conformations that are unfavorable for productive substrate binding. We compared our results to the two previously developed methods for specificity prediction that have been implemented in the Rosetta software. MFPred performed with comparable or greater accuracy than the sequence_tolerance [22] and pepspec [21] methods (Table 1). Additionally, MFPred was between 23-fold to 120-fold faster than the pepspec method and between 154-fold to 1154-fold faster than the sequence_tolerance method, depending on the number of peptide backbone conformations and rotamers (Table 1). For comparative benchmarking purposes, simulations were performed using a single AMD Opteron 6276 2. 3 GHz processor. Furthermore, MFPred is more accurate on single backbones and smaller backbone ensembles than the other two methods; when performed on a backbone ensemble generated from five substrate sequences, MFPred predicts 19 out of 31 positions with a significant p-value, whereas only 11 of the positions predicted by sequence_tolerance and 8 of the positions predicted by pepspec yield significant p-values (S7 Fig). When executed on a single backbone conformation, MFPred predicts 12 positions with a significant p-value, while both sequence_tolerance and pepspec predict only 8 positions with a significant p-value. Both sequence_tolerance and pepspec are designed to be used with larger peptide ensembles–their success is dependent on a diverse backbone ensemble–and, as expected, their prediction accuracy increases as the number of backbones in the ensemble rises (Fig 3A–3D), with sequence_tolerance predicting 15 significant positions and pepspec predicting 16 significant positions on the backbone ensemble generated from all cleaved sequences (S8 Fig). When performed on this expanded backbone ensemble, MFPred prediction accuracy was also higher, with 25 significant predictions. Thus, compared to two state-of-the-art existing methods, MFPred-based predictions are of comparable or higher accuracy, and can be obtained with 10-1000-fold higher computational efficiency. Besides informing us about the accuracy and speed of MFPred relative to existing methods, the comparison of MFPred to pepspec and sequence_tolerance allows us to categorize inaccuracies in MFPred predictions into those obtained from incorrect sequence sampling and those due to the Rosetta energy function or incomplete backbone conformational diversity. For example, MFPred on all cleaved backbones does not recover the experimentally determined high frequency for G at P2 of TEV-PR. Since both pepspec and sequence_tolerance also do not recover G at P2 with the same peptide backbone conformational ensemble, we attribute this inaccuracy to imperfections in the underlying Rosetta energy function and/or an incomplete peptide backbone ensemble used for prediction. Generally, MFPred predicts lower information content (i. e. flatter shape) for the profiles than both sequence_tolerance and pepspec (Table 1, Fig 3E–3H). In the cases of granzyme B protease and HIVPR1, the predicted lower information content is reflective of the experimentally determined profiles; however, in the case of TEV-PR MFPred underestimates the information content relative to pepspec and sequence_tolerance. All protocols underestimate the information content of the profile of HCV protease. This underestimation may be due to an incomplete experimental dataset or sampling/scoring inaccuracies as discussed above. Overall, the difference between the predicted information content and the experimental information content was smaller for MFPred than for sequence_tolerance and pepspec, especially when performed with smaller backbone ensembles. To investigate the generality of our method for specificity prediction, we utilized the MFPred method to predict the specificity profiles for a variety of peptide-recognition domains: kinase, SH2, SH3, PDZ, and MHC domains. We achieved 17 significant p-values out of 31 positions and high cosine similarities (0. 77–0. 85) for three out of five PRD classes: PKA (kinase), Src (SH2), and c-Crk (SH3) domains (Fig 4). However, these three systems had lower AUCs (0. 60–0. 65). This may be due to the inadequacy of AUC as a metric for scoring positions that have low information content in the experimentally-derived profile; if few of the experimental amino acid frequencies are greater than 10%, the AUC reveals little about the prediction accuracy. We predicted the specificity profiles of seven different PDZ domains: NHERF-2 PDZ2, PSD-95, AF-6 PDZ, Erbin PDZ, MPDZ-13, ZO-1 PDZ1, and DLG1-2 PDZ (Fig 4, S10 Fig). The specificity of NHERF-2 PDZ-2 was already predicted computationally by Zheng et al. [41], who were able to achieve good prediction via the use of CLASSY and FlexPepDock. King and Bradley previously predicted the specificity profile for PSD-95 computationally using pepspec [21], while the five other PDZ domain specificities were previously predicted by Smith and Kortemme via sequence_tolerance [22]. Six out of seven PDZ domains were predicted with medium to high accuracies, with cosine similarities of 0. 63–0. 86, AUCs of 0. 60 to 0. 88, and 25 out of 38 significant p-values. However, the prediction accuracy of the final PDZ domain, AF-6 PDZ was much lower, with a cosine similarity of 0. 43, AUC of 0. 59, and no significant p-values. This low accuracy may be due to the flexibility of the AF-6 PDZ domain, which has been known to bind in multiple binding modes and can be characterized as belonging to multiple classes of PDZ domain specificity [42,43]. Similar to the HIVPR1 case above, addition of receptor flexibility to MFPred may assist in AF-6 specificity profile recapitulation. Finally, we tested the performance of MFPred on predicting MHC-I peptide recognition specificities. We selected four MHC-I domains with crystallographic structure availability and a large pool of known peptide binders [44]. The experimentally derived specificity profiles for the MHCs were highly conserved at one or two positions but relatively flat at others (Fig 4, S11 Fig). The MFPred predictions reflected this pattern: while 30 out of 36 positions had p-values that were not significant, due to the high tolerance of a diversity of amino acid at those positions, the cosine similarity of the predictions was high (0. 63–0. 78), reflecting good overall profile recapitulation (Fig 4, S11 Fig). These results indicate that robust and accurate predictions of the specificity profiles of a variety of peptide-recognition domains can be obtained using the MFPred approach, pointing to its wide applicability, especially for cases where receptor backbone flexibility is minimal. Improved modeling of backbone conformational diversity, an area where methodological improvements are needed [45], is likely to improve prediction accuracy further. When used to design receptors for and against specificity profiles, MFPred should be able to accurately recapitulate changes in specificity profiles due to protease mutations, when simulations are performed on a constant set of backbones. As a proof of concept, we predicted the changes in the specificity profiles of two variants of granzyme B protease for which altered multispecificity has been experimentally determined (Fig 5). R192E granzyme B protease and R192E/N218A granzyme B protease have been shown to have decreased specificity for glutamic acid and increased specificity for lysine and arginine at P3 [46,47]. To investigate whether MFPred can recapitulate mutant specificity profiles without changing the peptide backbone, we modeled the variants of granzyme B protease by performing the necessary mutations in Rosetta on the five FastRelaxed granzyme B protease backbones. The MFPred-predicted specificity profile for the mutated structures accurately recapitulated the experimentally predicted specificity profile for the mutants. In the case of R192E, the change from a positively-charged arginine to a negatively-charged glutamic acid yields an increased frequency of positive amino acids such as lysine and arginine and a decreased frequency of negative amino acid glutamic acid. MFPred predicts the shift toward lysine and arginine and away from glutamic acid correctly, although it upweights the frequency of arginine and downweights the frequency of glutamic acid relative to the experimental profile. In the case of R192E/N218A, the shift towards arginine and lysine is even more pronounced in the experimentally-derived profile. Sterically, the mutation of N to A may allow for the longer sidechains of R and K (relative to E) to fit at P3. MFPred correctly predicts this shift as well. The sensitivity of MFPred to altered multispecificity at a given position due to a given receptor mutation should enable its use in designing for or against a given specificity profile.
Protein-peptide interactions underlie much of biology, and the ability to computationally manipulate these interactions would enable intervention in many biological processes. The rational design of receptor proteins, including enzymes that act upon peptide substrates, for and against peptide recognition specificity profiles is an open challenge. Such design would benefit from a specificity profile prediction technique that is both (i) rapid enough to be used in each step of the design process, and (ii) able to predict changed specificity for receptor variants with a constant peptide backbone conformational ensemble. The MFPred method developed here represents a step forward in achieving in both of these goals. MFPred is able to predict profiles for both proteases and a diverse set of PRDs, and it can recapitulate changes in the profile of variant granzyme B. This result sets the stage for application of the MFPred algorithm to enable the design of proteins for and against specificity profiles, by combining the MFPred algorithm with multi-state design [48]. The MFPred method, implemented in the context of the Rosetta software, performs specificity profile prediction with equivalent or better accuracy when compared to two previously developed methods (pepspec, sequence_tolerance) in the Rosetta framework, but with a significant decrease in run time (~10- to 1000-fold). Practically, this means that given a receptor variant and a peptide backbone ensemble, a specificity profile can be obtained, on a standard single processor, on a time-scale of seconds vs. hours required for other approaches. While pepspec and sequence_tolerance are less accurate on a smaller peptide backbone ensemble, MFPred is relatively robust to the size of the backbone ensemble. Additionally, MFPred can predict information content (determined from the amino acid frequency distribution at a given peptide position) better than other methods (Fig 3E–3H). The ability to recapitulate information content should enable design for a narrow or wide range of amino acid types at a given peptide position, thereby allowing greater control over binding selectivity. The speed, prediction accuracy on a small backbone ensemble, and robust recapitulation of information content of MFPred are due to the mean-field approach of MFPred: rather than attempt to enumerate many sequences on varying backbones, MFPred predicts a specificity profile by treating amino acid energies as a Boltzmann probability distribution. However, optimal sampling of the peptide backbone conformational space by MFPred does require some prior knowledge in the form of several (~5) recognized substrates, which is not required for pepspec or sequence_tolerance. While MFPred can rapidly and consistently generate recognition profiles with high accuracy compared to experimental data, it was not possible to achieve a perfect prediction using MFPred. Several reasons may underlie these limitations of MFPred. First, our experimental dataset may be incomplete: it comprises various in vitro and in vivo sources in the literature, each of which may have their biases. In vitro experimental profiles vary with the definition of a cleaved sequence; when few sequences are included in this definition, the profile will converge on a few optimal sequences. In vivo experimental profiles are subject to biases due to biological factors [21]. Second, any specificity prediction challenge is composed of several, smaller problems–sampling the vast sequence space, sampling the significantly larger conformational space, and scoring the structures–each of contributes multiplicatively to the error-rate. In our study, the sequence sampling problem is solved by MFPred itself. As it is an approximation, MFPred may not sample the sequence space effectively; the free parameters, which are optimized for overall success, are sub-optimal for each system. This is especially true in the case of the temperature parameter, which we found to be the most system-dependent. Thus, application of MFPred to domain families that are not included in our benchmark set may require further system-specific optimization of model parameters to achieve comparable accuracy. In terms of structure sampling, our method of utilizing a small number of known recognized peptides to generate a backbone ensemble is an attempt to more efficiently sample the large backbone conformational space (which also determines sidechain sampling due to the use of a backbone-dependent rotamer library [49]); however, this space is so large, especially in the case of a flexible binding pocket such as the HIV protease-1, that sampling efficiency is still limited. The sampling of receptor backbone flexibility is also required in such cases, as evidenced by a decreased prediction accuracy when the apo-structure of the complex is used (S12 Fig). Finally, we score the structures using an empirical energy function (from Rosetta); subtle errors in the energy function may also contribute to the observed inaccuracies. As both conformational and sequence sampling in the MFPred approach rely on, and are limited by, the underlying rotamer library and energy function as implemented in Rosetta, improvements in these features [49,50] should yield higher accuracy predictions.
We generated a flexible backbone ensemble by constructing models of the proteins bound to several cleaved sequences, and then diversifying those models via FastRelax [34], FlexPepDock [35], or Backrub [36] backbone sampling protocols, as described in detail below. For each protein, N cleaved sequences were chosen from the dataset by sorting the sequences in alphabetical order and then choosing evenly spaced sequences from the sorted dataset. Two alternative methods of picking cleaved sequences—randomly, or at even intervals from a set sorted by hamming distance from an arbitrarily chosen cleaved sequence—did not impact the results. Then those N cleaved sequences were threaded onto the original FastRelaxed protein-peptide complex to create N structure-sequence models. Each model was subjected to 10 trajectories of FastRelax simulations, 10 trajectories of FlexPepdock refine simulations, or 10 trajectories of Backrub simulations, and the resulting 10 models were considered to be the backbone conformational ensemble. As we found that the FastRelax protocol was more accurate than FlexPepDock and Backrub, we used FastRelax alone in the final version of the protocol. The model was constrained to active catalytic geometry for the proteases; we did not apply constraints to the PRD systems. Finally, the x lowest-scoring models for each sequence (with x dependent on the protocol in question, and generally set as 1) were chosen as the final backbone ensemble. Various self-consistent mean-field theory-based methods have been developed for use in protein sidechain packing and design [74–81]. In the canonical self-consistent mean field theory-based method for protein sidechain packing as proposed by Koehl and Delarue [74], the energy landscape is investigated by using an effective energy potential to approximate the effects of all possible rotamers at all positions to be modeled. Thus, the mean-field energy of rotamer r occurring at position i is determined by Eq 1: E (i, r) =e (ir) +∑j=1, j≠iN∑s=1Kje (ir, js) P (j, s) (1) e (ir) represents the one-body energy of the rotamer, or the energy between a residue and the fixed components of the protein. e (ir, js) represents the two-body energy between a rotamer r at position i and a rotamer s at position j. Energies are truncated at a threshold that we optimized as a free parameter. P (j, s) represents the probability of rotamer s occurring at position j and is initially given as 1/Kj, where Kj is the total number of available rotamers at position j (obtained from a rotamer library). A probability matrix (P) of size N × Kmax, where N is the number of positions to be analyzed and Kmax is the maximum number of rotamers at any position, is used to model the probabilities of each rotamer occurring. Once the effective energy of each rotamer is determined using (1), the probability of each rotamer is: P (j, s) =e−βE (j, s) ∑x=1Kje−βE (j, x) (2) β (= 1/kT) is also optimized as a free parameter. The algorithm iterates between the two equations until convergence is reached. We use a pre-calculated interaction graph in Rosetta [82] to store the one-body and two-body energies, which do not change between iterations, so the iteration is rapid. Convergence is improved with the use of a memory in the updating of P, so that the probability matrix after iteration x is given by Px = λPx−1 + (1−λ) Px, where λ is a free parameter between 0 and 1. Once convergence is reached, the probability matrix P can be used to obtain the probability for every rotamer. We extended the algorithm for use with a flexible backbone and with any given amino acid alphabet. Given an ensemble of backbone conformations, the probability matrix P is calculated for each backbone using the canonical self-consistent mean field method, while allowing each position to take on any amino acid, so that the vector for that position contains all the rotamers for all amino acids at that position. Paa (bb, i), the probability of amino acid aa occurring at position i in backbone bb, is determined for all amino acids at all positions in all backbones: Paa (bb, i) =∑r=1KaaPbb (i, r) Kaaγ∑x=120∑r=1KxPbb (i, r) Kxγ (3) where Kaa is the number of rotamers available to amino acid aa at position i, and γ is a free parameter optimized to 0. 8 in our implementation. Dividing the sum of probabilities over all rotamers for amino acid aa by Kaaγ thus corrects for cases where numerous rotamers of an amino acid artificially inflate the probability of a specific amino acid occurring (S1 Fig). The probability matrices for all backbones are then averaged together using a Boltzmann-weighting scheme in a two-step process. First, Ebb (i, aa), the weighted sum of the energies for rotamers of amino acid aa at position i in backbone bb, divided by Kaaγ, is calculated (Eq 4). Then Ebb (i, aa) is used to find W (i), the probability of backbone bb occurring at position i (Eq 5). M is the number of (peptide) backbones in the ensemble. Finally, a weighted average P is determined and taken to be the predicted specificity profile for that protease: P (i, aa) =∑bb=1MPaa (bb, i) W (i) (6) Thus, MFPred can be used for prediction of multispecificity for both one backbone and multiple backbone conformations. To optimize four free parameters for MFPred (λ, γ, threshold, and kT), we enumerated all combinations of λ (0. 25,0. 5,0. 75), γ (0,0. 2,0. 4,0. 6,0. 8,1. 0), threshold (5,10,50,100,250,500), and kT (0. 2,0. 4,0. 6,0. 8,1. 0). We selected 68 structures from the peptiDB (a peptide-protein complex database) [83] that met our criteria of having at least eight peptide residues. The structures were input into MFPred as a backbone ensemble and all combinations of the above parameters were tested. The resulting background specificity profiles were compared to the background residue distribution in the Rosetta database (S1 Fig, S9 Fig) and the combination of parameters with the lowest cosine distance from the known background distribution was chosen as our final set of parameters. While varying λ had little impact on the results, all other parameters had a significant, system-dependent impact on the results. Since the MFPred predictions include noise arising from limited sampling and the scoring function used (as mentioned above), we divided its predictions by the background profile to find the final prediction. The background profile was determined by averaging the frequencies of each position in the peptiDB profile. We divided each amino acid frequency in the initial predicted profile by the frequency of that amino acid in the background profile to find the final profile (S9 Fig). MFPred is available as a RosettaScripts Mover within the master branch of Rosetta. Sample cases for how to use MFPred can be found in S2 Note and in online Rosetta documentation. | Across biology, many proteins that recognize peptides are multispecific; they interact with multiple binding partners of disparate sequence. Computational prediction of these multiple peptide partners would enable greater understanding of individual protein-recognition domains. Additionally, the ability to customize protein-recognition domains by designing them to recognize and act upon a new set of peptides and not bind their original binding partners would be useful in drug design and biotechnology. Current methods for predicting multispecificity operate on a timescale that is too slow to be used for design. Here, we present a method, MFPred, for predicting multispecificity. MFPred robustly recapitulates protein-recognition domain specificity for a range of proteins, at comparable accuracy and with considerable speed-up relative to current methods. We apply MFPred to predicting altered multispecificity in a mutant protease to demonstrate its relevance to design. The rapidity and accuracy of MFPred should enable its use in investigating and modulating biological processes. | Abstract
Introduction
Results
Discussion
Methods | medicine and health sciences
crystal structure
pathology and laboratory medicine
protein interactions
chemical compounds
enzymes
pathogens
condensed matter physics
enzymology
microbiology
organic compounds
retroviruses
viruses
immunodeficiency viruses
rna viruses
basic amino acids
amino acids
acidic amino acids
crystallography
solid state physics
proteins
medical microbiology
hiv
microbial pathogens
chemistry
physics
biochemistry
peptides
arginine
organic chemistry
glutamic acid
viral pathogens
protein domains
biology and life sciences
proteases
physical sciences
lentivirus
organisms | 2017 | MFPred: Rapid and accurate prediction of protein-peptide recognition multispecificity using self-consistent mean field theory | 9,777 | 232 |
The hotspots of structural polymorphisms and structural mutability in the human genome remain to be explained mechanistically. We examine associations of structural mutability with germline DNA methylation and with non-allelic homologous recombination (NAHR) mediated by low-copy repeats (LCRs). Combined evidence from four human sperm methylome maps, human genome evolution, structural polymorphisms in the human population, and previous genomic and disease studies consistently points to a strong association of germline hypomethylation and genomic instability. Specifically, methylation deserts, the ∼1% fraction of the human genome with the lowest methylation in the germline, show a tenfold enrichment for structural rearrangements that occurred in the human genome since the branching of chimpanzee and are highly enriched for fast-evolving loci that regulate tissue-specific gene expression. Analysis of copy number variants (CNVs) from 400 human samples identified using a custom-designed array comparative genomic hybridization (aCGH) chip, combined with publicly available structural variation data, indicates that association of structural mutability with germline hypomethylation is comparable in magnitude to the association of structural mutability with LCR–mediated NAHR. Moreover, rare CNVs occurring in the genomes of individuals diagnosed with schizophrenia, bipolar disorder, and developmental delay and de novo CNVs occurring in those diagnosed with autism are significantly more concentrated within hypomethylated regions. These findings suggest a new connection between the epigenome, selective mutability, evolution, and human disease.
Array comparative genomic hybridization (aCGH) studies [1] and massively parallel sequencing [2] revealed that approximately 10% of the human genome is structurally polymorphic at the submicroscopic scale (<4 Mb), a much larger fraction than affected by single nucleotide polymorphisms (SNPs). Structural mutations that occur in a number of well studied structurally unstable loci cause disease [3]. The discovery of these structurally mutable disease-associated loci gave rise to the concept of genomic disorders [3], [4]. Their detailed analysis revealed the role of non-allelic homologous recombination (NAHR) and low copy repeats (LCR) in mediating recurrent deletions, duplications and inversions [5]. Genome-wide analyses of regions between paralogous LCRs in direct orientation have since led to the successful prediction of novel LCR-mediated genomic disorders [6], reinforcing the role of NAHR and LCRs. A potential role for LCR in inverted orientation has been elucidated recently for a specific type of complex duplication with an embedded triplicated segment in inverse orientation, DUP-TRP/INV-DUP [7]. The process of chromothripsis [8] has been proposed as a model to explain instability in 1–3% of all cancers resulting in a highly complex pattern of genomic rearrangements with multiple CNVs. The patterns of genomic instability observed in cancer have also been observed in complex genomic rearrangements (CGR) in human germline, pointing to similar mechanistic underpinnings [9]. The distribution of structural mutations in the human genome is highly selective, characterized by many hotspots of structural mutability. Evolutionary analyses of recent structural mutations in the human genome reveal that structural mutation hotspots frequently give rise to new LCRs [10], [11], indicating that a significant fraction of the observed association of LCRs and mutability may be explained by the increased production of LCRs at hypermutable loci. The recent discovery of a genome-wide association of LCRs with somatic mutability in cancer [12], and structural breakpoints in the mouse genome independent of LCR homology [13] further support the hypothesis that LCRs may not always cause instability but may preferentially arise at the loci that are inherently mutable both in cancer and in germline. Recent high-resolution genome analyses of genomic disorder loci revealed complex patterns of rearrangements not consistent with the NAHR mechanism [14], [15], [16], [17]. The mechanisms causing mutability in such structurally mutable hotspots remain elusive. Microhomologies and other sequence-level features point to the role of Fork Stalling and Template Switching (FoSTeS) and Microhomology-Mediated Break-Induced Replication (MMBIR) mechanisms [16] in the processing and repair of one-ended, double-stranded DNA [18]. However, these are repair mechanisms, are not causing mutations, and have not explained the highly selective distribution of structural mutability nor predicted genomically unstable loci. Multiple independent lines of evidence point to a possible role of the epigenome in structural mutability. Chromatin modifications are known to play a significant role in chromosome maintenance [19], including DNA repair [20], [21], and recombination [22], [23]. Chromatin and the epigenome regulate mutability at smaller scales, including increased mutability of 5-methyl cytosine [24], retroposon silencing [25], [26], [27], and preferential retrotransposition into specific chromatin states [28]. Genome-wide hypomethylation has been repeatedly observed in structurally unstable cancer genomes [29], [30]. Mutations in the methyltransferase DNMT3B have been shown to cause hypomethylation and genomic instability in juxtacentromeric regions in humans [31]. Mutations in the mouse homolog of methyltransferase DNMT1 have been shown to cause genomic instability [32]. Analyses of the structurally hypermutable genomes of gibbon species revealed association of hypomethylation with structurally mutable loci [33]. Finally, the recent discovery of the role of the DNA-break inducing base-excision repair pathway in genomic demethylation of primordial germ cells (PGCs) during fetal development in mouse [34] provides a possible mechanistic link between genomic hypomethylation and genomic instability in the mammalian germline. Genomic hypomethylation and LCR-mediated NAHR are therefore the two genome architectural features shown to be associated with structural changes. We here systematically examine and quantitate these associations. To assess the degree of association of germline methylation levels with structural instability, we examine four sperm methylome maps, including two high read coverage (15× combined coverage) from a recent study [35] and two maps we obtained by performing whole-genome bisulfite sequencing of sperm samples from two anonymous donors at low coverage (2. 5× combined coverage). To improve detection of structural mutations associated with LCRs and NAHR, we perform a comprehensive detection of human LCRs in the human genome and design an aCGH array for diagnostic use in the BCM Medical Genetics Laboratories (BCM-MGL) targeting NAHR susceptible regions between directly oriented paralogous LCRs (DP-LCRs) with size larger than 10 Kbp, separated by a distance less than 10 Mb of unique genomic sequence. We combine evidence of structural mutations from the following three sources: 1) human-specific genomic rearrangements; 2) structural polymorphisms in the human population, including copy-number variation (CNV) data from BCM-MGL and publicly available CNV data sets [36], [37], [38]; and 3) recent disease studies of schizophrenia [39], bipolar disorder [40], developmental delay [41], and autism [42]. Our analyses reveal a pattern of association of structural mutability with germline hypomethylation comparable in magnitude to the association between structural mutability and LCR-mediated NAHR.
To examine a potential association between germline methylation and structural mutability in humans, we first derived two sperm methylome maps by sequencing at combined 2. 5× genome coverage (one at 1. 2× and the other at 1. 3×) bisulfite-treated genomic DNA samples extracted from the sperm of two anonymous donors. Methylation levels were calculated for each of the 28,705 non-overlapping 100 Kbp windows covering the hg18 human genome assembly as the ratio between the number of methylated CpGs and the total number of CpGs sampled in reads mapping within the window. Windows with less than 20 CpG sampling events were removed from the subsequent analysis to avoid bias due to low sequence mappability. Both samples had more than 95% of windows with reads covering more than 40% of the CpGs within the window (Figure S7B). Due to the low 2. 5× combined coverage, the methylation levels of individual CpGs could not be determined with accuracy, but the average methylation levels at 100 Kbp level of resolution could be determined with high accuracy. Specifically, the methylation level of >98% windows was determined with <10% error with >95% probability (Table S10). The two methylomes were highly concordant at 100 Kbp level of resolution (linear correlation coefficient = 0. 96). For the purpose of our analyses, an average sperm methylome at 2. 5× coverage was constructed as an average of the two concordant methylomes. Methylation deserts were operationally defined as the 100 Kbp windows with the lowest 1% methylation level in the average sperm methylome. A 5% threshold was also used for some analyses, as noted below. We repeated our analyses using an independently obtained pair of sperm methylomes generated by Molaro et al. [35] from bisulfite sequencing data at a combined 15× genome coverage. To ensure deep sampling of CpGs in each window, only windows with more than 100 mapped reads and more than 100 CpG sampling events at 15× coverage were included in the subsequent analyses. To facilitate comparison, both combined methylomes (at 2. 5× coverage and at 15× coverage) were represented as methylation averages across the same set of 100 Kbp windows tiling the human genome. The 15× methylome showed high correlation with the 2. 5× methylome at the 100 Kbp resolution (r = 0. 82, p-value<2. 2e-16). Methylation deserts discovered at 2. 5× coverage using methylation percentile rank thresholds of 1% and 5% significantly overlapped those discovered at 15× coverage (Figure S21), indicating relatively stable genomic localization of methylation deserts across individuals. It has been suggested that directly-oriented paralogous LCRs (DP-LCRs) with high similarity, large size, and in close proximity would be most likely to mediate NAHR, resulting in deletions or duplications identifiable by aCGH [1], [3], [5], [6], [43]. We designed, implemented, and validated a new computational method for comprehensively detecting LCRs and DP-LCRs (see Materials and Methods: Computational Pipeline for LCR Identification). The method achieves higher sensitivity than previously applied methods [44] by using k-mer frequency sequence information to detect and cluster LCRs without remmatoving (repeat-masking) high copy-number repetitive elements (Materials and Methods: Whole-Genome Self-Comparison and Text S1 section 1. 1). In total, 268 regions between DP-LCRs were identified (Figure S3), a greater than two-fold increase over previously reported estimates (Text S1 section 1. 2 and Figure S4). We next examined the association of evolutionarily recent structural rearrangements in the human genome with both DP-LCR loci and germline hypomethylation. Assuming nearly neutral evolution [45], the distribution of structural variants that have accumulated in the human lineage since the branching of chimpanzee can be used as an indicator of structural mutability. By applying the Genomic Triangulation method [46] to genomic data from four non-human primate species (chimpanzee, rhesus macaque, orangutan and marmoset) and the human reference genome we detected 522 human-specific structural rearrangements (Materials and Methods: Identification of Human-Specific Rearrangements). The human-specific structural rearrangements were found to be highly associated with LCRs (six-fold enrichment, permutation test, p≈10−3), much higher than with other examined genomic features such as repetitive elements (Alu: 0. 89-fold; LINEs: 1. 1-fold; Microsatellites: 1. 2-fold). One-third of the rearranged regions were actually human LCRs, indicating a significant fraction of the association may be explained by segmental duplication events that produce LCRs. The rearrangements were found to associate specifically with DP-LCR loci to a lesser degree (three-fold enrichment, permutation test, p≈10−3). A striking association was detected between human-specific structural variants and hypomethylation. First, the methylation deserts comprising a total of 1% of the human genome contain ∼10% of the human-specific structural rearrangements, a tenfold enrichment (Figure 1A). Second, genome-wide comparison indicates a highly significant inverse association of human-specific rearrangements with methylation levels (Kolmogorov-Smirnov test, Dmax = 0. 23, p≈10−24) (Figure 1B). Additional permutation-testing experiments that are not based on fixed window size indicate that approximately 23% (Dmax = 0. 23) of human-specific rearrangements associate with hypomethylation (Figure S5A). The significance of this association gradually decreases with increasing distance from rearrangements (Figure 1C), suggesting that hypomethylation and structural mutability co-localize within relatively small chromosomal segments. The association could not be accounted for by considering a number of other potentially confounding factors including CpG islands, chromosomal bands, telomeric/centromeric locations and sex chromosome bias (Text S1 section 3; Tables S8, S9). We next directly compared the relative strengths of association of hypomethylation and DP-LCRs with human-specific rearrangements. The 100 Kbp windows covering the genome were each assigned to one or more of the following groups: (a) windows containing human-specific rearrangements; (b) windows that are methylation deserts; and (c) windows containing regions between DP-LCRs. The Venn diagram in Figure 2A illustrates proportions of windows across the three groups, based on which we calculated the statistical relative and attributable risks of rearrangements due to hypomethylation and DP-LCRs in Figure 2B (first row). Note that both genomic features confer significantly increased statistical risk, but the statistical relative risk due to hypomethylation is markedly higher than the risk due to DP-LCRs. Methylation levels in sperm are only a partial indicator of methylation levels in the whole human germline. To further examine the association between germline methylation and structural mutability in humans directly, one would ideally be able to measure DNA methylation in the entire male and female germline lineages, which are highly dimorphic [47]. To practically address this issue, we pursued an indirect approach by estimating methylation levels in the human germline (an average of male and female germlines), using the methylation index (MI) model [48] (Materials and Methods: Methylation Index Calculation at 100 Kbp Level of Resolution). Approximately 20% of the methylation deserts (defined as the lowest 1% methylation levels in sperm) occur within the 1. 5% fraction of windows with the lowest MI score (MI = 0), an indication that methylation deserts detected in sperm overlap substantially with hypomethylation in the germline as a whole (Figure S6A). The windows with MI = 0 contain ∼15% of the human-specific structural rearrangements, a similar tenfold enrichment as we observed for methylation deserts defined based on the sperm methylomes (Figure 1A). The sperm methylation scores of windows with MI = 0 show a bimodal distribution (Figure S6B), the lower mode including 35% with low methylation levels (<5%) in sperm and the higher mode is comprised of the remaining 65% that appear to have normal methylation levels in sperm. Because the higher mode could not be explained by obvious ascertainment biases (Materials and Methods: Examination of MI Ascertainment Biases), we hypothesize that this mode may either indicate hypomethylation specific to the female germline, given that male and female germline methylation patterns are highly dimorphic [47], or may be due to other germline hypomethylation detected by MI that is absent from sperm. Similar bimodal distribution was observed at 15× coverage (Figure S9B). As additional controls, five publicly available methylomes obtained by whole-genome bisulfite sequencing [49], [50] of human stem cells and fibroblasts were also compared over the same set of 100 Kbp windows. Methylation levels in sperm showed much higher correlations with the methylation levels in embryonic stem cells than with fibroblasts (Table S2), consistent with the more differentiated state of fibroblasts. Importantly, the methylation levels in sperm samples have higher correlations with the germline MI scores than either stem cells or fibroblasts (Table S2). Moreover, the bimodal distribution of hypomethylated regions is unique to sperm (Figure S9), consistent with sperm being the closest representative of the human germline. To examine structural mutability during more recent evolutionary time, we turned to the analysis of Copy Number Variants (CNVs) in the human population. De-identified aCGH data were collected from 400 human DNA samples analyzed by the BCM Medical Genetics Laboratories (BCM-MGL; http: //www. bcm. edu/geneticlabs/). These data were originally produced at BCM-MGL using a custom designed, whole-genome oligo-aCGH chip with a genomic distribution of probes more densely spaced between DP-LCRs as well as with lower but even distribution for the remaining regions of the genome (Materials and Methods: aCGH Probe Set Design and Analysis of CNVs in 400 MGL Samples). Approximately 12,000 non-unique CNVs seen in more than one individual larger than 500 bp were identified. More than 60% of the CNVs were not in public structural variation databases (Figure S10). A significant enrichment of LCRs (permutation test, three-fold enrichment, p≈0. 01) was found around the CNVs. When CNVs occurred between DP-LCRs, they were more likely to span the intervening region, a signature of NAHR, than those between non-paralogous LCRs (2-fold enrichment, p≈0. 001 by chi-square test, Figure 3G). However, such CNVs represent a small fraction (∼2. 5%, Figure 3A) of all CNVs. We next examined any potential association between LCRs and structural mutability using structural heterozygosity as a proxy. Assuming structural mutations are neutral, under the infinite allele model [51], the rate of structural heterozygosity is proportional to the mutation rate. Structural mutability can therefore be assessed using the rate of structural heterozygosity as a proxy (Figure 4A). Our results indicate that genome-wide structural mutability is directly correlated with LCR density and particularly with the LCRs that contain high copy-number repetitive elements (Figure S11). Following a similar approach as in Materials and Methods: Human-Specific Evolutionary Structural Rearrangements Associate More Strongly with Methylation Deserts Than with DP–LCR Regions, we next compared the distributions of sperm methylation levels for 100 Kbp windows containing CNVs and for those not containing any CNVs. The Kolmogorov-Smirnov test results indicate that the windows containing CNVs have significantly lower methylation in sperm (Figure S13). Permutation testing indicates that an excess of 9% of the CNVs is explainable by hypomethylation (Kolmogorov-Smirnov Dmax = 0. 09, Figure S5B). Association analysis also indicates a higher statistical relative risk due to hypomethylation than due to DP-LCRs (Figure 2B). We next compared structural mutability in methylation deserts with mutability in other genomic loci using structural heterozygosity rate as a proxy. The comparison indicated that the methylation desert loci have higher average structural heterozygosity rates (Figure 4B). The Kolmogorov-Smirnov test also indicates significant excess heterozygosity of CNVs in hypomethylated regions (Figure S14A). As an independent test for any potential association between hypomethylation and structural mutability, we performed analyses analogous to those discussed in the previous section using the following three publicly available CNV datasets: (i) aCGH data obtained from 270 HapMap samples using high-resolution Affymetrix SNP 6. 0 arrays [36]; (ii) aCGH data obtained from 450 HapMap samples using tiling oligonucleotide microarrays [37]; and (iii) CNV data generated on 19,000 samples [38] in a study of the role of common CNVs in eight common human diseases. The dataset (i) complements the 400-sample BCM-MGL data because it detects CNVs that overlap LCRs, and it provides high probe resolution in regions that are not associated with LCRs. Despite the bias away from known polymorphisms in the design of the custom array used to generate the 400-sample BCM-MGL dataset (Materials and Methods: aCGH Probe Set Design and Analysis of CNVs in 400 MGL Samples, Text S1 section 5 and Figure S12), analyses of the data set (i) confirmed the relative strengths of association of structural mutability with NAHR and with hypomethylation identified using the BCM-MGL data, as indicated in Figures 2B, 3,4, Figure S14, and Table S7. All three (i–iii) datasets confirmed significantly higher average heterozygosity rates of CNVs in methylation deserts (Figure 4). However, dataset (iii), which was biased against rare structural alleles [38], showed no significant difference in overall heterozygosity rate distributions between CNVs in the methylation deserts and the rest of the CNVs (Figure S14D), suggesting that rare variants may account for a significant fraction of association. In summary, despite the differences in array technologies, array design biases, and sample sets applied to the arrays, our analyses repeatedly point to a significant association of hypomethylation and structural mutability. We next asked if the association between structural mutability and hypomethylation is specific to germline, using the embryonic stem cell line H1 methylome [50] as a control. Germline methylation was assessed using the sperm methylomes both independently and in combination with the methylation index, as summarized in the five columns in Table 1. Recall that for windows with MI = 0, the sperm methylation scores showed a bimodal distribution (Figure S6B). As indicated in Table 1, significant enrichment of structural mutability could be observed for windows with MI = 0, and for both lower and higher modes of these windows. The enrichment observed in the higher mode (Table 1, column “MI = 0 & sperm>5%”) suggests the role of hypomethylation that is possibly present in the female germline and captured using the MI measurement but not present in sperm. The windows containing rearrangement/variation showed much lower methylation levels in the sperm methylome (Figure S15A–S15C). In contrast, an association with methylation levels in H1 could not be detected for the CNVs, except that windows containing human-specific evolutionary rearrangements did show association (Figure S15D–S15F). We found significant negative correlation between the methylation scores in sperm and the heterozygosity rates (CNVs from 400 MGL samples: r≈−0. 15, p≈10−9; CNVs from 270 HapMap samples: r≈−0. 20, p≈10−10). In contrast, no significant correlation between the H1 methylation scores and the CNV heterozygosity rates was detected. We next examined the difference in methylation levels between sperm and H1. As illustrated in Figure S16, the difference shows even stronger association with structural mutability than the absolute methylation levels in sperm. This result rules out possible ascertainment biases due to low mappability of sequencing reads in potentially unstable and repetitive hypomethylated regions. It also suggests that structural mutability is associated with germline-specific hypomethylation. We next examined the distribution of rare CNVs detected in the recent large-scale study by the International Schizophrenia Consortium [39]. CNVs in 3,391 individuals diagnosed with schizophrenia and 3,181 controls were identified and analyzed using Affymetrix SNP arrays. The study found that the individuals in the affected group have 15% more rare variants. We asked if the excess of variants in the affected group tends to occur in regions with low germline methylation levels. We first compared the distribution of the methylation levels for 100 Kbp windows containing the CNVs in the affected group with the distribution of methylation levels for windows not containing any CNVs. The same procedure was performed for the CNVs in the control group. Both the affected and control CNVs showed lower methylation. A significant enrichment of low MI values (Kolmogorov-Smirnov test, p≈10−5) was found for the affected group (Table S3), while no significant enrichment was found for the control group. We next identified those CNVs found only in the affected group and those found only in the control group. The two subsets were then further classified as being within or outside of regions showing lowest 5% methylation levels in sperm. The chi-square test indicates a 3-fold enrichment (p≈10−3) within low methylation regions of variants identified only in the affected group compared to those found only in the control group (Table 1). Similar enrichment was found in regions with MI = 0 (Table 1). We next examined distribution of CNVs identified in a recent bipolar disease study [40]. The study identified CNVs in 1001 bipolar disease cases and 1034 controls. An excess of large singleton deletions was found in cases relative to controls. We examined methylation of singleton deletions found only in bipolar cases to the methylation of the deletions found only in controls. As indicated in Table 1, compared to control-specific deletions the case-specific singleton deletions were enriched over 2-fold (p<1e-3 by Chi-square test) within the 100 Kbp windows having lowest 5% methylation levels in sperm. A recent autism spectrum disorders (ASDs) study [42] found a higher burden of rare CNVs in ASD patients. Trio analyses established that some of the CNVs were not present in parental genomes and were classified as de novo. We asked if the rare and de novo CNVs detected in the autism cases and controls associated with low methylation levels. The regions containing rare CNVs in both the cases and controls showed significant enrichment for both low methylation levels in sperm and for low MI values, when compared with regions without any rare CNVs (Table S3). The CNV variants identified only in the cases showed an approximately two-fold enrichment in hypomethylated regions compared to those found only in controls, but the enrichment did not reach statistical significance threshold due to a small number of variants detected (data not shown). Analysis of de novo and inherited CNVs found in cases revealed highly significant enrichment within hypomethylated regions of de novo relative to inherited CNVs. The enrichment was observed within hypomethylated regions in sperm (<5%), within windows of MI = 0, and especially in regions that met both criteria (Table 1). A recent study by Cooper et al. [41] identified CNVs in 15,767 children with intellectual disability and various congenital defects (cases) and in 8,329 unaffected adults (controls). We examined the enrichment of rare (<1% population frequency) case-associated CNVs within the windows with lowest 5% methylation in sperm relative to CNVs found in controls. Using Chi-square test, we observed a significant 2. 9-fold enrichment of the case-specific rare CNVs (p = 2. 78e-124) compared to the control CNVs. Out of the 59 pathogenic CNVs identified in this study, 12% are located in the methylation deserts, a 4. 7-fold (p = 3. 3e-5) enrichment compared with the control CNVs. Specific sub-classifications of phenotypic information was reported for almost half of the cases, including 575 cases with cardiovascular defects, 1,776 with the epilepsy/seizure disorder, 1,379 with the autism spectrum disorder and 3,898 with craniofacial defects [41]. We therefore repeated the same chi-square test for each sub-class, and observed enrichment of CNVs associated with each sub-phenotype vs. all control CNVs (11). Analysis of genomic features in the methylation deserts showed no enrichment for SINEs, LINEs or microsatellites (Figure S1C). Higher GC content was found for methylation deserts than elsewhere (Figure S1A), which may be due in part to the somewhat higher number of CpG islands in these regions than expected by chance (Figure S1C). Methylation deserts also showed higher average sequence conservation than the rest of the genome (Figure S1B). However, conserved coding sequences were slightly under-represented (0. 9 fold), and pseudogenes were over-represented (2 fold, Figure S1C). Overall, genes were under-represented (0. 7 fold) except for homeobox, cadherin, and histone families, all of which were highly enriched in methylation deserts (Table S1). Using the sperm gene expression data from previous studies by Pacheco et al. [52], we detected enrichment within methylation deserts of those genes that are highly expressed in sperm (Text S1 section 2). We next examined enrichment of promoters categorized by their CpG content into high-, intermediate- and low-CpG content promoters by Weber et al. [53]. We first observed a significant negative correlation between the methylation level and average CpG content across all 100 Kbp windows (r = −0. 35, p = 2. 5e-270). However, methylation deserts were not enriched for promoters with high CpG content (Table S6). Those with low CpG content showed slight under-representation in the methylation deserts (0. 65 fold). Interestingly, those with intermediate CpG content, which were also referred to as “weak CpG islands” and known to be more prone to de novo methylation during differentiation [53], [54] showed 3-fold enrichment in the methylation deserts (Table S6). According to Mohn et al. , almost all bivalent promoters (marked by both H3K27me3 and H3K4me2 during cellular differentiation) contain CpG islands, and a significant proportion of weak CpG promoters are bivalent and more likely to be methylated de novo [54]. We therefore examined the bivalent promoters as identified by Ku et al. [55] and found their 2. 6-fold enrichment in the methylation deserts (Table S6). The promoters that were both bivalent and had intermediate CpG content showed four-fold enrichment (Table 2). Because the Polycomb repressive complex 2 (PRC2) is known to regulate bivalent promoters, we next examined the distribution of PRC2 binding regions within methylation deserts, focusing specifically on the hyperconserved CpG domains (HCGDs) identified by Tanay et al. [56]. Tanay et al. used the COCAD (context-based CpG analysis of divergence) score to compare the actual rate of human–chimpanzee CpG divergence to the predicted rate. The HCGDs with low COCAD scores showed extensive overlap with regions bound by Polycomb repressive complex 2 (PRC2). Of the 194 non-overlapping genomic regions corresponding to HCGDs with COCAD scores below −5 (P<1E−6), a total of 60 (31%) are located in the methylation deserts (2. 5× coverage), showing a 37. 6-fold enrichment compared to the genomic background as determined by permutation testing (Table 2). Because tissue-specific regulation may involve changes in CpG methylation levels, we next investigated whether the methylation deserts are enriched for regions that are methylated in a tissue-specific manner. Toward this goal, we first examined the methylation data gathered at 1,413 CpG loci across 217 samples from 11 different human tissue types by Christensen et al. [57]. The CpG loci were divided into a group within germline methylation deserts and a group that did not fall within methylation deserts. Each CpG locus was assigned a score measuring the variation of methylation level across 11 tissues [57]. Kolmogorov-Smirnov test showed that CpG loci within the methylation deserts are significantly enriched for inter-tissue variability (Figure S22). To rule out the possibility that the excess variation is due to causes other than developmental regulation, the distributions of CpGs that exhibit aging-related variation and of those that exhibit environment-related variation were examined. None of the two groups of CpGs exhibited any preferential distribution within methylation deserts, indicating the methylation difference among cell lineages is more likely to be related to developmental regulation. We next examined whether the methylation deserts are enriched for regions involved in regulation of tissue-specific gene expression using the set of 269 putative genomic regulatory blocks (GRBs) and their target genes identified in the human genome by Akalin et al. [58]. The GRB target genes are most often transcription factors involved in embryonic development and differentiation. We examined the enrichment of GRB target genes or GRBs themselves in the methylation deserts (lowest 1% sperm methylation at 2. 5× coverage) using randomly selected genomic segments as controls. The GRB target genes showed 12-fold enrichment in the methylation deserts (p<1e-10). The GRBs on the other hand, showed around 2. 8 fold enrichment in methylation deserts, of which those that are multiple target GRBs showed a 4. 4 fold enrichment (both p<1e-3). Comparing distribution of other CpG island-overlapping genes outside GRBs to GRB target genes, by chi-square test we observed an extremely high 33-fold enrichment of GRB target genes within the methylation deserts (p∼1. 41e-146, Table 2). As an additional control, we examined ‘bystander’ genes defined by Akalin et al. as those intertwined with highly conserved non-coding elements but whose expression and function are unrelated to those of the GRB target genes. GRB target genes were enriched in the methylation deserts 9. 2-fold relative to the ‘bystanders’ (p∼1. 42e-43, by chi-square test, Table 2). Because methylation deserts are hotspots of evolution, we examined enrichment within methylation deserts of transcription factors (TFs) reported by Vaquerizas et al. [59] to be fast evolving in primates. We first applied permutation test to the coding sequences of all the ∼1300 manually curated sequence-specific TFs and observed a 3. 75 fold enrichment for their coding sequences in the methylation deserts (p<1e-3). We then examined the TFs within two clusters reported by Vaquerizas et al. [59] to be fast evolving in primates and detected an even higher 15-fold (p<1e-3) enrichment (Table 2).
Combined evidence from evolutionary, population-genetic and disease studies supports strong association between germline hypomethylation and selective structural mutability. Genome-wide, both relative and attributable risks of structural mutations due to methylation deserts are at least comparable to the corresponding statistical risks due to LCR-mediated NAHR. Our results show that 23% of human-specific evolutionary rearrangements are associated with hypomethylation. Methylation deserts comprise a total of 1% of the genomic sequence and contain about 10% of the 522 submicroscopic human-specific structural rearrangements identified by primate genome comparisons. The evolutionary findings are generally consistent with the results of analyses of CNVs in the human population. Our analysis reveals a two-fold genome-wide enrichment for deletions and duplications between DP-LCRs, the signature pattern of LCR-mediated NAHR. While the enrichment is statistically significant, the fraction of structural variation statistically attributable to NAHR is small, approximately 2. 5%. We show that methylation deserts exhibit higher association with CNVs (∼9%) and contain a disproportionately high fraction of CNVs that have high structural heterozygosity. The population-based analyses reveal less striking enrichment patterns than the evolutionary analyses. This may be explained by the fact that population based studies were generally of lower resolution (array-based, unlike sequence-based evolutionary analyses), were limited to copy-number changes, and were biased against rare variants. By demonstrating a higher association of structural mutability with hypomethylation than with NAHR, our results underscore the potential relative contribution of the role of microhomology-mediated break-induced repair in structural genomic instability [37] which is consistent with replication based mechanisms such as FoSTeS [14], MMBIR [18], and serial replication slippage (SRS) [16] rather than NAHR. Our results are consistent with the concept of a structural selective “mutability profile”, an epigenomic phenotype marked by the variation in germline methylation levels along the genome. Three questions regarding this mutability are of particular interest: heritability, mechanism, and evolution. First, does inter-individual variation in methylation-associated selective mutability profiles exist and if it does, is it heritable? As a first step toward answering these questions, we have generated preliminary results tentatively suggesting that inter-individual variation in selective structural mutability may be associated with methylation deserts (Text S1 section 6 and Figure S17). The second open question is the mechanism behind the selective mutability profile. One conceivable mechanism is genetic variation in DNA-break inducing base-excision repair enzymes involved in germline-specific demethylation [34]. Another possibility may involve unrepaired DNA breaks associated with active transcription because methylation deserts are highly transcribed in germline. Yet another possibility may be that transcription factors mediate structural rearrangements by bending chromatin, creating looping structures and DNA breaks, analogously to the role played by estrogen and androgen receptors in mediating structural instability in hormonally regulated tumors [60], [61], [62]. One specific possibility opened by this model is that selective structural mutability may be affected by the cellular and organismal environment and may be controlled experimentally or even therapeutically. Finally, assuming selective mutability profile variation is heritable, the question of its evolution arises (for a recent survey of the topic of “evolution of evolvability” see [63]). Specifically, does selective mutability evolve mostly neutrally by random drift? If not, what may be the nature of selection pressure acting on it? Assuming that selection indeed plays a role, it is useful to consider the payoff (higher probability of developing a favorable mutation that ultimately becomes fixed in the population) and risk (of mutation causing disease). A selective mutability profile with excess mutability concentrated in the loci with low payoff/risk ratios would then be less likely to produce mutations that ultimately become fixed than a mutability profile with mutability concentrated in the loci with high payoff/risk ratios. The latter would therefore be favored by selection. One testable corollary of this payoff/risk model is that de novo mutations will tend to cause diseases related to the phenotypes that are under positive selection in the human population. Assuming that brain function is under selection in the human population, this corollary predicts high incidence of brain-related diseases such as schizophrenia, bipolar disorder, autism, epilepsy, developmental delay and cranial features due to rare and de novo mutations. Our findings that the rare and de novo CNV variants in the individuals suffering from these diseases indeed concentrate within methylation deserts is consistent with this corollary. These findings suggest a novel type of connection between evolution and human disease [64]. The payoff/risk model is also consistent with highly mutable loci being responsible for tissue-specific phenotypes. This is because a mutation in a locus regulating a tissue-specific phenotype may not confer much risk to other tissues. The enrichment within methylation deserts that we observed for genes with tissue-specific patterns of expression and for transcription factors involved in cellular differentiation is therefore consistent with this payoff/risk model.
Two anonymous human sperm samples were collected from a local fertility clinic. Genomic DNA was isolated from the samples using the PureLink Genomic DNA kit (Invitrogen, CA, USA). A total 5 ug of DNA was sonicated with 30×30 s, 30 s interval, using Bioruptor (Diagnode, NJ, USA). Sonicated DNA was end repaired using the End-It Kit (Epicentre, WI, USA) and A-tailed in a 50 µl reaction containing 1 mM dATP mix, 10 U of 3′ to 5′ exo- Klenow DNA polymerase (NEB, MA, USA). Adaptor ligation was performed in 50 µl reaction containing 300 mM pre-methylated adapters and 1000 Unit T4 DNA polymerase and incubated at 16°C overnight. Adaptor-ligated DNA was subjected to a size selection on a 3% NuSieve 3∶1 agarose gel. DNA marker lanes were excised from the gel and stained with SYBR Green (Invitrogen, CA, USA). 250–350 bp slices were excised from the unstained gel and purified using MinElute spin column (Qiagen, CA, USA). Size-selected fragments were bisulfite-treated using the EpiTect Bisulfite Kit (Qiagen, CA, USA) with minor modifications by adding 5 more cycles (5 min 95°C followed by 90 min at 60°C). After bisulfite conversion, DNA was eluted in 40 µl EB buffer and 0. 8 µl DNA was used for analytical PCR reactions to determine the minimum number of PCR cycles required to get enough material for sequencing. Final PCR products were purified on MinElute columns (Qiagen, CA, USA) and assessed on 4–20% polyacrylamide Criterion TBE Gel (Bio-Rad, CA, USA) and quantified using Qubit fluorometer (Invitrogen, CA, USA). The libraries were sequenced on the Illumina Genome Analyzer II (one lane for each sample) following the manufacturer' s instructions. The Pash 3. 0 software [65] was used to map the resulting reads to the reference human genome (NCBI 36. 1/UCSC hg18). Pash 3. 0 maps bisulfite reads natively. Reads were hashed considering the space of all possible kmers (e. g. for ATCT, the kmers ATCT, ATCC, ATCCC, ATCCT will be hashed). The forward and the reverse strands of the reference genome were streamed against the kmer reads hash, and regular mapping was applied. T' s in the reads can map to both C' s and T' s in the reference. Pash 3. 0 performs gapped mapping, being sensitive to both indels and base pair substitutions. Only reads that map uniquely and with at least 90% identity were used for subsequent analysis. Duplicate reads were removed across the same library. In total, 82. 39% of the reads for sample1 and 83. 02% for sample2 passed quality filters, achieving genome coverage at 1. 3× and 1. 2× respectively. Each chromosome of the reference human genome (NCBI 36. 1/UCSC hg18) was divided into 100 Kbp windows, excluding assembly gaps. The methylation levels in each sample were estimated by examining every CpG dinucleotide within each read mapping into each of the 28,705 windows. The methylation level of a window was estimated by dividing the number of methylated CpGs by the total number of CpGs found in reads mapping within the window. Windows with less than 20 CpG sampling events were excluded from consideration. The average of the two methylation maps was used as a representation of the sperm methylome to compare with the inferred germline methylation index. For control purposes, five other methylomes of human embryonic stem cells and fibroblasts were constructed from publicly available whole-genome bisulfite sequencing data [49], [50], using the same pipeline. The MI model is based on the fact that in mammals DNA methylation predominantly occurs in CpG dinucleotides, increasing the probability of transitions to TpG or CpA dinucleotides. The MI calculation by Sigurdsson et al. [48] implicitly uses mutability of CpGs in the human genome as an indicator of methylation in the germline. We apply this method of by integrating four million non-redundant SNPs from the HapMap project. Methylation index values were calculated for the same set of 100 Kbp windows used for sperm methylome construction to facilitate comparison. The sites of likely human-specific structural rearrangements were identified using the Genomic Triangulation method [46]. Non-human primate fosmid end sequences (FESs) from chimpanzee (CHORI-1251 library), rhesus macaque (Washington University Genome Sequencing Center (WUGSC) MQAD library), orangutan (WUGSC PPAD library) and marmoset (WUGSC CXAG library) were downloaded from the NCBI Trace Archives (http: //www. ncbi. nlm. nih. gov/Traces/). The FESs were mapped to the human genome (NCBI 36. 1/UCSC hg18) using BLAT [68] with the parameters: tileSize = 11, minMatch = 2, minScore = 100, minIdentity = 0, maxIntron = 50. Alignment scores were calculated for BLAT mappings using the parameters: match = +2, mismatch = −1, gap opening = −2, gap extension = −1. BLAT mappings with an alignment score less than 200 were removed from consideration. BLAT results were also filtered to remove ambiguous reads anchoring to more than 12 locations with an alignment score within 5% of the top alignment score. FESs that mapped at a distance consistent with fosmid clone insert size (25–50 Kbp) and in correct orientation were used to infer orthologous blocks. FESs were allowed to consistently map to multiple locations so that shared segments could be covered. Overlapping orthologous blocks were merged, based on genomic coordinates, into “matepair chains”. Matepair chain gaps due to human assembly gaps were removed. The remaining 522 matepair chain gaps indicated sites of likely human-specific structural genomic rearrangements. A 105 K Agilent oligo CGH array was designed for the purpose of routine diagnostic CNV testing at MGL. Probe sequences were chosen from the Agilent Technologies HD CGH database. Oligos were searched for multiple homologies to the human genome (NCBI 36. 1/UCSC hg18) to avoid cross-hybridization. Only unique oligos were selected for the array design. The whole genome sequence was divided into three types of regions covered with probes at different densities. The genes between DP-LCRs associated with genomic disorders were probed at the highest probe density (1 probe/10 Kbp, or at least 10 probes/gene for small genes). The second-highest probe density (1 probe/15 Kbp, or at least 10 probes/region) was assigned to the identified regions between DP-LCRs. These regions were required to be gene-containing, with a length from 1 Kbp to 10 Mbp, and flanked by direct paralogous LCRs that are ≥10 Kbp in length, and sharing ≥94% similarity. Probes with the same density were also assigned to the regions within the genome sequence coordinates of BAC/P1 artificial chromosome clones that had already been validated for clone arrays used in clinical practice (Baylor College of Medicine (BCM) BAC Chromosomal Microarray V6, including 1472 BAC and PAC clones for over 270 known genetic syndromes, 41 unique subtelomeric regions, 43 unique pericentromeric regions, and the mitochondrial genome). The third probe density (1 probe/31 Kbp) was assigned to all the other regions in the genome, so-called “backbone” regions. All the probes were selected to avoid the highly repetitive elements, the LCRs, and the known CNVs in major public databases: TCAG Database [70] of Genomic Variants hg18. v1 (http: //projects. tcag. ca/variation/), and UCSC Structural Variation database (http: //genome. ucsc. edu/cgi-bin/hgTrackUi? db=hg18&g=cnp). De-identified array intensity data obtained from 400 human DNA samples were made available to us by MGL. The data were analyzed using the Circular Binary Segmentation (CBS) method [71], which splits array intensity data along the genome sequence into segments with equal copy number that are significantly different from the neighboring regions. To determine the extent of association of hypomethylation with human-specific rearrangements/CNVs and to avoid possible artifacts due to the fixed 100 Kbp window size for sampling, the distribution of the methylation levels for the structural rearrangements/CNVs was compared to the distribution of the methylation levels of randomly picked segments (100 random samplings for each of the rearrangements/CNVs) of matched sizes on the same chromosomes (Figure S5AB, Table S7 rightmost two columns). To examine the extent of hypomethylation in the regions flanking rearrangements, the average methylation level for 10 Kbp regions sampled at increasing distances (from 10 Kbp to 100 Kbp) from rearrangement breakpoints were compared with 10 Kbp regions at corresponding distances from the randomly selected segments across the same chromosome (Figure 1C, Figure S20B, Figure S23). To estimate the potential contribution of hypomethylation and DP-LCRs regions to the occurrence of structural rearrangements/CNVs, the 100 Kbp windows covering the genome were each assigned to one or more of the following groups: (a) windows containing structural rearrangements/CNVs; (b) windows that are methylation deserts; and (c) windows containing regions between DP-LCRs. Statistical relative and attributable risks were calculated using intersections among these groups or their complements, with the universal set defined as all windows. Using corresponding letters to represent frequencies of these groups and their complements, the statistical relative risk of rearrangements/CNVs of hypomethylation was calculated as, and the statistical attributable risk was calculated as. Similarly, the statistical relative and attributable risks of rearrangements/CNVs as effect of DP-LCRs can be estimated by substituting b with c in the above formulas. Assuming that mutations are neutral, under an infinite allele model for populations at drift-mutation equilibrium, for any two loci in the genome, the ratio of heterozygosity rates H1 and H2 is equal to the ratio of mutation rates μ1 and μ2 [72] (Figure 4A). Therefore, the relative mutation rates at different loci can be estimated by observed relative heterozygosity rates. Structural heterozygosity rates were defined as follows. The normal copy number signal was interpreted as a homozygous major structural allele and any signal other than normal, either gain or loss, was interpreted as indicating presence of minor structural allele. The structural heterozygosity rate at one locus was calculated as 2pq (p = frequency of normal copy number state; q = frequency of abnormal copy number state). Since subsets of the 400 MGL samples and the HapMap samples contained trios or father/mother-offspring pairs, the following correction was applied to related samples: if aberration from normal at the same locus was found for related samples (parent and child), its occurrence was counted only once for each related sample trio/pair when calculating allele frequency. Only genes with valid RefSeq IDs that were detected within CNV heterozygous segments were considered for functional classification. The Database for Annotation, Visualization, and Integrated Discovery (DAVID [73], http: //david. abcc. ncifcrf. gov) was used to perform functional annotation enrichment analysis. The enrichment analysis was performed by interrogating the gene lists against the Gene Ontology Biological Process (GOBP), Gene Ontology Cellular Compartment (GOCC), Gene Ontology Molecular Function (GOMF), cell signaling pathways (KEGG Pathway) and the Swiss-Prot/Protein Informatics Resource (SP-PIR) databases. Using all human RefSeq genes as background, the gene categories with significant EASE score (<0. 01) and Benjamini correction value (<0. 1) in any of these databases were reported as enriched. To compare gene enrichment within specific structural mutability levels, genes with different CNV heterozygosity rates as detected by the oligo array data were binned into lists, each list corresponding to CNV heterozygosity rates in the range [x, x+0. 1) where x took values from 0 to 0. 4 in increments of 0. 02. Each gene list was analyzed using DAVID as described above. The tool GFINDer [74] was used for the genetic diseases and clinical phenotypes enrichment analysis. GFINDer exploits textual information within the Online Mendelian Inheritance in Man (OMIM) database. All human Entrez genes were used as background, and resulting categories with p-value less than 0. 05 were reported. Tests both without any p-value correction and with FDR correction were applied. This research did not involve Human Subjects. All data and materials obtained from humans were either anonymized or de-identified prior to use in this research project. | The human genome contains many loci with high incidence of structural mutations, including insertions and deletions of chromosomal segments. This excessive mutability has accelerated evolution and contributed to human disease but has yet to be explained. Segments of DNA repeated in low-copy numbers (LCRs) have been previously implicated in promoting structural mutability in specific disease-associated loci. Lack of methylation (hypomethylation) of genomic DNA has been previously associated with high structural mutability in gibbons and in human cancer cells, but the association with structural mutability in the human germline has not been explored prior to this study. Our analyses confirm the role of LCRs in promoting structural mutability on the genome scale but also reveal a surprisingly strong association of genomic instability with hypomethylation. Specifically, evolutionary analyses reveal that methylation deserts, the ∼1% fraction of the human genome with the lowest methylation in human sperm, harbor a tenfold higher number of structural mutations than genome-wide average. Moreover, the structural mutations in individuals diagnosed with schizophrenia, bipolar disorder, developmental delay, and autism are significantly more concentrated within hypomethylated regions. Our findings suggest a new connection between methylation of genomic DNA, selective structural mutability, evolution, and human disease. | Abstract
Introduction
Results
Discussion
Materials and Methods | genomics
biology
computational biology | 2012 | Genomic Hypomethylation in the Human Germline Associates with Selective Structural Mutability in the Human Genome | 12,938 | 291 |
Sphingosine-1-phosphate (S1P) is a crucial regulator of a wide array of cellular processes, such as apoptosis, cell proliferation, migration, and differentiation, but its role in Leishmania donovani infection is unknown. In the present study, we observed that L. donovani infection in THP-1 derived macrophages (TDM) leads to decrease in the expression of S1pr2 and S1pr3 at mRNA level. We further observed that Leishmania infection inhibits the phosphorylation of sphingosine kinase 1 (sphK1) in a time-dependent manner. Exogenous S1P supplementation decreases L. donovani induced ERK1/2 phosphorylation and increases p38 phosphorylation in TDM, resulting in a decrease in the intracellular parasite burden in a dose-dependent manner. On the other hand, sphK inhibition by DMS increases ERK1/2 phosphorylation leading to increased IL-10 and parasite load. To gain further insight, cytokines expression were checked in S1P supplemented TDM and we observed increase in IL-12, while decrease IL-10 expression at mRNA and protein levels. In addition, treatment of antagonist of S1PR2 and S1PR3 such as JTE-013 and CAY10444 respectively enhanced Leishmania-induced ERK1/2 phosphorylation and parasite load. Our overall study not only reports the significant role of S1P signaling during L. donovani infection but also provides a novel platform for the development of new drugs against Leishmaniasis.
Leishmaniasis is a neglected tropical disease that affects about 12 million people worldwide [1] It is caused by intracellular protozoa parasite Leishmania that invades macrophages and selectively impairs host’s critical signaling pathways for its successful intracellular growth and proliferation [2]. In particular, alteration in lipid metabolic pathways, lipid relocation, modification, and accumulation during Leishmania infection has been proven to be a critical step in the progression of disease [3]. In addition, Leishmania infection leads to increase in ceramide generation that further depletes cholesterol from the membrane and disrupts lipid rafts resulting in weak CD40 mediated signaling that leads to increased ERK1/2 phosphorylation and impaired antigen presentation to the T cells, worsening the diseased condition [4,5]. S1P signaling is emerging as a prominent regulatory pathway in cells that governs myriad of downstream signaling events [6]. The cascade begins with the generation of S1P from sphingosine by the action of sphK. Afterward, S1P translocates to the outer membrane and binds to its receptors, namely S1PR1-5, which triggers the small G-proteins associated with them [7]. These G-proteins then activates different signaling proteins resulting in numerous effectors functions. Being such a crucial regulator of cellular processes, it has been seen that this signaling pathway often dysregulated during several diseases [8,9]. In addition, the possible therapeutic strategy can be designed by carefully monitoring the impairment of this signaling during several diseased conditions [8,9]. S1P signaling has been well established in bacterial and viral diseases. Till date, no study addresses the role of S1P signaling in Leishmania donovani infection in human macrophages, The role of S1P in another intracellular pathogen such as Mycobacterium has been well studied in vitro and in vivo [10]. It was documented that S1P possess antimycobacterial properties such as reduction in intracellular growth by enhancement of phagolysosome acidification, induction of IFN-γ and enhanced antigen processing and presentation in monocytes [10,11]. Apart from this, it was shown that host sphingosine kinase regulates antimycobacterial responses and its inhibition leads to sensitization of RAW 264. 7 macrophage to infection due to reduced expression of anti-mycobacterial effector functions such as pp38, inducible nitric oxide synthase (iNOS) and Lysosome-associated membrane protein 2 (LAMP 2) [12]. In Bordetella pertussis infection in mice, S1P mediated signaling through S1PR resulted in reduced pathology due to the infection [13]. Similarly, during Yersinia pestis infection, activation of S1PR1 mediated signaling by SEW2871 limits intra-nodal trafficking of infection [14]. Alveolar macrophages from sphK1-knockout mice showed an increased burden of intracellular fungal, Cryptococcus neoformans [15], which suggests a protective role of S1P against C. neoformans infection. Currently, the role of S1P is unknown during Leishmania donovani infection in human macrophages. Also, the involvement of S1PR mediated signaling during Leishmania infection is not studied till date. Hence, in our study, we examined the role of S1P signaling during Leishmania donovani infection that should be helpful for generation of host-directed therapies against Leishmaniasis. To explore the role of S1P signaling in Leishmania donovani infection we first analyzed the expression of S1PR1-5 in infected and uninfected TDM. We next checked the phosphorylation of sphK1, the enzyme responsible for S1P production, in both TDM and human monocyte derived macrophages (hMDM). For further studies, we checked the phosphorylation of MAPK such as ERK1/2 and p38 in presence of S1P in infected and uninfected macrophages. In addition, cytokines such as Interleukin (IL) 10 and IL-12 were checked at mRNA and protein levels and parasite load was checked in infected macrophages upon S1P supplementation. For further confirmation of our results ERK1/2 and p38 activation was also studied upon inhibition of sphK or S1P supplementation in hMDM As observed earlier, the expression of S1PR2 and S1PR3 were decreased during the infection, we checked for ERK1/2 phosphorylation, cytokine secretion and parasite load in the presence of S1PR2 and S1PR3 inhibitors, JTE-013 and CAY10444, respectively.
The study was approved by Institutional Ethical Committee (IEC), Jamia Millia Islamia, New Delhi for the human subject participation. Each donor provided written informed consent for the collection of blood and subsequent analysis. RPMI 1640, M199, Fetal Bovine Serum (FBS), and penicillin and streptomycin were purchased from Life Technologies. DMS (N-N Dimethyl-sphingosine), JTE-013 and CAY10444 were purchased from Cayman chemicals. S1P was purchased from Tocris. Antibodies such as ERK1/2, phospho-ERK1/2, p38, and phospho-p38 were purchased from Cell Signaling Technology. Phospho-sphingosine kinase 1 and total sphingosine kinase 1 antibody was purchased from ECM biosciences. ELISA kit for IL-0 and IL-12 were purchased from BD Biosciences. The standard strain of L. donovani: (MHOM/IN/83/AG83) was maintained in M199 media (Life Technologies) with 25mM HEPES (Sigma) and supplemented with 10% heat-inactivated Fetal bovine serum (Life Technologies) with 1% penicillin and streptomycin (Life Technologies) at 22°C. Fourth to fifth-day culture was used to infect differentiated TDM or hMDM. The THP-1 cell line was maintained in RPMI 1640 medium (Life Technologies) supplemented with 10% heat-inactivated FBS (Life Technologies) and 1% streptomycin-penicillin (Life Technologies) at 37°C in 5% CO2. PMA (phorbol 12-myristate 13-acetate; Sigma) was used for differentiation of THP-1 cells into macrophages by incubating cells for 24 hrs with 5 ng/ml PMA at 37°C in 5% CO2 in flat-bottom 6-well tissue culture plates (BD Biosciences). Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats obtained from healthy donors (M. A Ansari Health Center, Jamia Millia Islamia, New Delhi, India) on Histopaque (Sigma). PBMCs were seeded on 6 wells plate and allowed to adhere. For differentiation of hMDM, non-adherent cells were removed by gentle washing and adherent cells were incubated with 5ng/ml GM-CSF (Pepro-Tech) at 37°C in 5% CO2 for 24 hrs and replenished with supplemented RPMI 1640 containing 10% FBS for 6 to 7 days. Macrophages were harvested and distributed into six-well plates at 2 × 106 cells/well. Stationary-phase promastigotes were added to differentiated cells at an infection ratio of 1: 10 to for 6 hrs initiate infection. Infected macrophages were further replenished with supplemented RPMI 1640 containing 10% FBS for additional 42 hrs for different studies. 1×105 cells per well were seeded onto 96-wells plate and was treated with PMA for diiferentiation for 24 h. Next day cells were washed and media was replaced with fresh media and cultured for additional 24 h for resting. After 24 h the cells were treated with different inhibitor used in the study DMS (5 μM), JTE-013 (10 μM), and CAY10444 (10 μM) for 42 hours. Cell viability was determined using the MTT cell viability assay. 3- (4,5-Dimethyl-2-thiazolyl) -2,5-diphenyl-2H-tetrazolium bromide, MTT (Sigma–Aldrich) was applied at in dark following 4 h incubation at 37°C. The MTT containing medium was replaced with 100 μl of isopropanol-HCl (0. 1N) and kept at 37°C for 10 min to solubilize the formazan crystals. The samples were transferred to 96-well plates and the absorbance of the converted dye was measured at 570nm. The percent cell viability of the control (non-treated) cells was taken as 100%. RNA was extracted by Trizol (Sigma) as per user’s information and was quantified on Biophotometer (Eppendorf, Germany) and 1 μg of RNA was used to prepare cDNA. Levels of Il-10 and Il-12 expressions were determined in the treated and untreated IM by quantitative PCR (qPCR), with β-Actin taken as an endogenous control (Primer sequences in S1 Table). q-PCR was carried out in a final volume of 10μL in Lightcycler 480 (Roche). The reactions were carried out with an initial denaturation step of 10 minutes at 95°C, followed by 40 cycles of denaturation, for 15 seconds, at 95°C, and annealing/extension for 1 minute, at 60°C. Relative gene expression was analyzed by the Livak method. For parasite load, TDM were seeded on sterile coverslips placed in 6 well culture plates. Infection was given at 1: 10 for 6 hrs and infected cells were treated with different concentration of inhibitors/ S1P and cultured for next 42 hrs. Infected cells were washed with PBS and then fixed with ice-cold methanol for 10 min and air dried. Wells were immersed in Giemsa stain for 45 min and washed 2–3 times with PBS. At least 200 cells were observed from a minimum of 15 randomly selected fields for each condition to determine the average number of parasites per macrophage. The parasite load was calculated in percent for the different condition after taking parasite load in control as 100%. After treatment, macrophages were washed twice with PBS and lysed with ice-cold lysis buffer (50 mM Tris-HCl, [pH 7. 4], 150 mM NaCl, 1% Triton-X, 1 mM Sodium Orthovanadate, 10 mM Sodium Fluoride, 1X protease inhibitor cocktail (Cell Signaling Technology). Lysates were centrifuged at 14,000 × g at 4°C for 15 min, and the resulting supernatants were transferred to fresh tubes and stored at −80°C until required. 40–50 μg of protein were used for western blotting. IL-10 and IL-12-specific enzyme-linked immunosorbent assay (ELISA) was performed to detect the level of secreted IL-10 and IL-12 in the cell-free supernatant obtained from different experiments using ELISA kits as per manufacturer’s instructions. Immunoblots and PCR products were analyzed using ImageJ software, National Institute of Health, version 1. 50i. The band intensity was calculated and normalized to corresponding control. The results shown are representation from a minimum of three similar experiments which generated reproducible data. The statistical analysis was performed using GraphPad Prism, version 6. 0 (GraphPad, San Diego, CA, USA). P-value of less than 0. 05 was considered significant. The error bars of the values represent ± SD from the replicates. Tukey' s multiple comparisons test and Student t-test were performed to ascertain the significance of the differences between the means of the control and the experimental groups.
The expressions of S1pr1-5 were checked upon Leishmania donovani infection in PMA differentiated human macrophages at mRNA level by semi-quantitative PCR. The macrophages were infection with Leishmania donovani promastigotes (multiplicity of infection [MOI] 1: 10, macrophage to parasite) for 6 hours. The non-internalized parasite was removed by gentle washing 2–3 times with PBS. The macrophages were further for additional 42 hours. After the given time, mRNA was extracted and cDNA was prepared using 1μg RNA. In addition, we evaluated the fold change in the expression of these receptors using beta-actin as endogenous controls to normalize the S1P receptors gene expressions. We found the expression of S1pr1-3 in both infected and uninfected macrophages, however, S1pr3 was detected at very low level (Fig 1A). In our study, we found that there was a significant decrease in the expression of S1pr2 and S1pr3 in infected macrophages (Fig 1B), while there was a not significant change in the expression pattern of S1pr1 in infected macrophages. In addition, we couldn’t detect the expression of S1pr4 and S1pr5 in both infected and uninfected TDM (Fig 1A). The level of sphK1 phosphorylation was checked in infected and uninfected TDM. The proteins were extracted from both the given experiment and analyzed by western blot for the detection of sphK1 phosphorylation. Total sphK1 was used as loading control for both the experiment. In our experiment, we observed that there was a significant decrease in the phosphorylation of sphK1 in the infected TDM (Fig 2A and 2B). Similarly, we also checked in the sphingosine kinase phosphorylation at various time intervals and we found a there was a time-dependent decrease in the phosphorylation level of sphK1 (Fig 2C and 2D). The decrease in the phosphorylation of sphK1 was also observed in hMDM 48h post infection (S1A and S1B Fig). As sphK1 phosphorylation is necessary for S1P biosynthesis, we checked the effect of S1P supplementation on parasite load, pro-inflammatory and anti-inflammatory response. Firstly, we examined ERK1/2 and p38 activation in uninfected macrophages and infected TDM with L. donovani on S1P supplementation (10 μM) for 48 h. After given time, proteins were extracted and analyzed for the detection of phospho-ERK1/2 and phospho-p38. In our study, we found that S1P supplementation showed diminished ERK1/2 and increased p38 phosphorylation, as measured by western blot analysis (Figs 3A and S2A and S2B). Next, we checked for the expression of cytokines such as Il-10 and Il-12 at mRNA level by both semi-quantitative PCR and real-time PCR. We found that S1P supplementation decrease Il-10 expression and increases Il-12 expression in infected macrophages (Figs 3B and S2C). In case of infected hMDM, we found that there was an increase in Il-12 expression upon S1P supplementation. However, we also registered a nonsignificant change in Il-10 expression at mRNA level (S4C Fig). We also determine the parasite load in presence of increasing doses of S1P and we found a dose-dependent decrease in the parasite burden in infected macrophages (Fig 3C). The most significant decrease in the parasite load was observed at maximum concentration i. e 10 μM (Fig 3C–3E). To further validate our results, we checked for ERK1/2 and p38 phosphorylation levels cytokine expression, and parasite load in presence of sphingosine kinase inhibitor, DMS. In case of hMDM, DMS pretreatment increases ERK1/2 phosphorylation, while reduces p38 phosphorylation in infected hMDM (Figs 4A and 4B and S4A and S4B). Additionally, we found that DMS pretreatment in infected macrophages resulted in a significant increase in Il-10 expression, whereas decreased Il-12 expression was observed as compared to infected hMDM (S4C Fig). In uninfected and infected TDM pretreated with 5 μM DMS, we found an increase in ERK1/2 phosphorylation (Figs 5A and S3A). Additionally, cytokines expression were checked by both semi-quantitative PCR and real-time PCR and it was found that Il-10 expression was increased while Il-12 expression was decreased in DMS pretreated infected TDM (Figs 5B and S3B). We further evaluate the parasite load in presence of increasing dose of DMS and we found that DMS pretreatment leads to increase in parasite load in a dose-dependent manner with significant increase in parasite burden at 5 μM (Fig 5C–5E). As shown in the previous experiment that the expression of S1PR2 and S1PR3 was decreased during Leishmania infection, we further checked for ERK1/2 phosphorylation and parasite load in presence of S1PR2 and S1PR3 specific inhibitors, JTE-013 and CAY10444, respectively [16,17]. We found that pretreatment with S1PR2 or S1PR3 inhibitors leads to a dose-dependent increase in the intracellular parasite load with a significant increase at higher dose i. e 10 μM (S5A and S5B Fig). In addition, ERK1/2 phosphorylation was also studied in JTE-013 (10 μM) and CAY10444 (10 μM) pretreated infected TDM alone and in combination. We found a significant increase in the phosphorylation of ERK1/2 in presence of these inhibitors, alone or in combination which further supports our study (Fig 6A and 6B). We further checked the parasite burden by pretreatment of both the inhibitors in combination or alone and we found that they both in combination contributed to further increase in the parasite burden (Fig 6C–6G). We next investigated secretion of IL-12 and IL-10 upon modulation of S1P signaling by DMS, S1P supplementation, and S1PR2-3 inhibition, alone or in combination. As expected, we found an increase in IL-12 secretion, whereas IL-10 secretion was decreased upon S1P supplementation (Fig 7A and 7B). In contrast, DMS pretreatment leads to increased IL-10 and decrease IL-12 (Fig 7A and 7B). In addition, we observed that by blocking S1PR3 there was a significant increase in IL-10 secretion, however, not significant changes in IL-10 in presence of S1PR2 inhibition was observed. However, on combinational doses of S1PR2-3 inhibitors, there was a more significant increase in IL-10 secretion (Fig 7A). Interestingly, by blocking S1PR2 we notice a significant decrease in the secretion of IL-12 in infected TDM while no significant changes were observed by blocking S1PR3. Furthermore, the secretion of IL-12 was further decreased in infected macrophages that were treated with S1PR2-3 inhibitors in combination (Fig 7B).
S1P signaling is emerging as a novel therapeutic target for numerous infectious diseases. Many studies have acknowledged the fact that S1P signaling plays a critical role in numerous infectious diseases and hence careful manipulation of the signaling might provide a breakthrough therapy. Hence in this study, we, for the first time, examined the role of S1P signaling in one of the most neglected tropical disease, Leishmaniasis. SphK1 is a cytosolic enzyme that on activation translocated to the plasma membrane leading to S1P production [18]. Earlier studied showed that sphK plays a critical role in viral infection. Non-structural proteins from bovine viral diarrhea virus (BVDV) have been shown to binds and inactivate sphK1 activity in a time-dependent manner, which favors viral growth by inhibition of apoptosis in Madin–Darby bovine kidney (MDBK) cells [19]. SphK1 inhibition was also reported in dengue virus (DENV) infection in HEK-293 cells [20]. It was demonstrated that thymocytes from Trypanosoma cruzi-infected mice showed a decrease in the activity of sphK1 and sphK2 at mRNA level [21]. S1P has been shown to regulate the phosphorylation of several MAPK (mitogen-activated protein kinase) including ERK1/2 and p38. S1P induces the phosphorylation of both ERK1/2 and p38 with maximum activation between 10–30 min of stimulation [22,23]. S1P induced activation of p38 is shown to be mediated by S1PR2 in SVEC endothelial cells line and pre-incubation with JTE-013 leads to inhibit the ability of S1P to induced p38 phosphorylation [23]. ERK1/2 and p38 phosphorylation, are established biomarkers for the progression and suppression of Leishmania infection [4,24]. Increase in ERK1/2 phosphorylation induces IL-10 production that has been associated with disease progression [4]. Moreover, it was found that IL-10 neutralization restores p38 activation and promotes parasite clearance, in contrary, IL-12 neutralization increases parasite burden [25–27]. Leishmania infection inhibits p38 phosphorylation that reduces Il-12 mRNA expression [28] and on the other hand, increase in p38 phosphorylation leads to IL-12 production that has been associated with disease suppression [4]. Altogether, IL-10 and IL-12 are considered to be important cytokines that regulate Leishmania donovani infection [29]. In this study, we observed that S1P results in decreased Leishmania induce ERK1/2 phosphorylation, while it increases p38 phosphorylation in macrophages. This further leads to induction of anti-leishmanial response by increase in IL-12 at mRNA and protein level, while a decrease in disease-promoting IL-10 at mRNA as well as in protein level, which altogether resulted in reduced parasite burden. DMS as a specific inhibitor of sphK which is reported by experiments done by other groups [30–32]. In our study, we also observed that DMS pretreatment inhibits sphK1 activity by inhibition of the phosphorylation of sphK1 in TDM (S6A Fig). Pharmacological inhibition of sphK by DMS or sphK1 specific siRNA has previously shown to decrease p38 phosphorylation in mouse macrophage cell line RAW264. 7 [33]. Similarly, inhibition of sphK inhibition by same above approach in mice also has been shown to down-regulate pro-inflammatory cytokines [34]. These reports suggested regulation of MAPKs and cytokines by inhibition of sphK. We have shown that Leishmania donovani infection in human macrophages results in decreased sphK1 phosphorylation in a time-dependent manner. In addition, pharmacological inhibition of sphK1 by DMS have been shown to exacerbate the infection by increasing disease promoting ERK1/2 induced IL-10 secretion, while on the other hand, decreasing p38 activated IL-12 production. Earlier, we showed that Leishmania infection significantly reduces the expression of S1pr2-3 in infected macrophages. To further gain insight on the role of these receptors mediated signaling, we checked the expression of ERK1/2 in presence of S1PR2 and S1PR3 specific antagonist JTE-013 (10μM) and CAY10444 (10 μM). Earlier studies have shown that inhibition of S1PR2 by JTE-013 results in increased ERK1/2 phosphorylation [35,36]. In our study, we found that there was an increase in the expression ERK1/2 on inhibition of S1PR2-3, alone or in combination. We also confirm this by checking the intracellular parasite burden and we found that combinational does of both the inhibitors increase the intracellular parasite burden in a very significant manner. Furthermore, we also observed increased IL-10 secretion in infected macrophages upon S1PR3 inhibition and combination doses of S1PR2 and S1PR3 inhibitor, but no observable changes were seen upon S1PR2 inhibition. Interestingly, inhibition of S1PR2 significantly decreases IL-12 secretion but no significant decrease was found upon S1PR3 inhibition. Notably, combinational doses of both these inhibitors significantly decrease IL-12 secretion and which may contribute to the increased parasite survival as observed by increase in parasite load. Hence this study suggests that S1PR2-3 inhibition reciprocally regulates IL-10/IL-12 balance during Leishmania infection which altogether makes macrophages more susceptible to Leishmania infection. In summary, we showed that S1P signaling plays a protective role in Leishmania infection and S1PR2-3 can be considered as novel and attractive therapeutic target against leishmaniasis. Our study suggested S1P mediated signaling abolishes Leishmania induce ERK1/2 phosphorylation resulting in low parasite load, on the other hand, S1P induces activation of p38 pathway that leads to IL-12 production for further clearance of the intracellular parasite (Fig 8). Additionally, blockage of S1PR2-3 mediated signaling by specific inhibitors in alone and in combinational doses resulted in activation of the ERK1/2 pathway leading to IL-10 production and increase parasite load. Hence, our study thus provides a novel and important aspect of S1P signaling during Leishmania donovani infection that may be helpful for the generation of a new line of anti-leishmanial drugs. | Leishmania donovani is an intracellular parasite which is internalized by host macrophages by subverting several intracellular signaling events. During infection suppression of p38 MAPK and activation of ERK1/2 MAPK have been acclaimed for survival and proliferation of these protozoan parasites. In this study, we show novel signaling pathways that interact with these MAPK that further contributes to determine the final fate of the disease. Sphingosine-1-Phosphate (S1P) is a bioactive lipid that binds to a family of G-protein coupled receptors known as S1P receptors. TDM infected with Leishmania donovani showed a decrease in the expression of S1PR1-3. Moreover, the enzyme that catalyzes S1P production, Sphingosine Kinase 1, showed decreased activation in a time-dependent fashion in infected cells. Furthermore, exogenously supplementation of S1P clears intracellular parasite burden by a decrease in ERK1/2 phosphorylation and IL-10 at mRNA and protein level. On the other hand, S1P induces anti-leishmanial response by activating p38 phosphorylation and IL-12 at mRNA as well as at protein level. To further gain insight on the receptors subtypes involved in the anti-leishmanial response, we specifically blocked S1PR2 and S1PR3. In this study, we found a tremendous increase in the parasite load as a result of increased IL-10 secretion and ERK1/2 phosphorylation on combination of these inhibitor doses. Taking together, our study thus suggested the possible involvement of S1PR2-3 during Leishmania donovani infection in human macrophages. These findings thus elaborate our knowledge in understanding the interaction of signaling intermediates during Leishmania infection which may lead to the discovery of novel therapeutic interventions. | Abstract
Introduction
Materials and methods
Results
Discussion | blood cells
phosphorylation
innate immune system
medicine and health sciences
immune cells
immune physiology
cytokines
immunology
parasitic diseases
parasitic protozoans
physiological processes
developmental biology
protozoans
leishmania
molecular development
white blood cells
animal cells
proteins
gene expression
leishmania donovani
immune system
biochemistry
eukaryota
cell biology
post-translational modification
physiology
secretion
genetics
biology and life sciences
cellular types
macrophages
organisms | 2018 | Sphingosine-1-phosphate signaling in Leishmania donovani infection in macrophages | 6,603 | 442 |
Mitosis in eukaryotic cells employs spindle microtubules to drive accurate chromosome segregation at cell division. Cells lacking spindle microtubules arrest in mitosis due to a spindle checkpoint that delays mitotic progression until all chromosomes have achieved stable bipolar attachment to spindle microtubules. In fission yeast, mitosis occurs within an intact nuclear membrane with the mitotic spindle elongating between the spindle pole bodies. We show here that in fission yeast interference with mitotic spindle formation delays mitosis only briefly and cells proceed to an unusual nuclear division process we term nuclear fission, during which cells perform some chromosome segregation and efficiently enter S-phase of the next cell cycle. Nuclear fission is blocked if spindle pole body maturation or sister chromatid separation cannot take place or if actin polymerization is inhibited. We suggest that this process exhibits vestiges of a primitive nuclear division process independent of spindle microtubules, possibly reflecting an evolutionary intermediate state between bacterial and Archeal chromosome segregation where the nucleoid divides without a spindle and a microtubule spindle-based eukaryotic mitosis.
Mitosis is a feature of all known eukaryotic cells, essential for the generation of viable progeny. Upon entry into mitosis the duplicated centrosomes that serve as microtubule organizing centers separate and organize a bipolar array of spindle microtubules. Microtubules are essential for chromosome segregation and eukaryotic nuclear division is not known to occur in their absence. Spindle microtubules attach to kinetochores, specialized protein complexes that assemble on centromeres of each chromosome [1]. After sister chromatid cohesion is lost at anaphase, microtubules pull the sister chromatids apart to opposite poles of the spindle. A surveillance system, the spindle checkpoint, monitors mitotic progression and prevents the onset of anaphase until all chromosomes have achieved bipolar attachment and can segregate [2]. The spindle checkpoint monitors kinetochore-microtubule attachment and a single detached or misaligned kinetochore is thought to be sufficient to trigger the checkpoint delaying the onset of anaphase and cytokinesis as well as blocking chromosome replication in the following cell cycle [3]. Defects in the spindle checkpoint result in the premature onset of anaphase and lead to chromosome mis-segregation. Genetic screens aimed at the isolation of mutants hypersensitive to microtubule depolymerizing drugs have identified many components of the spindle checkpoint [4], [5] such as mad1, mad2, mad 3, bub1, and bub3, which are highly conserved from yeast to humans [6]–[8]. The fission yeast Schizosaccharomyces pombe undergoes a closed mitosis with the nuclear membrane remaining intact and a microtubule-based spindle extending within the nucleus between two spindle pole bodies (SPB), the centrosome equivalent [9]. As in other organisms, a spindle checkpoint delays mitotic progression in the presence of microtubule defects which disrupt the spindle. The extent of the mitotic delay due to spindle checkpoint activation is variable and depends on the nature of the mitotic insult. The β-tubulin cold sensitive nda3-KM311 mutant has no spindle microtubules and blocks in pre-prophase with condensed chromosomes [10], [11]; in contrast the temperature sensitive nda3-1828 mutant proceeds through an aberrant mitosis and cytokinesis [12]. Here we show that in fission yeast, although depolymerization of spindle microtubules activates the spindle checkpoint, it delays mitosis only for a short time, especially at high temperature. Also, unexpectedly, in the absence of spindle microtubules, cells undergo an alternative nuclear division process and proceed into the next cell cycle. This process requires SPB separation, sister chromatid separation, and actin polymerization. We suggest that this process might represent a primitive kind of eukaryotic nuclear division.
We assessed the extent of mitotic and cytokinesis delay due to spindle checkpoint activation by treating wild type and spindle checkpoint deficient mad2Δ fission yeast cells [13] with 50 µg/ml of carbendazim (MBC), which disrupts the mitotic spindle by depolymerizing microtubules [14]. Cells that fail to segregate their chromosomes but escape the spindle checkpoint proceed through to cytokinesis and septation without completing mitosis, generating a “cut” phenotype with the septum cutting through the nucleus (Figure 1A) [15]. After MBC addition we observed a delay of cytokinesis of 2 h at 20°C and 45 min at 25°C, while at the higher temperatures of 32°C and 36. 5°C no significant delays were observed (Figure 1B). These results indicate that in fission yeast MBC-dependent spindle depolymerization activates the spindle checkpoint and delays cytokinesis only transiently at low temperatures and barely at all at high temperatures. We therefore tested whether at high temperature (36. 5°C) MBC treated cells could re-enter the next cell cycle and replicate their DNA. We used a temperature sensitive mutant, defective in septation initiation network signaling, which prevents cytokinesis and thus the cutting of the nucleus by the closing septum. At 36. 5°C, cytokinesis defective cdc11-119 cells treated with MBC continued DNA replication at a rate comparable to control DMSO-treated cells (Figure 1C). Similar results were obtained when additional MBC was added every hour to ensure the presence of active drug in the medium (unpublished data) and also when the cytokinesis defective mutants cdc4-8, cdc8-27, cdc12-112 (Figure 1D), as well as cdc7-24 and cdc3-124 (unpublished data) were used [16], [17]. In contrast no DNA replication was observed when cdc25-22 mutant cells blocked in G2 were treated with MBC (Figure 1D and unpublished data) [18], [19]. We conclude that if cytokinesis and septation are blocked, MBC-treated cells can proceed into the next cell cycle and undergo DNA replication. Consistently we observed no difference in DNA replication in cdc11-119 (Figure 1C) and cdc11-119 mad2Δ cells (Figure 1E) treated with MBC at the restrictive temperature. The spindle checkpoint monitors kinetochore-microtubule attachment and a single unattached kinetochore is sufficient to activate the spindle checkpoint delaying the metaphase to anaphase transition and mitotic exit. We reasoned that given that MBC treated cells re-enter interphase, the spindle checkpoint may be inactive. To test this we determined whether at 36. 5°C the checkpoint control was able to detect unattached kinetochores and therefore recruit checkpoint proteins such as Mad2 to the kinetochores. We investigated Mad2-GFP accumulation on kinetochores at 36. 5°C and found that 21%±4% of MBC treated cells had discrete Mad2GFP loci compared to only 6%±1% in control DMSO treated cells (Figure 2A), indicating that unattached kinetochores were present and were recognized by the checkpoint machinery. Next we tested whether fission yeast cells were able to block cell cycle progression at 36. 5°C in the presence of spindle microtubule defects. We used the kinesin 5-related mutant cut7-446, which fails to form a functional bipolar spindle due to lack of spindle microtubule interdigitation in the central region [20]. As shown in Figure 2B double mutant cdc11-119 cut7-446 cells blocked cell cycle progression and did not replicate their DNA during the 5 h time course. MBC treatment, however, was sufficient to allow cells to proceed through the cell cycle and replicate their DNA (Figure 2B), suggesting that fission yeast cells are competent to activate the spindle checkpoint at high temperature but not in the presence of MBC. As the spindle checkpoint senses unattached kinetochores we reasoned that in MBC treated cells either the checkpoint was never activated or residual undetected microtubules bound to kinetochores inactivated the checkpoint. To distinguish between these two possibilities we utilized a strain bearing a temperature sensitive mutation in the kinetochore protein Nuf2 [21]. The nuf2-2 allele at restrictive temperature abolishes microtubule binding, leaving the kinetochore competent to activate the spindle checkpoint [21]. As shown in Figure 2C, at their restrictive temperature cdc11-119 nuf2-2 cells delayed mitosis and re-entered interphase more slowly than cdc11-119 cells under the same conditions. However, when cdc11-119 nuf2-2 cells were shifted to the restrictive temperature in the presence of MBC, DNA replication occurred more readily than in DMSO treated control cells (Figure 2C). This indicates that failure to arrest in mitosis upon MBC treatment is unlikely to be caused by the inactivation of the spindle checkpoint by residual stable kinetochore microtubules. We next tested whether MBC treated cells completely lacked mitotic spindles. First, we stained for α-tubulin using antibodies and detected only very short microtubule remnants less than 1µm in length (Figure 3A), consistent with previously published data [14]. Second we could not detect mitotic spindles using a strain bearing a GFP-tagged version of α-tubulin (Atb2-GFP) (Figure 3B); only very short microtubule stubs were occasionally observed in the cytoplasm. However, despite the absence of mitotic spindle microtubules, staining with the DNA dye 4′, 6-diamidino-2-phenylindole (dapi) revealed the presence of multiple DNA masses in MBC treated cdc11-119 cells. After 5 h at 36. 5°C, 38% of the cell population had at least two well-separated DNA masses (Figure 3D). Visualization of the nuclear membrane with the marker Cut11-GFP (Figure 3A, C, D) [22] established that these DNA masses represented individual nuclear fragments enclosed by nuclear membrane. Time-lapse videos of cdc11-119 cut11-GFP atb2GFP cells at 36. 5°C showed that the nucleus was undergoing a division process. However, unlike a normal mitosis there was no elongation of the nucleus into a dumbbell shape. Instead the nuclear envelope acquired a wobbly ruffled aspect and then eventually pinched into two nuclear masses (Figure 3E and Videos S1, S2). To confirm that the division of the nucleus was occurring without spindle microtubules, we used the double mutant cdc11-119 cut7-446, which fails to undergo mitosis at the restrictive temperature due to formation of monopolar spindles [20]. After 5 h at 36. 5°C, 28% of the MBC treated cells had two nuclear masses (Figure 3F). Therefore, a nuclear division process takes place independently of a functional microtubule spindle. Occasional nuclear fragmentation has been reported in blocked nda3-KM311 cells [23], and we found that after 5 h at 19°C, dapi staining of nda3-KM311 cells showed that 30% of cells contained 2–3 nuclear bodies (Figure S1A, B). Membrane ruffling was also observed in these cells, as assayed using the nuclear envelope marker Uch2GFP [24], [25] (Video S3). We conclude that in the absence of spindle microtubules or a functional bipolar spindle, fission yeast cells can undergo an unusual nuclear division associated with ruffling of the nuclear membrane. Because the clearest characteristic of this process is fission of the nucleus, we have called it nuclear fission. As normal mitotic progression is under surveillance of the spindle checkpoint, we tested whether this control was operative during nuclear fission. We reasoned that if nuclear fission was subject to the spindle checkpoint, then mad2Δ checkpoint deficient cells would undergo division more efficiently. However, we observed no difference in the accumulation of binucleates between control cdc11-119 cells and cdc11-119 mad2Δ cells, suggesting that nuclear fission is independent of the spindle checkpoint control (Figure 3G and Figure S2). Fission of the nuclear membrane is the final event of mitosis, so we determined whether earlier events of mitosis were also taking place during nuclear fission. At the onset of mitosis, the duplicated SPBs separate and the mitotic spindle forms between them [9], [26]. To assess whether SPB separation occurred during nuclear fission, we used two SPB markers, Cut12-GFP (Figure 4A) [27] and Sad1-DsRed (Figure 5A, B) [28]. After 5 h in MBC 82% of the cdc11-119 cells had at least 2 SPBs (Table 1), and after 6 h cells with up to 8 SPBs were observed (Figure 4A). Significantly, almost all of the dapi-staining DNA masses (98%) were associated with at least one SPB (Figure 4A), suggesting that SPB separation was part of the process of nuclear fission. It has been previously shown that the SPBs facilitate nuclear envelope division during mitosis [25] and if SPB function is also required for nuclear fission, then impairing SPB maturation should block nuclear fission. To investigate this we monitored the appearance of binucleates in a cut11-2 mutant that fails to anchor the SPB in the nuclear envelope and exhibits defective maturation of a new SPB [22]. As shown in Figure 4D, nuclear fission was reduced from 38%±3% in cdc11-119 control cells to 7. 5%±4% in cdc11-119 cut11-2 cells, indicating that efficient nuclear fission, like mitosis, requires functional SPBs. A second important event of mitosis is sister chromatid separation, which is induced by degradation of the cohesin complex component Scc1 at the metaphase-to-anaphase transition [29]. We monitored chromosome I segregation during nuclear fission using a cen1-GFP expressing strain to mark the centromere of chromosome I. Within 5 h, all cells showed at least 2 cen1-GFP marked dots, establishing that separation of chromosome I centromeres was taking place (Figure 4B and Table 1). Centromeres were observed to segregate to different nuclear masses in 73% of the cells, which contained two nuclear masses. Similar results were obtained for chromosome II using a cen2-GFP strain (unpublished data). To monitor chromosome III segregation we used Clp1-GFP, which marks the nucleolus and co-segregates with chromosome III [30], and found that Clp1-GFP also partitioned to different nuclear masses in 70% of the cells with two nuclear masses (Figure 4C). These results indicate that sister chromatid cohesion is lost during nuclear fission, allowing sister chromatids to move away from each other and to segregate within the different nuclear masses. If chromatid separation is required for nuclear fission, then blocking the release of sister chromatid cohesion should reduce fission efficiency. In a separase mutant (cut1-645) [31] sister chromatids retain some cohesion and do not separate completely. After 5 h treatment with MBC, 11. 5%±2% cdc11-119 cut1-645 cells contained two nuclear masses compared to 43%±8% in control cdc11-119 cells (Figure 4E), demonstrating that nuclear fission is significantly reduced if chromosome separation does not take place. We conclude that during nuclear fission SPBs and sister chromatids separate in the absence of spindle microtubules, that some level of chromosome segregation can take place, and that for efficient nuclear fission functional SPBs and sister chromatid separation are required. During interphase, fission yeast centromeres cluster at the nuclear envelope in the vicinity of the SPBs [32] in a microtubule independent fashion [9], [32]. This clustering is lost upon entry into mitosis when kinetochores associate with mitotic spindle microtubules [32]. We considered that kinetochores might remain associated with SPBs in the absence of mitotic spindle microtubules. We therefore monitored centromere clustering at SPBs in MBC treated cells, using a strain bearing a centromere I marked with GFP and SPB tagged with Sad1-DsRed. In contrast to a normal mitosis we did not observe the cen1GFP signal dissociating from Sad1-DsRed labeled SPBs (Figure 5A), indicating that when microtubules are depolymerized by MBC treatment SPB-centromere association persists. We confirmed that the association was maintained for all three fission yeast centromeres using an ndc80-GFP bearing strain to label all three chromosomes simultaneously. In the presence of MBC at 36. 5°C, no dissociation of the Ndc80-GFP signal from Sad1-DsRed tagged SPBs was detected (Figure 5B), suggesting that centromere clustering near the SPB persists during nuclear fission. However, it was possible that centromeres transiently dissociate from the SPBs upon mitotic commitment and then are quickly recaptured. To investigate this possibility we performed time lapse imaging of SPBs and centromeres in the presence of MBC. Time-lapse imaging of cdc11-119 sad1DsRED ndc80GFP showed that in the presence of MBC, SPB separation occurred without centromeres declustering (Figure 5C and Video S4). The two SPBs appeared to move apart from each other with their associated set of centromeres. As declustering occurs upon mitotic commitment, we further analyzed kinetochore clustering in a cdc11-119 cut11mcherry ndc80GFP strain. Cut11mcherry accumulates on SPBs from early mitosis through to the metaphase to anaphase transition, and therefore acts as a marker of mitotic commitment. We observed that MBC treated, early mitotic cells (as defined by SPB-Cut11mcherry accumulation) contained Ndc80GFP labeled dots that remained in close proximity to the SPB (Figure 5D and Video S5). The Ndc80GFP labeled dots moved away from each other only after Cut11GFP came off the SPB (Figure 5D 90″). Thus, during nuclear fission, unlike mitosis, centromeres remained clustered around the SPB. If centromere-SPB association is important for nuclear fission, then a mutant, which fails to maintain clustering of kinetochores at the SPBs, should impair nuclear fission. We used an ima1Δ strain, which disrupts kinetochore clustering at SPBs in 75% of cells [33]. We observed that after 6 h at the restrictive temperature 14%±2% of cdc11-119 ima1Δ MBC treated cells underwent nuclear fission compared with 34%±4% in control cdc11-119 cells. MBC treated cdc11-119 ima1Δ cells showed no significant change in nuclear ruffling compared to cdc11-119 cells (Video S6). Thus, failure to maintain the association between centromeres and SPBs significantly reduces the efficiency of nuclear fission. Given that there are no microtubules present to generate the force necessary for nuclear fission, we ascertained whether nuclear fission required filamentous actin. Cdc11-119 cells were treated with MBC for 2 h, followed by addition of either DMSO or 10 µM latrunculin A (LA), an actin depolymerizing drug. In MBC treated cells, we observed that LA treatment completely blocked nuclear fission (Figure 6A). Nuclei of LA treated cells were rounder than those of control DMSO treated cells (Figure 6B) and time-lapse videos of cdc11-119 cut11GFP cells showed no nuclear membrane ruffling (Figure 6C and Video S7). Mitosis proceeded normally in the presence of 10 µM LA when MBC was not present. As in cdc11-119 cells, LA treatment blocked nuclear fission in nda3-KM311 cells at 19°C (Figure S1C). Despite the dependency of nuclear fission on the actin cytoskeleton we were unable to detect actin structures in or around the nucleus (Figure S3). We next examined whether SPB separation was also affected in LA treated cells, by blocking cells expressing the SPB marker Sad1-RED for 1 h in the presence of MBC and then treating them either with DMSO or 10 µM LA. After 1 h MBC treatment, 24%±3% of cells had 2 SPBs. After a further 2 h in the presence of LA, 23%±2% of cells had 2 SPBs, compared to control DMSO treated cells, which showed 64%±2% of cells with 2 or more SPBs (Figure 6D). The SPBs in LA treated cells also remained in close proximity to each other compared to control cells (Figure 6E). Similarly we observed no increase in the number of cells with 2 Ndc80-GFP labeled kinetochores in LA treated cells (unpublished data). Thus, we conclude that nuclear fission depends on filamentous actin.
This work describes an unexpected process whereby in the absence of a mitotic spindle the fission yeast nucleus can undergo nuclear division. This process, which we have called nuclear fission, requires SPB maturation and sister chromatid separation and leads to some sister chromatid segregation. We propose the following mechanism for nuclear fission (Figure 7). As cells exit G2, sister chromatids remain clustered at the SPBs. The duplicated SPBs separate slowly by moving within the nuclear membrane, and as the sister chromatids lose cohesion, they move apart as a consequence of their association with SPBs. This mechanism assumes that the two sister chromatids are associated non-randomly with different SPBs. As the chromatids separate through the nucleus, the nuclear membrane deforms around the two DNA masses generating two nuclear bodies. Preventing SPB maturation or maintaining sister chromatid cohesion interferes with the separation of DNA masses and blocks nuclear fission. Although nuclear fission occurs in the context of closed mitosis, it is reminiscent of the formation of multiple nuclei in metazoan cells when a portion of the DNA becomes separated form the bulk of the chromosomes and is encapsulated in a separate nuclear entity. This happens under pathological conditions, for example in cancer cells as a consequence of inappropriate chromosome segregation or chromosome breakage [34], [35], or during oocyte meiosis and early developmental stages in mice deficient for the chromokinesin Kid [36], which show incomplete chromosome compaction. It also occurs in more physiological situations such as the formation of karyomeres in the early embryonic divisions of Xenopus, sea urchin, and polychaetes, where individual chromosomes are separated and engulfed by the nuclear envelope [37], [38]. Important for the formation of separate nuclear entities are the necessity for a minimal distance between DNA masses and for sufficient nuclear membrane to be available. Similarly in fission yeast, nuclear fission occurs only when SPB-chromosomes masses move away from each other and when lipid biosynthesis is up-regulated during the expansion of the nuclear envelope to accommodate the elongating spindle [39]–[41]. Understanding the regulation of nuclei formation during nuclear fission might shed light on the mechanism that controls the formation of a single nucleus around each chromosome complement at the end of mitosis. Actin polymerization, which is required for nuclear fission, might be involved in the membrane redistribution associated with the increase in nuclear envelope area observed during early anaphase B. Nuclear envelope ruffling could be a consequence of a rapid redistribution of membranes between ER and nuclear envelope. During nuclear fission, ruffling is most obvious after cut11GFP comes off the SPBs at a stage corresponding to anaphase B when spindle elongation takes place during a normal mitosis. Therefore, nuclear envelope ruffling might be expected to be more obvious during nuclear fission when there is no spindle elongation to stretch the nuclear membrane. Consistently, nuclear ruffling in fission yeast is also observed during mitosis if spindle elongation is blocked, as in the kinetochore mutant nuf2-3 (our unpublished observation). Nuclear membrane expansion was also observed in budding yeast cells, blocked in mitosis with nocodazole to depolymerize microtubules. It has been suggested that this expansion is a consequence of the up-regulation of lipid biosynthesis normally taking place during mitosis [42]. Nuclear membrane extensions appear also upon deregulation of phospholypid biosynthesis by spo7 inactivation [41], [42]. However, such extensions are not observed if spo7 inactivation takes place during anaphase probably because of the incorporation of extra membrane into the elongating nucleus [41]. Further studies will be required to clarify whether phospholipid biosynthesis has a role in nuclear fission or if actin is involved in the nuclear expansion observed during a normal mitosis. However, there could be other roles for the involvement of actin in nuclear fission. Actin could generate either a pushing force causing nuclear membrane distortion as is seen during lamellipodia protrusion [43] or a pulling force separating SPBs and their cargo of chromosomes. In this context it is of interest that bacterial chromosome segregation is driven by polymerization of the actin-like MreB/ParM protein [44], [45] and also that in vertebrates the driving force for centrosome separation is provided by actin filaments [46]–[48]. Differently from a normal mitosis, nuclear fission appears not to be under spindle checkpoint control and takes place irrespective of checkpoint engagement. We observed nuclear fission both in nda3-KM311 cells, which activate the spindle checkpoint blocking cells in pre-prophase, and in MBC treated cdc11-119 cells, which do not delay mitotic exit. Interestingly, in both circumstances cells accumulate Mad2 on kinetochores, suggesting that unattached kinetochores have been detected. However in MBC treated cells, where microtubules are almost completely depolymerized, no mitotic delay is observed and cells re-enter interphase. This result suggests that Mad2 accumulation on kinetochores, although necessary for checkpoint activation, might not be sufficient to maintain the mitotic block. Alternatively the checkpoint might be activated normally but only transiently. As a monopolar spindle does activate the checkpoint (cut7-446) and we have excluded the possibility that the binding of residual kinetochore microtubules inactivates the checkpoint (nuf2-2), something other than microtubules would have to be involved in the inactivation. It has been shown that in budding yeast activation of the APC inhibits the mitotic checkpoint through APC-mediated degradation of the checkpoint kinase Mps1 [49]. This feedback between APC and the spindle checkpoint implies that if enough APC is activated the checkpoint is turned off. In S. pombe, such a mutual inhibition would mean that if the nuclear fission process were to generate enough active APC, then the checkpoint block would be overcome. Alternatively, maintenance of the SPB-kinetochore interaction might cause kinetochores to be less readily available for full strength checkpoint activation, leaving sufficient levels of active APC in the nucleus to turn off the checkpoint. Further studies will be required to understand how MBC-mediated microtubule depolymerization inactivates the spindle checkpoint in fission yeast. We speculate that nuclear fission might be a vestige of an ancient mechanism of nuclear division that predates the appearance of a mitotic spindle. It might reflect an evolutionary intermediate state between the mechanism of chromosome segregation seen in bacteria and Archea [44], [50] and that seen during mitosis in eukaryotic cells. In the intermediate evolutionary state, the replicated sister chromatids would remain attached to the centrosomes and become segregated by movement of the divided centrosomes within the nuclear membrane. Later in evolution, addition of a mitotic spindle between the centrosomes would have increased the efficiency of centrosome separation and the microtubule-based polar movement of sister chromatids would have led to a more efficient and accurate conventional mitosis. It has been observed that in the absence of the tubulin homologue FtsZ, the L-form of Bacillus subtilis acquires an unusual mode of proliferation with cells undergoing membrane ruffling prior to the formation of protrusions, which then resolve into independent round bodies [51]. It has been suggested that this extrusion-resolution mode could either be driven by force generated by the actin homologue MreB or by active chromosome segregation followed by collapse and resealing of the membrane. These mechanisms may be relevant to the nuclear fission process we have observed in fission yeast in the absence of spindle microtubules. A spindle independent mechanism (SIM) has also been reported for nucleolar segregation during Aspergillus nidulans mitosis [52]. The mechanism underlying SIM in A. nidulans is not yet clear, but the nuclear envelope is believed to play a critical role to generate the force necessary for nucleolar separation. In summary, we suggest that nuclear fission represents the vestiges of a primitive nuclear division process that existed early in the eukaryotic lineage prior to the evolution of a mitotic spindle and mitosis as known today.
All S. pombe strains used in this study are listed in Table S1. Standard methods were used for growth and genetic manipulation [53]. All experiments, unless otherwise stated, were performed in YE4S (yeast extract with added 250 mg/l histidine, adenine, leucine, and uridine). Cells were grown at 25°C to 1–2×106 cell/ml density before shifting to the restrictive temperature (36. 5°C). After 3 h (unless otherwise stated), cultures were split in two and treated with either 50 µg/ml MBC (freshly made in DMSO) or DMSO at 36. 5°C, unless otherwise stated. It should be noted that some batches of MBC are more toxic for cells and these were not used in this study. For lat A treatment, following a 2 h block at 36. 5°C in the presence of either 50 µg/ml MBC or DMSO, cells were treated with 12. 5 µM lat A. For immunofluorescence, cells were collected by filtration and then fixed. Cells were fixed in −80°C methanol for 1 h and then processed as previously described [54]. For microtubule detection, TAT1 antibody (a-tubulin antibody; a kind gift of Prof. K. Gull) was used at 1∶200 dilution and Alexa fluor 546-linked anti-mouse (Molecular Probes) at 1∶1000 dilution as secondary antibody. For actin staining, cells were fixed by adding formaldehyde (final concentration 3. 7%) to the medium for 25 min, washed twice with PEM (100 mM Pipes, 1 mM EGTA, 1 mM MgSO4 pH = 6. 9), permeabilized with 1% Triton X-100, washed twice with PEM, and stained with rhodamine phalloidin (Molecular Probes). For dapi staining, cells were either heat fixed (70°C) or fixed in 70% cold ethanol, then re-hydrated in distilled water, and stained with 2 µl of 50% glycerol, 0. 1 M Tris pH 8 containing 1 µg/ml dapi. For immunofluorescence, cells were fixed in cold methanol at −80°C overnight and then processed as previously described [54]. Images were taken using a Deltavision microscope. The percentages are averages of 3–8 experiments and the standard errors were calculated and reported. For live imaging cells were attached to coverslips using soya bean lectin (100 µg/ml) and imaged in minimal medium either containing DMSO or 50 µg/ml MBC at 36. 5°C, using a Deltavision microscope supplied with a temperature controlled chamber. | The process of cell division, mitosis, ensures that chromosomes are accurately segregated to generate two daughter cells, each with a complete genome. Eukaryotic cells use a microtubule-based mitotic spindle to ensure proper chromosome segregation. In the fission yeast Schizosaccharomyces pombe, mitosis is “closed”: that is, the nuclear envelope does not break down, and the mitotic spindle forms within the nucleus. Unexpectedly we have found that in certain circumstances division of the fission yeast nucleus and progression into the next cell cycle can take place without the mitotic spindle. We call this nuclear division process “nuclear fission” because the nucleus separates into two bodies. We show that nuclear fission requires filamentous actin and functional spindle pole bodies, which are the fission yeast equivalent of the centrosome in other organisms. We also show that nuclear fission requires sister chromatid separation and is accompanied by some level of chromosome segregation. We propose that nuclear fission is a vestige of a primitive nuclear division process and might reflect an evolutionary intermediate between the mechanism of chromosome segregation that takes place in bacteria and the microtubule-based mitosis of modern eukaryotes. | Abstract
Introduction
Results
Discussion
Materials and Methods | cell biology/cell growth and division
cell biology
cell biology/cytoskeleton | 2010 | Fission Yeast Cells Undergo Nuclear Division in the Absence of Spindle Microtubules | 8,113 | 316 |
The Saccharomyces cerevisiae RNA-binding protein Nab4/Hrp1 is a component of the cleavage factor complex required for 3′ pre-mRNA processing. Although the precise role of Nab4/Hrp1 remains unclear, it has been implicated in correct positioning of the cleavage site in vitro. Here, we show that mutation or overexpression of NAB4/HRP1 alters polyA cleavage site selection in vivo. Using bioinformatic analysis, we identified four related motifs that are statistically enriched in Nab4-associated transcripts; each motif is similar to the known binding site for Nab4/Hrp1. Site-directed mutations in predicted Nab4/Hrp1 binding elements result in decreased use of adjacent cleavage sites. Additionally, we show that the nab4-7 mutant displays a striking resistance to toxicity from excess copper. We identify a novel target of alternative 3′ pre-mRNA processing, CTR2, and demonstrate that CTR2 is required for the copper resistance phenotype in the nab4-7 strain. We propose that alternative 3′ pre-mRNA processing is mediated by a Nab4-based mechanism and that these alternative processing events could help control gene expression as part of a physiological response in S. cerevisiae.
3′ pre-mRNA processing severs the nascent transcript from the elongating polymerase to liberate a new polyadenylated mRNP particle [1,2]. Since the placement of the cleavage site can include or remove RNA sequences that influence transcript stability, transcript localization, protein expression, or protein localization, cleavage site choice can have profound influences on gene expression. In higher eukaryotes, regulation of gene expression via alternative 3′ pre-mRNA processing plays essential roles in development and tissue-specific functions [3–7]. In fact, approximately 50% of human genes are suspected to have alternative polyadenylation sites based on bioinformatics analysis [8,9]. In Saccharomyces cerevisiae, the in vivo mechanism of cleavage site selection and the physiological consequences of alternative cleavage remain largely unknown. Only a few cases of multiple polyadenylation sites have been confirmed by having their 3′ ends mapped [10–12]. Interestingly, the site of polyadenylation for half a dozen of these alternatively cleaved transcripts are sensitive to the growth condition of the cell [10,11]; raising the tantalizing possibility that alternative polyadenylation may be dynamic and regulated. In both yeast and metazoan systems, the multi-subunit cleavage and polyadenylation machinery is assembled on virtually all RNA polymerase II transcripts. The protein components between yeast and mammalian systems that were once thought to be so divergent are now known to be relatively well-conserved [13,14]. In striking contrast, the sequence elements that recruit the cleavage machinery are phylogenetically diverged. In S. cerevisiae, five sequence elements have been identified that contribute to cleavage site selection: the efficiency element, the positioning element, the near-upstream site, the cleavage site, and the near-downstream site [14–16]. However, no single element is absolutely required and each element can be degenerate, making it difficult to accurately predict the 3′ end for most yeast transcripts. These issues are compounded when considering alternative 3′ pre-mRNA processing signals, which may diverge more significantly than a typical 3′-end processing site. In contrast, the AAUAAA hexamer found in mammalian sequences has long been thought to be an invariant signal for polyadenylation. Interestingly, recent bioinformatics analysis suggests that the variability of mammalian polyadenylation signals may be more akin to those found in S. cerevisiae [17]. Almost a dozen variants to the AAUAAA hexamer have been suggested to play roles in polyadenylation [18]. It has been suggested that the variability in sequences may be utilized as part of the mechanism of alternative 3′ pre-mRNA processing [9,17]. Two known mechanisms of regulated 3′ pre-mRNA processing in metazoans are based upon controlling key components of the cleavage machinery. The best studied example of regulated alternative processing in mammalian cells involves CstF64 and the transcript for immunoglobulin M [4,5]. In resting cells, low levels of CstF64 allow the production of a long form of the transcript that encodes a transmembrane domain, leaving the protein tethered to the cell. Upon B-cell activation, levels of CstF64 rise, which causes a weaker, upstream cleavage site to be used, eliminating the transmembrane domain and creating a secreted protein. The heterodimer CFIm is another component of the mammalian cleavage machinery recently discovered to influence cleavage site selection [19,20]. In fact, CFIm can influence the cleavage site selection of one of its own subunits [19]. Although the consequences of this potential auto-regulation remain unknown, the data underscore the notion that control of a component of the cleavage machinery can regulate cleavage site selection. The closest S. cerevisiae ortholog of CFIm may be Nab4/Hrp1 [20]. Nab4 is an essential heterogeneous nuclear ribonucleic acid (hnRNP) protein that can shuttle in and out of the nucleus [21]. In addition, Nab4 has been biochemically isolated as part of the cleavage factor complex [22,23]. Although the involvement of Nab4 in 3′ pre-mRNA processing is undisputed, the precise function of Nab4 during this process remains controversial. It is unclear if Nab4 is involved in the cleavage reaction itself or is only required to correctly position the cleavage site. When Nab4 is excluded from in vitro cleavage reactions, the activation of cryptic cleavage sites dramatically increases, leading to the hypothesis that Nab4 is involved in the discrimination between correct and cryptic sites [24]. In addition to its role in 3′ pre-mRNA processing, Nab4 has been implicated in mRNA export and nonsense-mediated decay [25]. Unlike other members of the cleavage and polyadenylation machinery, Nab4 appears to be retained on the message after 3′ pre-mRNA processing and is escorted with the message out of the nucleus. Once in the cytoplasm, it disengages from the transcript and is recycled back into the nucleus by the import receptor Kap104 [26]. It remains unknown whether the roles of Nab4 in export and decay are downstream consequences of its role in 3′ pre-mRNA processing or if they represent independent functions. To better understand the mechanism and consequences of alternative 3′ pre-mRNA processing in S. cerevisiae, we analyzed the function of Nab4. We show that alternative 3′ pre-mRNA processing is sensitive to the levels of this component of the cleavage complex, similar to the regulated cleavage site selection seen in mammalian cells. In addition, we have uncovered an unexpected role of Nab4 and alternative 3′ pre-mRNA processing in the response to toxic copper concentrations. These data support the hypothesis that not only does alternative cleavage occur, but that the control of alternative cleavage is important for cell physiology.
We previously identified transcripts preferentially associated with hnRNPs (heterogeneous nuclear ribonucleic acid proteins) on a genome-wide scale. One of the transcripts that co-immunoprecipitated with Nab4 is SUA7 [27]. SUA7, which encodes the transcription initiation factor TFIIB, is known to have two different cleavage sites in the 3′ UTR [11]. To determine if Nab4 could play a role in alternative 3′ pre-mRNA processing in vivo, we analyzed the SUA7 transcript in a nab4-7 mutant. Total RNA samples were collected from NAB4 and nab4-7 cells in both log phase and stationary phase. RNA samples were analyzed by 3′ rapid amplification of cDNA ends (RACE) for qualitative analysis (Figure 1A) and by Northerns for quantitative analysis (Figure 1B). Notably, we found that the nab4-7 strain displays an aberrantly high ratio of the long form to short form for the SUA7 transcript. In our strain background, the ratio of long form to short form is 1. 65 in exponentially growing cells. In the presence of the nab4-7 mutation, the ratio increases to 2. 37. This altered ratio is observed at the permissive temperature, where there is no significant growth defect. Interestingly, in wild-type cells the predominant cleavage site used is sensitive to the growth condition of the cell. The long form predominates during log phase growth and the short form predominates during stationary phase growth [11]. As expected, the control strain displayed the shift from long form to short form during stationary phase growth, reflected in the change in ratio from 1. 65 in log to 0. 89 in stationary phase (Figure 1). In striking contrast, the nab4-7 strain showed no appreciable shift in cleavage site usage (a ratio of 2. 37 in log phase and 2. 27 in stationary phase). Therefore, Nab4 can influence cleavage site usage in vivo for a transcript that is known to have alternative 3′ pre-mRNA processing sites, and a mutant version of Nab4 is defective in a known shift in cleavage site usage. We next tested if increased expression of an otherwise wild-type version of Nab4 could influence cleavage site selection. Total RNA was collected from strains containing a plasmid with Nab4 under the control of a galactose-inducible promoter. Overexpression in the inducible strain was observed by Western blotting; the control strain showed no change in Nab4 levels over the course of the induction (unpublished data). Using 3′ RACE analysis, we found that as Nab4 is overproduced, the relative ratio of long SUA7 transcript to short transcript sharply decreases (Figure 2). Therefore, as excess Nab4 is produced, cleavage at the upstream site to produce a shorter transcript is favored. The shift in cleavage site usage is not due to the change in carbon source, as the addition of galactose to an otherwise identical strain without overexpressed Nab4 shows no such shift in SUA7 cleavage. Therefore, the cellular concentration of wild-type Nab4 can influence cleavage site selection in vivo. In the simplest scenario, Nab4 would exert its influence on cleavage site selection through a direct protein-RNA interaction with the substrate transcript. To identify potential Nab4-binding sites, we took an unbiased bioinformatics approach using the word-finding algorithm Regulatory Element Detection Using Correlation with Expression (REDUCE) [28]. This analysis was similar to previous work in which we identified transcripts that were preferentially associated with Nab4 using an IP-microarray approach [27]. However, in the current analysis, the Ty and Y′ elements that are associated with multiple RNA-binding proteins were eliminated from the dataset prior to REDUCE analysis. We used the algorithm on this filtered dataset to identify short words, up to seven letters long that were overrepresented in transcripts associated with Nab4. The genome was searched from 600 bases upstream to 300 bases downstream of every gene contained in our microarray dataset. This region has been estimated to include the 3′ UTR for 98% of the genome [16]. Motifs that were identified in three independent microarray-REDUCE analyses are shown in Table 1 with the lowest p-value calculated for the given motif. We identified the word “TATATAA” that is nearly identical to the known Nab4-binding motif of “UAUAUA. ” In addition to the TA-rich motif, two other related motifs were identified that contain single nucleotide deviations from the core UA-repeat motif of the known Nab4-binding site, “ATAAATA” and “TACATA. ” The core UA-repeat motif occurs more frequently in UTR sequences than in open reading frame (ORF) sequences. The deviant motifs display a similar bias as the more canonical motifs in their occurrence in UTR sequences, indicating that the deviant motifs could also be used as 3′ pre-mRNA processing elements (Figure 3). The bias to noncoding regions is not likely due to the AT bias of the S. cerevisiae genome, as a similar motif identified for the hnRNP protein Nab2, “AAAAAAG, ” showed no such 3′ bias. Given that our analysis implicated Nab4 in cleavage site selection in vivo, we wanted to determine if our newly found Nab4 motifs contributed to cleavage site selection as well. To do so, we created mutations in the potential Nab4-binding sites from the SUA7 3′ UTR. The SUA7 transcript contains three potential Nab4-binding sites that are upstream of the known cleavage sites. The first motif, “AAAAAT, ” is located 116 nucleotides from the stop codon and is almost identical to a Nab4 motif identified in our bioinformatics analysis, “ATAAATA. ” The second motif “TACATA” and the third motif “TATATATATA” lie directly adjacent to each other, 224 nucleotides from the stop codon. The second motif is identical to a motif identified above and the third motif is an extended form of the canonical efficiency element (Figure 4A). Six nucleotides of each motif were replaced with an unrelated sequence. By replacing instead of removing the motif, the spacing in the primary sequence was preserved between other elements in the 3′ UTR. Due to technical reasons, we were unable to obtain a deletion of the second motif. Total RNA was collected from strains containing a plasmid encoding the only copy of the SUA7 gene, and 3′ RACE analysis was performed. As shown in Figure 4, the removal of either the first motif or the third motif resulted in a decrease in cleavage at the adjacent downstream cleavage sites (Figure 4). Mutation of motif 1, the first motif downstream of the stop codon, resulted in a relative decrease in the amount of short form (Figure 4B, lane 3). Likewise, mutation in motif 3, the farthest from the stop codon, resulted in a relative decrease in the amount of long form (Figure 4B, lane 4). The relative shifts in cleavage site usage due to mutations in the Nab4 motifs are apparent in both log and stationary phase (Figure 4B). Since cleavage at adjacent sites are reduced, but not eliminated, the Nab4 motifs cannot be absolutely required to elicit cleavage. Considering that multiple sequence and protein elements work in concert to determine cleavage, the lack of a strict requirement of the element to regulate cleavage site usage is perhaps not surprising. Nonetheless, these data support a role for the influence of Nab4 on cleavage site selection in vivo. Although 3′ pre-mRNA processing reactions uncovered a role for Nab4 in establishing correct cleavage in vitro, two non-mutually exclusive mechanisms have been proposed: Nab4 could actively promote correct cleavage or it could inhibit incorrect cleavage. Our in vivo results suggest that the Nab4 motifs promote cleavage events. If alternative cleavage site selection is important for a physiological response, then nab4 mutants that alter cleavage site usage may be defective for that particular response. Interestingly, Nab4-associated transcripts containing the “TACATA” motif are enriched in the functional category of transition metal ion transport, as defined by the Gene Ontology term finder available on the Saccharomyces Genome Database (http: //www. yeastgenome. org) (unpublished data). To determine if Nab4 plays a role in the expression of genes in this category, we tested if the nab4-7 mutant displays a growth phenotype in the presence of excess copper. The response to copper stress is an ideal assay as cells are sensitive to both the absence and the excess of copper. Serial dilutions of NAB4 and nab4-7 mutant strains were grown on a series of CuSO4 -containing plates at room temperature. Surprisingly, the nab4-7 mutant displays a striking resistance to high levels of copper (Figure 5A). Even in the presence of 1 M CuSO4, the nab4-7 mutant displays robust growth in comparison to its wild-type counterpart (Figure 5A, right panel). The nab4-7 mutant strain shows no apparent growth defect in normal growth conditions (trace copper) (Figure 5A, left panel). Therefore, we have identified a previously unknown role for Nab4 in copper homeostasis. With a striking growth phenotype in hand, we hypothesized that the resistance of the nab4-7 strain to excess copper may be due to a defect in alternative 3′ pre-mRNA processing. To identify alternatively processed transcripts that may play a role in the copper phenotype, we used microarrays to visualize genome-wide changes in transcript abundance in the nab4-7 strain. We reasoned that an alternate cleavage event could change the stability of the given transcript. Total RNA from NAB4 and nab4-7 strains at high-temperature and stationary phase growth were analyzed by standard yeast ORF microarray analysis (unpublished data). Hundreds of genes displayed significant changes in the nab4-7 background. Notably, the CTR2 transcript was severely affected, showing an increase in transcript levels of more than 8-fold. CTR2 encodes for a vacuolar membrane protein that helps to maintain proper intracellular levels of copper by controlling the flux of copper between the vacuole and the cytosol [29,30]. Given the surprising robustness of the nab4-7 strain to copper stress and given that the major transcript affected by Nab4 depletion was a gene encoding a copper transporter; we tested if the CTR2 transcript could be an alternatively processed substrate of Nab4. The CTR2 gene contains three Nab4 motifs at 188,426, and 581 nucleotides downstream of the stop codon. The first two motifs contain the TATATA motif and the farthest motif contains the TACATA motif. 3′ RACE analysis confirms the presence of three isoforms of the CTR2 transcript (Figure 5B). All three isoforms were confirmed to be CTR2 products using two additional PCR primers in independent reactions (unpublished data). We found that as functional Nab4 is depleted by a temperature shift of a strain containing the nab4-7 allele, the ratio between the forms changes, dramatically increasing the proportion of the longest form (Figure 5B). These results add CTR2 to the growing list of transcripts with multiple 3′ pre-mRNA cleavage sites. Moreover, these results demonstrate that the 3′ pre-mRNA processing of the CTR2 transcript is sensitive to the presence of functional Nab4. Since CTR2 was the most dramatically affected transcript of the copper regulon affected by the nab4-7 mutation, we hypothesized that it may be involved in the strong resistance of the nab4-7 strain to excess copper. To test this hypothesis, we created a strain containing either the NAB4 or the nab4-7 allele and a deletion of the CTR2 gene. Serial dilutions of these strains were grown on a series of copper plates (Figure 5C). In the wild-type strain background, deletion of ctr2 conferred no resistance or sensitivity to excess copper (Figure 5C, second panel from left). Deletion of ctr2 has no effect on the temperature-sensitive phenotype of the nab4-7 mutant (Figure 5C, left panel). However, the ctr2Δ completely suppresses the nab4-7 resistance to copper stress Figure 5C, right two panels). These data show that ctr2Δ is epistatic to nab4-7 with respect to copper stress. Therefore, CTR2 and possibly its alternative 3′ pre-mRNA processing play a central role in the copper resistance phenotype of the nab4-7 strain.
Alternative cleavage site selection may be more common and more important than is currently appreciated. Based on our microarray and bioinformatic analysis, we have identified a previously unknown role for Nab4 in copper homeostasis. In our previous work, we identified other functional categories of transcripts that are preferentially associated with Nab4 [27]. We anticipate that transcripts in these other functional categories will also display physiological consequences if their 3′ pre-mRNA processing is misregulated. In addition, estimates of alternative cleavage events were taken from EST data from a single phase of growth. Since alternative cleavage events can be sensitive to the growth condition of the cell, these estimates are likely to be low [10,11]. Between the three motifs identified here and the previously known canonical Nab4-binding element, approximately 59% of transcripts in the yeast genome contain more than one Nab4 motif in the region encompassing the ORF and 500 nucleotides downstream of the stop codon (unpublished data). It would not be surprising if the number of actual alternative cleavage events rivals the 54% currently predicted for the mammalian transcriptome [9]. Although we have concentrated our analysis on Nab4, we suspect that other players also contribute to regulated 3′ pre-mRNA processing. The work of Minvielle-Sebastia that originally motivated this work also implicated CF1a, another member of the cleavage complex, in establishing proper cleavage [24]. Additionally, it has been suggested that a third member of the cleavage complex, Rna14, affects the 3′-end processing of its own transcript [12]. It will be interesting to determine if other members of the cleavage complex can also influence cleavage site choice in vivo similar to Nab4. We have identified a correlation between alternative processing, Nab4, and the copper response. We have shown that a nab4-7 strain displays a striking resistance to excess copper and that this resistance requires the presence of the CTR2 gene. We demonstrated that the CTR2 transcript has multiple 3′ ends and that the cleavage site used is sensitive to the presence of functional Nab4. Interestingly, this exceptionally long 3′ UTR, almost 600 nucleotides, places the CTR2 transcript in the top 2% of predicted yeast 3′ UTR lengths. The average yeast transcript is estimated to have a 3′ UTR of less than 100 nucleotides [16]. Moreover, the quantity of CTR2 transcript in the cell rises dramatically as the nab4-7 mutant strain is shifted to the restrictive temperature. Exactly how the alternative processing affects the function of CTR2 and how this allows the nab4-7 mutant cells to resist excess copper remains unknown. Whether the alternative 3′ pre-mRNA processing of CTR2 is directly responsible for the copper phenotype in unknown; however, the correlation between Nab4, CTR2, and alternative 3′ pre-mRNA processing is too enticing to ignore. Interestingly, the 3′ UTR has recently been discovered to play an important role in another metal-stress responsive pathway. During iron-stress conditions in S. cerevisiae, the cell relies on mRNA surveillance mechanisms to respond to low levels of iron. In iron-replete conditions, specific mRNAs are targeted for degradation via sequence elements in the 3′ UTR of target transcripts [31]. Additionally, the absence of RNA surveillance pathways leads to sensitivity to high iron conditions in S. cerevisiae [32]. It remains to be seen if alternative 3′ pre-mRNA processing of CTR2 plays into a similar mRNA surveillance pathway, as seen in iron stress. Nonetheless, the importance of the 3′ UTR in different stress conditions emphasizes the potential for alternative 3′ pre-mRNA processing to effect cell physiology. Our study of alternative 3′ pre-mRNA processing illustrates the utility of studying this process in a genetically tractable organism. Given our unexpected discovery of a role for Nab4 in copper toxicity, we suspect that other growth and stress conditions will reveal even more regulated, alternative processing events. Moreover, identifying the mechanism and consequences of regulated 3′ pre-mRNA cleavage has proven time-consuming in metazoan cells. In S. cerevisiae, this type of analysis should be more facile than in metazoan cells. Just as the machinery of 3′ pre-mRNA processing between yeast and mammals is more closely related than first expected, we hypothesize that the regulation of alternative processing also remains conserved. In the simplest model of alternative 3′ pre-mRNA processing, Nab4 associates with its binding site and promotes cleavage at an adjacent site. Changing the association of Nab4 with a given site would then change the cleavage site to be used. Known instances of alternative 3′ pre-mRNA processing in mammalian cells, such as for CstF64 and the transcript for immunoglobulin M, are characterized by at least three defining features: first, multiple sequence elements differentially recruit the cleavage complex; second, an RNA-binding protein recognizes a specific sequence element; and lastly, the activity or concentration of the RNA-binding protein is regulated. We have shown that Nab4 fulfills each of these requirements. Similar to CstF64, Nab4 is a component of the processing machinery. Just as control of the levels of CstF64 is a fundamental component to regulating cleavage site selection in B cells, we have found that the levels of Nab4 can likewise affect cleavage site selection in vivo. Moreover, the association of CstF64 with its binding element leads to activation of a nearby cleavage site. Similarly, we find that deletion of predicted Nab4-binding motifs reduces cleavage efficiency at adjacent sites. These data are not consistent with an alternate model proposed for Nab4 function which suggests that Nab4 association blocks inappropriate cleavage sites [24]. Therefore, just as the protein and sequence elements have been found to be conserved from yeast to mammals, we propose that the regulation of alternative cleavage site selection is similarly conserved. A key feature of regulated 3′ pre-mRNA processing in mammalian cells is the ability of CstF64 to discriminate between strong and weak polyadenylation signals. If the mechanism of regulated cleavage is indeed conserved, then we expect Nab4 to also show such a bias between sites. In the simplest case, such a discrimination of the protein would be based on its binding affinity to the efficiency element. We identified three motifs that are enriched in messages preferentially associated with Nab4, one that contains the canonical efficiency element and two that are single nucleotide deviations from the canonical. For the SUA7 transcript, the deviant motif lies upstream of the canonical motif and the upstream site is the less preferred site during log phase growth. We predict that the deviant motifs will display a lower affinity for Nab4 binding than the canonical motifs. Additionally, we have identified just three motifs via our microarray and bioinformatics analysis. Bioinformatics analysis has shown that other single-nucleotide deviations away from the core UA repeat motif are also statistically enriched downstream of ORF sequences [15,16,33]. While the affinity of Nab4 for the canonical efficiency element has been estimated, the specificity of Nab4 remains uncharacterized. It would be interesting to determine if the affinity of Nab4 for these different sequence elements correlates with the ability to undergo alternative 3′ pre-mRNA processing. Based on our findings that the overexpression of Nab4 alters cleavage site selection, we predict that the cell should exert some regulation over Nab4 itself. Moreover, if Nab4 is regulated, then cleavage site selection may also be regulated. Nab4 is already known to be subjected to both methylation and phosphorylation [34,35]. Both modifications have the potential to alter Nab4 function. Another indication of Nab4 regulation is that both too little (NAB4 is essential) and too much Nab4 lead to cell mortality (unpublished data; M. Swanson, personal communication). In fact, several lines of evidence imply that the levels of Nab4 are continuously adjusted based upon growth conditions. First, NAB4 transcript levels are sensitive to several growth conditions as shown by microarray analysis. Interestingly, NAB4 transcript levels decrease during stationary phase [36], increase during the early phase of sporulation [37], and fluctuate during metabolic cycling in nutrient-limited conditions [38]. In addition to regulating the total concentration of Nab4, the local concentration of Nab4 could be affected. Since cleavage site selection is presumably a nuclear event, changes in the nuclear concentration of Nab4 should affect cleavage site choice as well. Interestingly, certain stress conditions, such as hypo-osmotic stress, can affect the localization of Nab4 to move from mainly nucleoplasmic to largely cytoplasmic [39]. We predict that conditions which change the total or local Nab4 protein levels or the activity of Nab4 will correlate with changes to cleavage site choice in transcripts important in these conditions. In this paper, we address two critical issues concerning alternative 3′ pre-mRNA processing: how is alternative cleavage site choice mediated in vivo and what are the physiological consequences of alternative cleavage? We propose a mechanism of regulation via titration of a component of the cleavage machinery that is similar to the known mechanism utilized in mammalian cells. More importantly, this work describes a relationship between alternative cleavage and a physiological response in S. cerevisiae. The unexpected discovery of a robust growth phenotype of the nab4-7 strain to excess copper demonstrates the utility of our approach of identifying preferentially associated binding partners of RNA-binding proteins and using bioinformatics to uncover clues as to the function of the protein. We predict that copper stress is the first of many physiological responses that depend, in part, upon alternative cleavage.
Yeast manipulations were executed according to Guthrie and Fink [41]. The nab4 mutant strains were generously provided by the Swanson Laboratory and were first described by Minvielle-Sebastia et al. [24]. Each strain has the endogenous NAB4 gene deletion covered by CEN plasmid carrying either a wild-type or mutant version of NAB4. The Swanson Laboratory also generously provided the Nab4 overexpression strain. This strain is a diploid, heterozygous at the NAB4 locus (one wild-type and one deletion), containing a plasmid with NAB4 under the control of the Gal4, galactose-inducible promoter. The Δctr2 and nab4-7, Δctr2 double mutant strain were constructed in the same strain background as above according to the method in Longtine et al. [40]. Total RNA was isolated using four organic extractions: once in hot, acidic phenol-choloform; twice in cold, acidic phenol-chloroform; and a final extraction in chloroform-isoamyl alcohol. The RNA was then precipitated with sodium acetate and ethanol and resuspended in water. Equal amounts of RNA by OD were diluted into loading buffer and loaded onto an agarose gel between 1. 5%–2% agarose in TBE buffer. RNA was transferred onto a membrane using capillary action for at least 6 h or overnight. Both the RNA on the gel and on the membrane was visualized to ensure successful transfer. The RNA was immobilized on the membrane with UV-crosslinking. Radioactive probes for the Northern analysis were 32P end-labeled oligos. Rapid-Hybe buffer (Amersham, http: //www. amersham. com) was used during the hybridization. Quantitation was performed using a Storm Phosphorimager. Total RNA was isolated using the same procedure described from Northern analysis. 20–25 μg of total RNA was reverse transcribed using Superscript III (Invitrogen, http: //www. invitrogen. com) and an anchored dT primer. The cDNA was diluted 10-fold and then PCR-amplified using one primer specific to the ORF and one primer specific to the dT oligo. The primer for the SUA7 analysis was “ataacttaccgggcgttg” and the primer for CTR2 analysis was “tggggcaatatggggtaattaca. ” The 3′ RACE reactions were separated on 1. 5% agarose gels. Strains were grown to log phase, then diluted to the same density by OD. 3- to 4-fold serial dilutions were prepared and plated onto a series of plates containing various concentrations of CuSO4 or rich media (YPD). Except where noted, strains were grown at room temperature for 3–5 d. | A fundamental step in gene expression is the generation of the terminal edge (3′ end) of the mRNA transcript by appropriate cleavage of the longer pre-mRNA. In general, the processing site that emerges first is used, but there are interesting examples where alternative sites are used. Because the choice of alternative sites can add or delete sequences which can affect transcript stability, localization, or translation, it is important to understand how this process is regulated. We have addressed this question in the genetically tractable yeast Saccharomyces cerevisiae. Nab4 is a sequence-specific RNA-binding protein involved in 3′ processing. We showed that cells that make too little or too much Nab4 exhibit significant changes in the ratios of alternative 3′ ends. Interestingly, we found that a nab4 mutant is able to grow on concentrations of copper that are toxic to normal cells. We identified a gene required for this capability, CTR2, and showed it has alternative 3′ ends sensitive to the presence of Nab4. We predict that by influencing levels of other alternative 3′ ends, Nab4 is also important for the appropriate biological responses to many other stresses. | Abstract
Introduction
Results
Discussion
Materials and Methods | cell biology
eukaryotes
molecular biology
genetics and genomics
saccharomyces | 2007 | Alternative 3′ Pre-mRNA Processing in Saccharomyces cerevisiae Is Modulated by Nab4/Hrp1 In Vivo | 7,878 | 259 |
The peroxide response transcriptional regulator, PerR, is thought to contribute to virulence of group A Streptococcus (GAS); however, the specific mechanism through which it enhances adaptation for survival in the human host remains unknown. Here, we identify a critical role of PerR-regulated gene expression in GAS phagocytosis resistance and in virulence during pharyngeal infection. Deletion of perR in M-type 3 strain 003Sm was associated with reduced resistance to phagocytic killing in human blood and by murine macrophages in vitro. The increased phagocytic killing of the perR mutant was abrogated in the presence of the general oxidative burst inhibitor diphenyleneiodonium chloride (DPI), a result that suggests PerR-dependent gene expression counteracts the phagocyte oxidative burst. Moreover, an isogenic perR mutant was severely attenuated in a baboon model of GAS pharyngitis. In competitive infection experiments, the perR mutant was cleared from two animals at 24 h and from four of five animals by day 14, in sharp contrast to wild-type bacteria that persisted in the same five animals for 28 to 42 d. GAS genomic microarrays were used to compare wild-type and perR mutant transcriptomes in order to characterize the PerR regulon of GAS. These studies identified 42 PerR-dependent loci, the majority of which had not been previously recognized. Surprisingly, a large proportion of these loci are involved in sugar utilization and transport, in addition to oxidative stress adaptive responses and virulence. This finding suggests a novel role for PerR in mediating sugar uptake and utilization that, together with phagocytic killing resistance, may contribute to GAS fitness in the infected host. We conclude that PerR controls expression of a diverse regulon that enhances GAS resistance to phagocytic killing and allows adaptation for survival in the pharynx.
Group A Streptococcus (Streptococcus pyogenes or GAS) is a strictly human pathogen that causes a spectrum of disease ranging from superficial infection of the pharyngeal mucosa and the skin to invasive infection of deep tissues and the bloodstream. Like other lactic acid bacteria, GAS lacks oxidative phosphorylation machinery and instead obtains energy by fermentation. Despite this lack of oxygen requirement, GAS has evolved to grow rapidly in the oxygen-rich human host environment and to resist killing by reactive oxygen species (ROS), such as hydrogen peroxide (H2O2) and superoxide (O2−), which are formed by reduction of atmospheric oxygen or produced by phagocytes upon activation of the host inflammatory response. Unlike some other Gram-positive pathogenic bacteria, GAS lacks catalase (a heme-containing peroxidase) that degrades H2O2 to H2O and O2, but produces several alternative peroxidases that degrade organic and inorganic peroxides formed in the presence of O2 or as a consequence of inflammation. Among these, the NADH peroxidase (npr), alkyl hydroperoxidase (ahpC) and glutathione peroxidase (gpoA) have been shown to contribute to GAS aerotolerance and ROS detoxification during culture in vitro as well as during infection in mice [1]–[4]. Resistance to O2− toxicity is mediated by superoxide dismutase (sodA), which converts O2− to H2O2, which is detoxified in turn by peroxidases. Superoxide dismutase is also critical for GAS aerotolerance and survival in oxidative environments, as a sodA mutant was not only more sensitive to O2−-generating agents but also was unable to grow under standard aerobic growth conditions [5]. GAS adaptive responses to oxidative stress are coordinated, at least in part, by the peroxide stress response regulator PerR, which belongs to the Fur (ferric uptake repressor) super-family of metal-binding transcriptional regulators [2], [6]. PerR was originally shown to control oxidative stress responses in Bacillus subtilis through direct binding of conserved promoter sequences known as Per boxes [7]. In the absence of oxidative conditions, PerR represses PerR regulon expression in B. subtilis by remaining bound to Per boxes in PerR-regulated promoters. In this form, PerR contains a structural zinc atom, as well as a regulatory ferrous ion, binding of which is coordinated by several amino acids including three histidines. Under oxidative conditions, the bound ferrous ion catalyses oxidation of two of three histidine residues, promoting release of oxidized PerR from target promoters and de-repression of the PerR regulon [8], [9]. In B. subtilis and Staphylococcus aureus, PerR controls expression of peroxidases (catalase, alkylhydroperoxide reductase) among other genes, and coordinates oxidative stress responses and iron homeostasis [7], [10], [11]. Such coordinate control is critical for bacterial survival as free intracellular iron reacts with H2O2 to form highly oxidizing hydroxyl radicals (HO.) in what is known as the Fenton reaction. Certain studies have suggested that expression of known peroxidases and all three described iron acquisition systems of GAS [12]–[14] is PerR-independent, and therefore, PerR regulation in GAS differs from other Gram-positive species [2], [4], [15]. However, decreased expression of superoxide dismutase and the MtsABC iron transport system has been reported in an M-type 1 GAS perR mutant [6]. More recent evidence suggests that PerR may be involved in coordinating oxidative stress responses and metal homeostasis in GAS. Characterization of the PerR regulon in M-type 5 GAS, by comparing a perR mutant transcriptome to that of wild-type during mid-exponential phase growth, suggested strong regulation (3-fold or higher) of six genes by PerR; however, only pmtA, a gene encoding a putative metal efflux protein, was proposed to be under direct PerR control [16]. Interestingly, the previously reported enhanced resistance of perR mutants to H2O2 killing in vitro was attributed to pmtA overexpression, as mutation of pmtA in the perR mutant background led to lower H2O2 resistance, and single pmtA mutants were over 10-fold more sensitive to H2O2 killing than wild-type GAS. Deregulation of the remaining five genes in the M-type 5 perR mutant was proposed to be an indirect effect of metal starvation resulting from pmtA upregulation, and was thought to be mediated by a second transcriptional regulator protein, AdcR, although regulation of adcR itself by PerR was not shown [16]. Even though PmtA was associated with resistance to H2O2 killing in vitro, no PerR-regulated gene identified in that strain, including pmtA, was linked to GAS survival and virulence in vivo. Despite their increased resistance to H2O2 challenge, M-type 1 and M-type 5 perR mutants of GAS show sensitivity to superoxide in vitro and significant loss of virulence following subcutaneous or intraperitoneal inoculation in mice [2], [4], [6]. However, the mechanism through which PerR contributes to GAS virulence and the specific contribution of PerR-regulated gene expression to bacterial survival at particular host sites during infection remains unknown. In this study, we demonstrate a critical role of PerR-regulated gene expression in GAS resistance to phagocytic killing and in pharyngeal colonization in primates. In addition, using genomic microarrays, we find that the GAS PerR regulon is substantially more extensive and more diverse than previously appreciated, a finding that might explain the pivotal role of PerR in GAS phagocytosis resistance and pharyngeal infection.
To study the role of perR in GAS virulence and pathogenesis, a perR deletion mutant was derived from strain 003Sm, a spontaneous streptomycin resistant (SmR) variant of wild-type M-type 3 strain DLS003 [17]. Mutagenesis was designed based on the genome sequence of M-type 3 strain MGAS315 [18]. An in-frame 426 bp deletion of the 468 bp coding sequence of perR (spyM3_0147) was constructed in the temperature-sensitive shuttle vector pJL1055, and the resulting plasmid pJLperRΔ was used to generate the perR deletion mutant strain 003SmperRD by allelic exchange (Figure 1). The mutant exhibited wild-type growth rates under aerobic conditions during culture in THY broth or on THY-blood agar. As reported for perR mutants constructed in M-type 1 and M-type 5 GAS [2], [6], resistance of 003SmperRDΔ to H2O2 challenge (10 mM) in liquid culture was approximately 15-fold higher than that of parent strain 003Sm, with survival rates of 23% versus 1. 6%, respectively (Figure S1). The involvement of PerR in oxidative stress adaptive responses of GAS in vivo was evaluated by testing phagocytic killing resistance of the perR mutant strain 003SmperRΔ in the Lancefield bactericidal test [19]. Wild-type strain 003Sm and perR mutant 003SmperRΔ were rotated end-over-end in heparinized human blood at 37°C, and survival was assessed over time. As indicated by the viable counts recovered 3 h post-inoculation, wild-type bacteria showed a net growth of 4-fold relative to the inoculum in sharp contrast to the perR mutant that did not exhibit any net growth over the same time period (Table 1). In control assays in which bacteria were incubated for 3 h without rotation (required for physical contact between leukocytes and bacteria for efficient phagocytosis), both mutant and wild-type GAS increased by 30 to 50-fold. Thus, the reduced proliferation of the mutant relative to wild-type bacteria in rotating blood was due to its increased sensitivity to killing by phagocytes. To further validate this result, we assessed phagocytic killing resistance of the perR mutant strain expressing PerR constitutively from plasmid pORI-perR in trans, or the perR mutant strain carrying empty vector pORI23 as control. For reasons that are unclear, culture of GAS on erythromycin (for maintenance of pORI-perR and pORI23) increased baseline survival in human blood more than 10-fold. Even so, survival of strain 003SmperRΔ (pORI-perR) was significantly higher (1. 8-fold) than that of control strain 003SmperRΔ (pORI23), a finding that also supported the role of PerR in GAS phagocytic killing resistance (P<0. 01 for comparison of survival at 3 h of perR mutant carrying pORI-perR versus empty vector pORI23). Taken together, these data provide the first direct evidence that PerR-regulated gene expression is required for optimal GAS resistance to phagocytic killing during infection. The mechanism by which PerR contributes to GAS survival inside phagocytic cells was further investigated using a mouse macrophage cell line. Wild-type strain 003Sm or perR mutant 003SmperRΔ was used to infect RAW264. 7 macrophages and the ability of the mutant to survive intracellularly was compared to that of wild-type bacteria. Intracellular survival was measured in the absence or presence of diphenyleneiodonium chloride (DPI), an inhibitor of both NADPH oxidase and nitric oxide synthase, in order to investigate the effect of the macrophage oxidative burst on both wild-type and mutant GAS. In the absence of DPI, survival of the mutant inside macrophages was 3- to 4-fold lower than that of wild-type, as indicated by the number of intracellular bacteria recovered from infected macrophages during gentamicin/penicillin exclusion assays (Figure 2). This survival attenuation of the mutant inside macrophages was in agreement with reduced resistance to phagocytic killing in human blood shown above. DPI treatment of infected macrophages resulted in an approximate 10-fold increase in the number of viable intracellular wild-type bacteria. Whereas the number of intracellular perR mutant bacteria recovered from untreated macrophages was significantly lower than that of wild-type, in the presence of DPI intracellular survival of the mutant was similar to wild-type GAS (Figure 2). The finding that the perR mutant is hypersensitive to macrophage killing and that its intracellular survival defect is abrogated by DPI suggests that GAS killing by the macrophage oxidative burst is counteracted, at least partially, by PerR-dependent gene expression. A major goal of this study was to define the contribution of PerR in GAS survival and persistence in the pharynx, the most commonly infected human host site. To achieve this goal, we used a baboon pharyngeal colonization model that recapitulates several aspects of GAS human infection [20]. Quantitative throat cultures were used to assess pharyngeal colonization by wild-type strain 003Sm compared with that of the perR mutant integrant strain 003Sm-Int-perRΔ in five baboons co-infected with approximately 1. 5×108 cfu of each of the two strains. Co-infection with parent strain 003Sm served as an internal reference in each animal and was adopted to control for the variability in GAS pharyngeal colonization levels previously observed among different animals. Expression of perR in strain 003Sm-Int-perRΔ is abolished by the stable chromosomal integration of plasmid pJLperRΔ (Figure 1) that also confers chloramphenicol (Cm) resistance, a marker that was exploited to differentiate mutant strain 003Sm-Int-perRΔ from the Cm-sensitive parent strain. In addition, to control for possible nonspecific attenuating effects of plasmid insertion in the GAS chromosome or potential excision of the plasmid from the chromosome during infection, a second group of five animals was co-infected with similar doses of 003Sm and integrant strain 003Sm-Int-WT in which pJLperRΔ integration had occurred immediately downstream, and not upstream, of perR (Figure 1). As predicted by their respective genotypes, strain 003Sm-Int-perRΔ did not express PerR and was hyper-resistant to H2O2 challenge similarly to the deletion mutant strain 003SmperRΔ while strain 003Sm-Int-WT produced wild-type amounts of PerR, as shown in immunoblots using PerR-specific antiserum (Figure 3), and exhibited similar H2O2-sensitivity as the wild-type strain (data not shown). All 10 animals were infected by direct pipeting of pre-mixed suspensions of one of the two strain pairs onto the posterior pharynx (003Sm/003Sm-Int-perRΔ or 003Sm/003Sm-Int-WT). A total of 11 sequential throat swabs were collected from each animal at various time points up to 42 d post-inoculation for quantitative culture on both Sm and Cm plates. Relative colonization of wild-type versus each of the two integrant strains was calculated by dividing the wild-type cfu count by the cfu count of the plasmid integrant strain studied, and was represented as the competitive colonization index (CCI) at each time point for each animal. All baboons were colonized for 28 d or longer except for one animal in the control group (003Sm/003Sm-Int-WT) that was colonized for 14 d (Figure S2). At 1 and 3 h post-infection, the GAS counts in both groups were reduced drastically compared to the inoculum; however, the counts of each of the two integrant strains were similar to those of the wild-type parent strain, as indicated by CCIs of approximately 1 in nine of 10 animals (Figure 4). A further decrease in colonization occurred in most animals at 24 to 48 h (Figure S2), a pattern similar to that observed in previous primate studies of GAS throat colonization [20]–[22]. Interestingly, the 003Sm/003Sm-Int-perRΔ median CCI increased from approximately 1 at 3 h to 4. 8 at 48 h post-infection, whereas the median CCI in the control group (003Sm/003Sm-Int-WT) was approximately 11-fold lower at 0. 43. Following the initial drop, colonization levels increased in most animals of both groups after day 3. In the control group, the two strains were recovered in similar relative numbers throughout the experiment (Figure S2), as indicated by median CCIs of approximately 1 (Figure 4). In contrast, the median CCI in the 003Sm/003Sm-Int-perRΔ infected group increased further after day 3 reaching values of 53,21, and 32 on days 7,14, and 21, respectively, which were 33- to 81-fold higher than those recorded for the control group at respective time points (Figure 4). The competitive infection experiments were analyzed also as time-to-clearance curves, that is, the time to persistently negative cultures for each of the three strains used. This analysis showed that the perR mutant strain 003Sm-Int-perRΔ was cleared by two of five baboons within hours of infection and was not recovered from four of five animals after day 7 (Figure 5). In sharp contrast, the wild-type strain colonized all five animals in the same group for at least 28 d, with three of five animals staying GAS-positive for the duration of the study (42 d). As expected, in the control animals the time-to-clearance of the control integrant strain 003Sm-Int-WT did not differ from that of wild-type strain 003Sm. Together, these data demonstrate that the perR mutant strain 003Sm-Int-perRΔ is severely attenuated in its capacity to colonize the baboon pharynx. The rapid elimination of the mutant from the baboon throat underlines the importance of PerR in virulence and suggests a critical role of PerR-regulated gene expression in GAS survival and persistence in the pharynx during human infection. To gain insight into the gene network controlled by PerR in strain 003Sm and to identify candidate loci contributing to GAS pharyngeal infection and virulence, we compared the transcriptome of mutant strain 003SmperRΔ with that of 003Sm using GAS genomic microarrays. Total RNA was isolated from both strains grown to mid- or late-exponential phase and mRNA was converted to fluorescently labeled cDNA and hybridized to the GAS microarray. A total of 42 genes, including five encoding hypothetical proteins, were differentially regulated in wild-type bacteria: 20 of these were regulated at mid-exponential phase and 31 at late-exponential phase growth (Figure 6). Five genes (SPy0714, SPy1434, SPy1871, SPy2000, spyM3_1095) showed similar regulation in both growth phases and were among the most highly downregulated loci. Unexpectedly, 21 of the 42 PerR-dependent genes exhibited higher expression levels in wild-type bacteria, a finding that suggests PerR may act, directly or indirectly, as an activator, as well as a repressor. Sixteen of these 21 genes were identified during late-exponential phase growth, an indication that PerR-dependent gene activation is growth phase-dependent and may involve specific physiological conditions and/or additional factors absent during early exponential phase (Figure 6). The microarray data were validated by quantitative RT-PCR (qRT-PCR) using total RNA isolated from three independent cultures of wild-type and perR mutant bacteria in each of two growth phases. Based on the data obtained for a total of 17 genes representing the majority of PerR-dependent loci (Table 2), there was strong correlation between qRT-PCR and microarray data analyses (r = 0. 89). PerR regulation seemed to be somewhat variable for three of these genes (SPy0033, SPy1159, SPy1707), as their regulation, though confirmed in the RNA preparations used for microarray studies (r = 0. 98), could not be validated in independent RNA preparations. Several PerR-dependent genes could be associated with GAS oxidative stress adaptation responses and virulence (Table S1). Among these, pmtA (SPy1434), the most highly PerR-repressed gene identified, encodes a metal transport ATPase recently shown to be under PerR control in M-type 5 GAS. In that study, PmtA was suggested to contribute to H2O2 resistance in vitro but its role in vivo remains unclear [16]. Also, among these genes was SPy2079-SPy2080 encoding alkyl hydroperoxidase and NADH oxidase/alkyl hydroperoxide reductase (AhpC and AhpF, respectively) (Figure 7). AhpCF is thought to catalyze the NADH-dependent reduction of peroxides, and has been shown to contribute to GAS oxidative stress resistance and virulence in mice [1], [2], [4]. Although regulation of this locus was not strong, at least under the conditions tested, AhpCF is the first GAS peroxidase system found to be PerR-dependent. In addition, expression of the ribonucleotide reductase operon nrdF. 2-nrdI. 2-nrdE. 2 (SPy0425-SPy0427) was PerR-dependent (Figure 7). Ribonucleotide reductases are required for the conversion of ribonucleotides to deoxyribonucleotides and may be critical for DNA replication under oxidative stress for repair of DNA damage by ROS [23]–[25]. The M-type 3-specific prophage gene spyM3_1095, encoding DNase/mitogenic factor 4 (Mf4) was also included in this group. Expression of this extracellular DNase was shown to be activated under oxidative stress, as well as following DNA damage in M-type 3 strain MGAS315 [26]. Finally, SPy0714 encoding the putative adhesin AdcA was strongly regulated at both growth phases studied. It was striking that 40% (17 of 42) of PerR-dependent loci are predicted to encode products involved in sugar metabolism and/or transport, with the majority being upregulated by PerR (directly or indirectly) during late-exponential phase growth. These loci included two phosphotransferase systems (PTSs), SPy1710-SPy1711 and SPy1916-SPy1923 (lacA. 2-lacG), with putative function in lactose/galactose utilization [27], [28]. Moreover, a PTS transcriptional regulator that exhibited similar PerR-dependent up-regulation mapped immediately upstream and in the opposite orientation of each of these two loci (SPy1712, SPy1924), an indication that PerR regulation of these two PTS loci might be indirect through control of their respective cognate regulators (Figure 7). These findings suggest an important role of PerR in sugar uptake and utilization, as well as oxidative stress adaptive responses of GAS. In summary, 35 of a total of 42 genes shown to be PerR-dependent in strain 003Sm were not previously considered to be under PerR control. These included genes mainly regulated during late-exponential phase growth, such as most PTS/sugar metabolism genes, as well as genes with putative function in oxidative stress responses and virulence. These findings suggest that PerR controls, directly or indirectly, a more extensive and functionally more diverse regulon than previously reported in GAS. In order to investigate whether the PerR-dependent genes identified in microarray studies were directly regulated by PerR in strain 003Sm, the M-type 3 strain MGAS315 genome [18] was scanned for the previously proposed GAS Per box operator sequence [4]. This analysis revealed canonical Per boxes upstream of pmtA and ahpCF, which is in agreement with earlier reports of PerR-regulated gene expression in GAS [4], [16]. Additional Per box-like promoter sequences were identified upstream of several PerR-dependent loci in strain 003Sm, such as nrdF. 2-nrdI. 2-nrdE. 2, PTSs SPy1710-SPy1711 and SPy1916-SPy1923 (lacA. 2-lacG) as well as their cognate regulators lacR. 1 and lacR. 2 (Table 3). A non-canonical Per box sequence has been previously reported for the DNA-binding peroxide resistance gene mrgA also thought to be PerR-regulated in GAS, although PerR binding to the mrgA promoter has not been shown [4]. Since several of the differentially regulated loci in the M-type 5 perR mutant had been previously suggested to be under AdcR control [16], with their differential regulation a secondary effect of pmtA upregulation in the mutant, a second M-type 3 genome search was performed for AdcR consensus sequences [29]. AdcR motifs were identified upstream of four loci differentially regulated in perR mutant strain 003SmperRD: adcRCB (SPy0092-SPy0094), adcA (SPy0714), rpsN. 2 (SPy1871) and phtD (SPy2006). Our results of AdcR motif search in M-type 3 GAS are in line with those reported by Brenot et al. [16] and imply that regulation of the latter loci is also under AdcR control in strain 003Sm. Following in silico analyses, PerR binding to promoters encompassing predicted Per boxes was tested by electrophoretic mobility shift assays (EMSAs). Recombinant PerR was expressed in E. coli as an N-terminal his-tagged protein and was purified under native conditions. Promoters tested in EMSAs included those of the putative metal transport ATPase pmtA and the alkyl hydroperoxidase system ahpCF, both of which include canonical Per boxes, the promoter upstream of the PTS SPy1916-1923 (lacA. 2-lacG) as well as its divergently transcribed putative cognate regulator lacR. 2 (SPy1924), and the promoter sequence in the intergenic region upstream of PTS SPy1710-1711 and its putative cognate regulator gene lacR. 1 (SPy1712) that contains a common Per box (Table 3). As shown in Figure 8, recombinant PerR bound to both pmtA and ahpC promoters but did not bind to promoters encompassing non-canonical Per boxes, such as those upstream of the PTS loci (data not shown). It is possible that PerR binding to the latter promoters requires different conditions from those tested and/or additional factors that may be present in GAS in vivo. Alternatively, these genes may not be directly regulated by PerR, with their altered expression in mutant strain 003SmperRΔ being a secondary effect of PerR deletion.
The present investigation provides direct new evidence that PerR plays a key role in GAS adaptation for survival in the human host and uncovers a mechanism through which it contributes to GAS virulence. PerR-dependent gene expression was associated with enhanced resistance to ROS-dependent phagocytic killing by macrophages in vitro and with improved survival in human blood. The PerR system also enhanced GAS pharyngeal colonization in a primate model that mimics GAS pharyngitis in human beings [20]. In this model, wild-type GAS showed a clear competitive advantage over a perR mutant in both its ability to maintain high colonization levels and to persist in the pharynx over time. The marked attenuation of the perR mutant is similar to that reported in the baboon model for GAS mutants lacking the hyaluronic acid capsule or M protein, two of the most important virulence determinants of the species [20]. To our knowledge, this is the first study to demonstrate the key role of PerR-dependent gene expression in GAS pharyngeal infection and to directly associate PerR with phagocytic killing resistance in human blood. Our comparative analysis of global gene expression in M-type 3 strain 003Sm versus its isogenic perR mutant revealed a PerR-dependent regulon of 42 genes. Although some of these had been reported earlier [16], the majority of the PerR-dependent loci in strain 003Sm were not previously identified. Discovery of new PerR-dependent genes in the present study may reflect the fact that many were identified in late-exponential phase samples, whereas earlier investigations compared wild-type to perR mutant transcriptomes during mid-exponential phase. It is likely that some of the differences reported also reflect strain-dependent variation among GAS isolates. For instance, expression of both superoxide dismutase (sodA; SPy1406) and the metal transport system MtsABC (SPy0453-SPy0455) was moderately affected in an M-type 1 GAS perR mutant [6], but no such evidence was found in this or other studies involving GAS strains of M-type 1,3, and 5 [15], [16]. In addition, the Dps-like peroxide resistance gene mrgA (SPy1531), the putative cold shock protein gene csp (SPy2077), and the cation transport gene czcD (SPy0845) previously shown to be under PerR control in M-type 5 GAS were not identified in the present study [4], [16]. Several of the PerR-regulated gene products have been implicated in GAS oxidative stress responses and resistance mechanisms. The metal transporter PmtA has been linked to H2O2 resistance in vitro but its role in virulence remains unclear, as low pmtA expression was observed in GAS recovered from mouse skin ulcers [16]. Nonetheless, the strong regulation of pmtA by PerR in both the M-type 5 and M-type 3 strain backgrounds argues for an important role of this metal transporter in GAS physiology and perhaps virulence. Increased metal efflux due to pmtA upregulation has been suggested to lead to H2O2 hyper-resistance of the perR mutant during growth in metal-replete conditions in vitro [16]. In the metal-deplete host environment, however, PmtA upregulation could lead to metal starvation that might be in part the reason for reduced fitness and virulence of the perR mutant during infection. Thus, coordinate control of PmtA expression by PerR might vary depending on the host environment encountered. The role of the alkyl hydroperoxidase system (AhpCF) in oxidative stress resistance and virulence of GAS has been demonstrated [1], [2], [4]. Although the degree of ahpCF regulation by PerR was only moderate (1. 7- to 2. 0-fold) under the experimental conditions of this study, such regulation may still have a significant effect on oxidative stress resistance of GAS. Initial studies suggest 2- to 3-fold upregulation of ahpC following GAS challenge with H2O2 during culture in vitro; however, additional microarray studies with peroxide and perhaps other stress agents (oxidative and non-) will be essential to identify the physiological environmental stimulus for PerR and the resulting PerR-mediated regulation events in GAS. PerR-dependent expression was also demonstrated for the M-type 3 specific phage-encoded DNase Mf4 (spyM3_1095) that was previously shown to be induced under oxidative stress and DNA damage [26]. DNases (phage-encoded and non-) have been shown to enhance extracellular phagocytic killing resistance of GAS by degrading human neutrophil extracellular traps [30], [31]. A similar function in phagocytic killing resistance of M-type 3 GAS is possible for Mf4. Interestingly, expression of the ribonucleotide reductase operon nrdF. 2-nrdI. 2-nrdE. 2 was also PerR-dependent. Ribonucleotide reductase systems catalyze deoxyribonucleotide synthesis which is essential for DNA replication and repair. Such fundamental function is critical for bacterial survival and may be required for DNA replication under all growth conditions. Upregulation of this operon could also allow repair of DNA damage due to ROS under oxidative conditions. A second ribonucleotide reductase system encoded by spyM3_1048-spyM3_1050 and spyM3_1706, representing NrdH-NrdE. 1-NrdF. 1 and NrdI. 1, respectively, is found in both M-type 1 and 3 GAS genomes [18], [32]. While the nrdIEF locus seems to be essential for growth of B. subtilis [23], [25], an absolute requirement of one or both ribonucleotide reductase systems in GAS has not been established. Approximately half of the PerR-dependent genes in strain 003Sm are activated by PerR (directly or indirectly), and a significant proportion of these are predicted to encode products involved in sugar metabolism and transport. This result suggests a novel role for PerR in regulating sugar uptake and utilization by GAS. PerR activation (direct or indirect) of PTSs and/or other sugar utilization pathways may reflect the increased energy utilization required to counteract cell toxicity inflicted upon the bacteria by the inflammatory response. Utilization of multiple sugar sources during infection is likely to increase bacterial survival and could be an important contribution of PerR-regulated gene expression to GAS virulence. The dramatically attenuated colonization capacity of the perR mutant in the baboon model could be the result of poor uptake and/or utilization of available sugar sources in the oropharynx together with reduced resistance to phagocytic killing. Recent evidence shows that GAS utilization of maltodextrins is specifically upregulated in human saliva, an event that has been proposed to contribute to GAS survival in saliva and to efficient colonization of the mouse oropharynx [33]. Currently, whether any of the PerR-dependent PTSs contributes to GAS fitness in the host remains unknown. In B. subtilis, PerR represses target gene expression by binding to conserved Per box promoter sequences. Bound PerR is released from promoters under oxidative conditions upon its auto-oxidation by a regulatory ferrous ion [9]. The PerR regulation mechanism in GAS may be similar, as purified recombinant PerR contains both zinc and ferrous ions (I. Gryllos and D. Kurtz Jr; unpublished observations). In addition, a Per box consensus sequence similar to that of Bacillus has been proposed for GAS; however, except for pmtA and ahpCF, this sequence is not fully conserved in promoters thought to be controlled by PerR [4], [16]. PerR binding to a canonical Per box upstream of ahpCF was reported previously in GAS, but PerR-dependent regulation of this locus was very weak, if any, at mid-exponential phase growth [4]. In the present study, moderate regulation of ahpCF was reported during late-exponential phase growth with weaker regulation observed during mid-exponential phase growth. Binding of recombinant PerR to both pmtA and ahpCF promoters in electrophoretic mobility shift assays confirms direct PerR regulation of the ahpCF locus and also demonstrates direct regulation of pmtA. In addition, Per box-like sequences were identified upstream of several other PerR-dependent loci in strain 003Sm, of which some were predicted to be activated, and not represssed, by PerR. Unlike the promoters encompassing fully conserved Per boxes, the promoters of the PerR-activated PTS loci did not bind recombinant PerR in vitro. Interestingly, the majority of gene activation events were observed in late-exponential phase growth, which suggests that specific conditions and/or additional factors may be required for PerR regulation at these promoter sites. At present, whether additional factors affect PerR regulation of certain promoters by interacting with the promoter sequence, with PerR, or both, remains unclear. Nonetheless, the presence of Per box-like sequences upstream of loci that require PerR for activation implies a potential PerR interaction with Per box sequences during gene activation as well as repression. Ferrous ion-mediated PerR oxidation resulting in a conformational change could be part of an activation mechanism. PerR is also thought to be required for maximal expression of several genes in B. subtilis, and direct PerR gene activation through binding to two tandemly arranged non-canonical Per boxes has been reported for the surfactin (srfA) operon [11], [34]. Initially considered to be exclusively a repressor, the prototype of the PerR family, Fur, has been suggested to also activate gene expression [35], [36]. Currently, the possibility that PerR does not activate gene expression in GAS through direct interaction with specific promoters, but instead perR deletion leads to indirect effects in sugar utilization and other loci cannot be excluded. Taken together, our data show a pivotal role of PerR-controlled gene expression in GAS resistance to phagocytic killing and a critical contribution to host adaptation for survival in the pharynx. In addition, the evidence presented expands significantly on our knowledge of the PerR regulon in GAS. The diverse functions of newly identified PerR-dependent loci suggest a previously unrecognized role of PerR that may contribute to increased bacterial fitness in the host. Future investigations into the contribution of these loci to GAS virulence, as well as the molecular events governing PerR gene regulation, should further elucidate PerR function during human infection.
GAS strain DLS003 is an M-type 3 strain isolated from a patient with necrotizing fasciitis [17]. GAS was grown at 37°C in Todd-Hewitt broth (Difco Laboratories) supplemented with 0. 5% yeast extract (THY), or on THY agar or trypticase-soy agar, both supplemented with 5% defibrinated sheep blood. For cloning experiments, Escherichia coli DH5α or XL1-blue was grown in Luria-Bertani (LB) broth (Difco Laboratories) or on LB agar. When appropriate, antibiotics were added at the following concentrations: ampicillin 100 µg/ml; chloramphenicol 20 µg/ml for E. coli, 10 µg/ml for GAS; erythromycin 300 µg/ml for E. coli, 0. 5 µg/ml for GAS; gentamicin 200 µg/ml; kanamycin 30 µg/ml; penicillin 20 µg/ml; streptomycin 200 µg/ml. Plasmid pJL1055 is a temperature-sensitive E. coli–Gram-positive shuttle vector previously used for allelic replacement mutagenesis in group B Streptococcus [37]; pORI23 is an E. coli–Gram-positive shuttle vector previously used for complementation studies in GAS [38], [39]. Restriction endonuclease digestions, DNA ligations and transformations of chemically competent E. coli were performed using standard protocols [40]. GAS chromosomal DNA isolation and GAS electroporations were performed as described [41], [42]. Total GAS RNA was isolated using the RNeasy mini kit (Qiagen) from bacterial lysates obtained by shaking with glass beads on a dental amalgamator [43]. Primers used were designed based on the M-type 3 GAS strain MGAS315 genome sequence [18] (Table S2). PCR was performed with Taq Platinum DNA polymerase (Invitrogen). Semi-quantitative RT-PCR was performed using the Access RT-PCR system (Promega) for 20–30 amplification cycles and 30 ng total bacterial RNA template. qRT-PCR was performed on an ABI PRISM 7000 Sequence Detection System (Applied Biosystems) using the QuantiTect SYBR green RT-PCR kit (Qiagen) as described [39]. Expression levels of each test gene were normalized to those of recA (spyM3_1800), which did not show any change in expression as a consequence of perR mutation. Data were reported as mean relative expression levels (±standard deviation) in wild-type versus perR mutant. The perR deletion mutant was derived from strain 003Sm, a spontaneous streptomycin resistant (SmR) variant of wild-type strain DLS003. Sm resistance was required for baboon throat colonization experiments (see below). Wild-type perR (spyM3_0147) on 003Sm chromosome was exchanged with a truncated perR copy by allelic exchange as described [43]. Approximately 500 bp of perR upstream and downstream flanking sequence encompassing approximately 20 bp of the 5′-end and the 3′-end of perR, respectively, was obtained by PCR using the 003Sm chromosome as template and primer pairs spyM3_0146-F/perR-R and perR-F/spyM3_0148-R, respectively. The forward primer binding to the 3′-end of perR (perR-F) was designed to carry complementary sequence to the reverse primer binding to the 5′-end of the gene (perR-R). The overlapping sequence incorporated in the two PCR products allowed fusion of the two fragments in a subsequent PCR by using both fragments as template and the two outside primers (spyM3_0146-F and spyM3_0148-R) for amplification. The resulting fragment carrying the truncated perR copy was cloned into the temperature-sensitive E. coli-Gram-positive shuttle vector pJL1055 to create plasmid pjLperRΔ. The perRΔ construct was introduced into 003Sm by electroporation and crossed into the chromosome by homologous recombination with subsequent excision of the wild-type copy as described [43]. The perR deletion mutant strain 003SmperRΔ was obtained from integrant strain 003Sm-Int-perRΔ following several passages at 30°C in the absence of antibiotics in order to promote pJLperRΔ excision from the chromosome. The chromosomal deletion and the lack of perR specific mRNA in strain 003SmperRΔ were confirmed by PCR and reverse transcriptase PCR (RT-PCR), respectively. PCR and qRT-PCR amplification of perR flanking genes (spyM3_0146 and spyM3_0148) yielded identical results in wild-type and mutant bacteria indicating no upstream or downstream effects during perR mutagenesis. Loss of PerR expression in the mutant was confirmed by immunoblotting of mutant protein extracts with PerR antiserum raised against purified recombinant PerR. For perR complementation, the perR coding sequence including the predicted ribosomal binding site was amplified from the wild-type strain 003Sm chromosome using primer pair perR-F-RBS (BamHI) /perR-R (PstI) and was cloned into BamHI/PstI-digested shuttle vector pORI23. The resulting plasmid pORI-perR was introduced into perR mutant strain 003SmperRΔ by electroporation. PerR expression in trans was confirmed by immunoblotting of cell lysates of 003SmperRΔ (pORI-perR), or 003SmperRΔ (pORI23) as control, using PerR specific antiserum. These studies showed significantly higher PerR expression level in strain 003SmperRΔ (pORI-perR) than in the wild-type strain 003Sm (data not shown). A 465 bp fragment encoding PerR without its start codon was amplified by PCR from GAS strain 003Sm chromosome using primers perR-F (BamHI) and perR-R (HindIII). The product was cloned into BamHI/HindIII-digested pQE30 (Qiagen) downstream of the RGS-his6 tag sequence, and expression of the 19 kDa his-tagged recombinant peptide was performed in E. coli strain M15 (pREP4) using IPTG induction (1 mM final concentration). Purification from E. coli lysates was achieved by Ni2+-affinity chromatography using Ni-NTA resin (Qiagen) under native conditions. The purified recombinant peptide was used to immunize rabbits for PerR antiserum production. GAS protein preparations were obtained from 10 ml cultures grown to mid-exponential growth phase as described [39]. Preparations were analyzed by SDS-PAGE under reducing conditions using 10% NuPAGE Bis-Tris gels (Invitrogen). For immunoblotting, proteins were transferred onto a nitrocellulose membrane and blocked in PBS-5% milk (Difco Laboratories). Anti-PerR rabbit antiserum was added at 1∶5000 in PBS-5% milk for 1 h at room temperature followed by 1 h incubation in peroxidase-conjugated secondary antibody (GE Healthcare). Membrane development was performed using the ECL detection kit (GE Healthcare). Resistance of GAS to oxidative killing by H2O2 was tested as described [6] with slight modifications. Bacteria were grown overnight (16 h) in 11 ml THY broth in sealed culture tubes and then sub-cultured in 10 ml broth at a starting density of A600nm = 0. 05. Cultures were grown for approximately 2 h to early exponential phase (A600nm = 0. 15) at which point they were challenged with H2O2 at a final concentration of 10 mM. Following 1 h incubation with H2O2 at 37°C, culture samples were removed for quantitative GAS culture on tryptic soy-blood agar under aerobic conditions. Percent survival was calculated by dividing the GAS counts obtained after H2O2 challenge with those obtained immediately prior challenge. Data represent the mean (±standard deviation) obtained from three independent determinations. GAS sensitivity to phagocytic killing by human phagocytes was assessed using the Lancefield bactericidal test [19]. Approximately 104 GAS cfu grown to early exponential phase (A600nm = 0. 15) in THY broth were rotated end-over-end at 37°C with heparinized human blood from healthy volunteers. Proliferation or killing of GAS was determined by comparing bacterial counts at 3 h post-incubation with those at time of inoculation (0 h), as obtained by quantitative culture of samples on tryptic soy-blood plates. Negative control assays consisted of GAS-blood suspensions incubated without rotation that is required for efficient phagocytosis. For positive control, an M-type 3 protein mutant [44] that shows reduced survival in human blood was assayed in parallel. Mouse RAW264. 7 macrophages [45] were maintained in Dulbecco' s modified Eagle' s medium (DMEM, Gibco) supplemented with 10% fetal calf serum. For GAS infections, wild-type and perR mutant bacteria were grown to early exponential phase (A600nm = 0. 15–0. 2), washed and suspended in DMEM (no serum), then used to infect macrophages grown in 24-well culture plates at 37°C in 5% CO2 at a multiplicity of infection of 0. 5–1. Following 1 h infection, non-associated bacteria were removed by washing twice with the infected media and the cells were lifted with 0. 25 ml H2O and lysed with an equal volume of 0. 1% Triton X-100 (0. 05% final concentration). Macrophage-associated (adherent and internalized) GAS counts were determined by plating serial dilutions of macrophage lysates on tryptic soy-blood agar. For determination of intracellular GAS counts, gentamicin-penicillin exclusion assays were performed. In brief, macrophages were infected for 1 h as described above and extracellular bacteria were killed by the addition of gentamicin-penicillin (200 µg/ml–20 µg/ml, respectively) for an additional hour. Antibiotics were removed with three PBS washes and intracellular GAS counts were determined in macrophage lysates obtained with Triton X-100 as described above. For certain experiments, the general oxidative burst inhibitor diphenyleneiodonium chloride (DPI) (Sigma) was added during the 1 h antibiotic treatment period at a final concentration of 10 µM. Data are presented as GAS counts recovered from each well (∼106 macrophages) and represent the mean cfu (±standard deviation) obtained from three independent determinations each performed in duplicate. Reduced ROS production in DPI-treated macrophages, versus untreated cells as control, was confirmed using the ROS indicator dye dichlorodihydrofluorescein diacetate (H2-DCF-DA) (Invitrogen) at 10 µM final concentration. DCF fluorescence due to reaction with ROS inside macrophages was measured with a Synergy 2 fluorometer (Bio-Tek) with excitation and emission filters of 485±20 and 528±20 nm, respectively. DPI did not affect macrophage viability as monitored by trypan blue staining of DPI-treated and untreated cells. Experiments in baboons were performed at the primate center of the University of Oklahoma Health Sciences Center (OUHSC) as described with minor modifications [21]. Prior to inoculation, throat cultures of all animals used were confirmed GAS-negative. Groups of five baboons (2–2. 7 y), matched for age and sex, were co-infected with a 1∶1 mixture of stationary phase-grown wild-type GAS strain 003Sm and integrant strain 003Sm-Int- perRΔ, or with wild-type strain 003Sm and integrant strain 003Sm-Int-WT, as a control group. Strains 003Sm-Int-perRΔ and 003Sm-Int-WT were derived by chromosomal integration of plasmid pJLperRΔ either upstream or downstream of perR (spyM3_0147) resulting in perR mutant and wild-type phenotype, respectively. Lack of PerR expression in strain 003Sm-Int-perRΔ and wild-type levels of PerR in strain 003Sm-Int-WT was confirmed both by qRT-PCR and immunoblotting of protein lysates of the two strains using anti-PerR rabbit antiserum. No loss of Cm resistance due to pJLperRΔ spontaneous excision from the chromosome could be detected in either of two integrant strains following overnight broth culture (16–18 h) at 37°C in the absence of antibiotics, as determined by quantitative culture of duplicate samples on blood agar carrying Sm or Cm. Thus, chromosomal integration of pJLperRΔ in the two integrant strains was stable. Neither of the two integrant strains showed any competitive advantage or disadvantage in growth when each of the two strains was co-cultured with wild-type strain 003Sm in THY broth in vitro, as determined by quantitative culture of samples obtained before and after co-culture on Sm or Cm. For inoculation, animals were anaesthetized and the posterior pharynx infected by pipeting 1 ml of bacterial suspension carrying 1. 5×108 cfu of wild-type strain and an equivalent number of one of two integrant strains. Throat swabs were collected from anaesthetized animals at 1,3, 24,48, and 72 h and at 7,14,21,28,35 and 42 d post-inoculation. The bacteria on each swab were suspended in 2 ml of THY-10% glycerol, the suspension was mixed vigorously and immediately frozen on dry ice and stored at −80°C. Samples of the inoculum were processed similarly. Freezing of GAS in THY-10% glycerol preserves viability for at least 3 mo. The frozen samples were thawed, serially diluted, and plated in duplicate on THY-blood agar containing either streptomycin to determine the total GAS cfu in each sample, or chloramphenicol to determine the proportion of GAS cfu representing plasmid integrant strains 003Sm-Int-WT or 003Sm-Int-perRΔ. Each of the two antibiotics inhibited growth of the baboon pharyngeal microflora. The competitive colonization index (CCI) for each animal at each time point represents the cfu count of wild-type bacteria divided by the cfu count of the plasmid integrant strain studied. When no colonies were recovered for one of the two strains studied in each sample, the animal was considered negative for that strain and zero counts are shown. However, a cfu count of nine, and not zero, was used for CCI determination in these samples. Since the throat swab culture detection limit was 10 cfu/animal for each strain, 9 cfu would be the highest possible undetectable count and, therefore, such approach represents a conservative analytic strategy. A CCI value was not recorded and is not shown for swab samples that yielded no colonies of either of the two co-infecting strains. An animal was considered to have cleared a strain at the time point beyond which subsequent throat cultures did not yield any colonies of that strain. The GAS DNA microarray used is a modified version of that reported previously [39] and represents all ORFs of M-type 1 GAS strain SF370, in addition to 71 M-type 18 and 21 M-type 3 specific ORFs. cDNA synthesis and labeling, microarray hybridizations and data analyses were performed as described [39]. Transcriptome comparisons were performed between wild-type strain 003Sm and perR mutant strain 003SmperRΔ grown to mid- (A600nm = 0. 25) or late-exponential (A600nm = 0. 6) phase. For each strain, total bacterial RNA was extracted from four independent cultures, pooled and used for cDNA synthesis and labeling. In each experiment, the wild-type and mutant RNA samples to be compared were converted to cDNA and fluorescently labeled with Cy3 or Cy5 in direct (Cy3-Cy5) and dye swap (Cy5-Cy3) labeling reactions to correct for dye-dependent variation of labeling efficiency, and co-hybridized on the arrays. Statistical significance of the difference in mean fluorescence intensity calculated for each gene in the two RNA samples compared was determined by an unpaired 2-tailed Student T-test. Genes whose relative expression (fold-change) exceeded 1. 6-fold and had P values ≤0. 01 were considered as differentially expressed. The False Discovery Rate, estimated by the procedure described by the National Institute of Aging array analysis tool, was ≤0. 05 ([46]; http: //Igsun. grc. nia. gov/ANOVA/). The primary microarray data have been submitted to ArrayExpress database of the European Bioinformatics Institute under accession number E-MEXP-1625. The M-type 3 strain MGAS315 genome sequence [18] was searched for PerR [TTANAATNATTNTAA] and AdcR [TTAAC (T/C) (A/G) GTTAA] conserved motif sequences using both ACMES (Advanced Content Matching Engine for Sequences) [47] provided by the University of Missouri-Columbia and the PredictRegulon web server [48]. Binding of recombinant PerR to DNA fragments representing putative PerR-regulated promoters was performed as described with modifications [8]. Promoter DNA fragments (220–300 bp) were obtained by PCR using wild-type strain 003Sm chromosome as template and primers described in Table S2. Purified PerR was diluted in TDG buffer (50 mM Tris, pH 8. 0; 0. 1 mM DTT; 5% glycerol) and was mixed with promoter DNA for 15 min at room temperature in 10 µl reactions that contained the following: 45 mM Tris (pH 8. 0), 1 mM DTT, 50 ng/µl bovine serum albumin, 6. 5% glycerol. Promoter DNA was added at a final concentration of 50 nM. Reactions were resolved on 6% native polyacrylamide DNA retardation gels (Invitrogen) at 100 V for 100 min and gels were stained with ethidium bromide. Negative control assays were carried out with the 246 bp promoter of guaB, encoding inosine monophosphate dehydrogenase [49], which is not PerR-regulated. | Group A Streptococcus (GAS) is the leading cause of bacterial pharyngitis (strep throat) in children. It is also associated with the life-threatening conditions of necrotizing fasciitis, streptococcal toxic shock, and the post-streptococcal autoimmune syndrome of rheumatic fever and heart disease. The peroxide response regulator PerR is thought to enhance GAS pathogenesis by coordinating adaptive responses of the bacteria in oxidative environments, but its specific contribution to GAS adaptation and virulence at various host sites during infection is poorly understood. In studies comparing phagocytic killing resistance of wild-type and a perR mutant strain of GAS, we find that PerR contributes to bacterial survival in the oxidative environment of phagocytes either during growth in human blood or during interaction with mouse macrophages. Moreover, our studies in a primate model of human pharyngitis suggest that PerR is critical for pharyngeal infection, as PerR mutant bacteria were able to establish infection in one of five inoculated animals in contrast to wild-type bacteria, which established infection in all five animals used. Initial investigation into the genes controlled by PerR in GAS identified a range of bacterial products that may contribute to GAS survival and virulence during infection. | Abstract
Introduction
Results
Discussion
Materials and Methods | infectious diseases/bacterial infections
microbiology/cellular microbiology and pathogenesis
infectious diseases/respiratory infections | 2008 | PerR Confers Phagocytic Killing Resistance and Allows Pharyngeal Colonization by Group A Streptococcus | 14,038 | 312 |
The flavivirus NS5 harbors both a methyltransferase (MTase) and an RNA-dependent RNA polymerase (RdRP). Both enzyme activities of NS5 are critical for viral replication. Recently, the full-length NS5 crystal structure of Japanese encephalitis virus reveals a conserved MTase-RdRP interface that features two conserved components: a six-residue hydrophobic network and a GTR sequence. Here we showed for the first time that these key interface components are essential for flavivirus replication by various reverse genetics approaches. Interestingly, some replication-impaired variants generated a common compensatory NS5 mutation outside the interface (L322F), providing novel routes to further explore the crosstalk between MTase and RdRP.
The genus of Flavivirus within the family Flaviviridae contains large amounts of arthropod-borne viruses, which includes Japanese encephalitis virus (JEV), West Nile virus (WNV), Dengue virus (DENV), Tick-borne encephalitis virus (TBEV) and yellow fever virus (YFV) [1]. Most of these viruses are important human and animal pathogens. So far, no effective antiviral drug is available to treat flavivirus infections [2]. The study of viral replication mechanism will help to develop an efficacious antiviral therapy against flaviviruses. The genome of flaviviruses is a positive-sense single-stranded RNA which contains a 5′ non-translated region (NTR) with type I cap structure at its 5′ end, an open reading frame (ORF) and a 3′ NTR without a poly (A) tail. The ORF encodes a polyprotein that is cleaved into three structural proteins (Capsid [C], premembrane [prM] and Envelope [E]) and seven non-structural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5) by a combination of viral and host proteinases [3]. The non-structural proteins play critical roles during viral RNA replication, virion assembly, and evasion of host immune responses [4]–[9]. NS5, the largest and most conserved flavivirus protein, is a multi-function protein that comprises an N-terminal methyltransferase (MTase) and a C-terminal RNA-dependent RNA polymerase (RdRP). The MTase domain carries out both guanine-N7 (N7) and nucleoside-2′-O (2′-O) methylation steps in the 5′ end capping processes of the viral genome [10]. The N7 methylation is essential for viral replication and the 2′-O methylation is mainly involved in host immune response [11]–[13]. RdRP is responsible for viral RNA replication through a de novo initiation mechanism in a primer-independent fashion [14]. In addition, it has been reported that RNA guanylyltransferase (GTase) also resides in the MTase domain, and NS3 could stimulate GTase activity of NS5 [15]. The intact NS5 protein also interacts with viral protein NS3 [16]–[17] and different host proteins [17]–[19], and modulates innate immune response [20]–[24] in viral infection. The deciphering of NS5 intra-molecular interaction will help to understand the versatile functions of NS5 during viral infection. Although the interactions between MTase and RdRP have been demonstrated by reverse genetics, biochemical, and structural approaches [25]–[28], it was only until recently that the high-resolution details of the intra-molecular interactions between MTase and RdRP of flavivirus NS5 was identified with the crystal structure of the integral JEV NS5 [29]. The MTase-RdRP interface (Fig. 1A) contains two key components, which are a hydrophobic network and a GTR sequence hypothesized to mediate the interface formation [29]. The hydrophobic network is composed of three residues P113, L115, and W121 within the RdRP interacting module (residues 112–128) of MTase and three residues F467, F351, and P585 in three finger subdomains of RdRP (Fig. 1A). These six residues are arranged in an alternating pattern, thus forming a conserved hydrophobic network in the heart of the interface. The GTR sequence (residues 263–265) is the last three residues of the MTase and is located at the edge of the interface and spatially near W121 of the hydrophobic network (Fig. 1A). In the full-length JEV NS5 structure, the GTR sequence mediates the MTase-RdRP interactions mostly through hydrogen bonding interactions (G263 and R265 form a total of seven hydrogen bonds with the rest of MTase, while T264 forms two hydrogen bonds with RdRP), and the flexibility offered by G263 likely accounts for the ≈60° turn of the main-chain direction at this position [29]. All nine residues of the two key interface components are highly conserved in flaviviruses except for residue 115 that is generally hydrophobic (Fig. 1B), indicating their functional relevance in NS5. In this study, we first explored the biological functions of the intra-molecular interactions between MTase and RdRP identified by the JEV NS5 crystal structure using JEV and DENV-2 infectious clones. We showed that both hydrophobic network and GTR sequence in the MTase-RdRP interface are essential for replication of JEV and DENV-2, which indicated the conservative functions of MTase-RdRP interactions in flavivirus life cycles. Using a transient replicon system of JEV, we further demonstrated that the mutations of the conserved residues in the interface impaired viral RNA replication. Moreover, a common compensatory mutation L322F identified outside of the interface can restore the P113D and F467D RNA replication, paving a way for further investigation of the interregulation between MTase and RdRP.
Baby hamster kidney cells (BHK-21) was propagated in Dulbecco' s modified Eagle' s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 units/mL of penicillin and 100 µg/mL of streptomycin in 5% CO2 at 37°C. Monoclonal antibody against envelope protein of St. Louis encephalitis virus (SLEV) was used as first antibody, which is cross-reactive to JEV envelope protein. Texas Red-conjugated goat anti-mouse IgG was used as secondary antibody. The infectious clones of pACYC-JEV-SA14 [30] and pACYC-NGC [31] were used as the backbone to construct mutant genome-length cDNA clones with different NS5 mutations for JEV and DENV-2, respectively. All mutations were engineered by fusion PCR using standard procedures. The JEV NS5 mutations were inserted into pACYC-JEV-SA14 at BamHI and XbaI restriction sites except that the P585 mutation at XbaI and SalI restriction sites. All DENV-2 NS5 mutations were cloned to pACYC-NGC by restriction digest with NruI and MluI. A JEV replicon containing Renilla luciferase (Rluc) reporting gene [30] was used to examine the effects of compensatory mutations on viral replication. To construct Rluc-Rep mutants containing various NS5 mutations, BamHI/XbaI DNA fragment from the JEV infectious clone containing NS5 mutations was inserted into the JEV Rluc-Rep at the same sites. All constructs were verified by DNA sequencing before they were used in the subsequent experiments. The infectious clone and replicon cDNA plasmids of JEV were linearized with XhoI and the infectious clone of cDNA plasmids of DENV-2 were digested with ClaI to linearize. The purified linearized cDNAs were then subjected to in vitro transcription using the T7 in vitro transcription Kit. All procedures were performed according to the manufacturer' s protocols. The resulting RNA was resolved in RNase-free water and stored at −80°C in aliquots. Approximately 10 µg RNA was electroporated into 8×106 BHK-21 cells in 0. 8 ml of ice cold PBS buffer (pH 7. 5) in a 0. 4 cm cuvette at 0. 85 kV and 25 µF, pulsing three times at 3 sec intervals. After a 10-min recovery at room temperature, the transfected cells were mixed with 25 ml pre-warmed DMEM containing 10% FBS. The RNA-transfected cells were seeded on a Chamber Slide. At 24,48,72 and 120 hours post transfection (hpt), the cells were fixed in cold (−20°C) 5% acetic acid in methanol for 10 min at room temperature. The fixed cells were washed with PBS three times and then incubated with anti-SLEV envelope protein monoclonal antibody (1∶250 dilution with PBS) for 1 h at room temperature. Cells were washed with PBS three times and then incubated with goat anti-mouse IgG conjugated with Texas-Red at room temperature for 45 min. Following three times of PBS washing and a 10-min DAPI incubation to stain the nuclei, the slide was mounted with 95% glycerol and visualized under a fluorescence microscope. Cell images were taken at 200× magnification. Virus stock was produced by harvesting the supernatant of RNA-transfected BHK-21 cells. Two methods (double layer and monolayer) were performed for plaque assay. Virus titer and morphologyapp: addword: morphology were determined by double layer plaque assay. Briefly, a series of 1∶10 dilutions were prepared by diluting 15 µl virus stock with 135 µl DMEM containing 10% FBS, and then 100 µl of each dilution were seeded onto each well of 6-well plates containing confluent BHK-21 cells (5×105 cells per well, plated 1 day in advance). The plates were incubated at 37°C with 5% CO2 for 1 h before the first layer of agar was added. After 72 h of incubation at 37°C with 5% CO2, a second layer of agar containing neutral red was added. Plaques were photographed and numbered after neutral red incubation of the plates for another 12 to 24 h. The viral titer was calculated as plaque formatting unit (PFU) per milliliter. The limit of detection is 10 PFU/ml. Virus production assay was analyzed by monolayer plaque assay. As described above, viral supernatants of different time points were serially diluted 10-fold and were used to infect confluent BHK-21 cells (1×105 cells per well, plated 1 day in advance) in 24-well plates. The infected cells were incubated at 37°C with 5% CO2 for 1 h before the layer of medium containing 2% methylcellulose was added. After 4 days of incubation at 37°C with 5% CO2, the cells were fixed in 3. 7% formaldehyde and then stained with 1% crystal violet. The viral titer was calculated as plaque formatting unit (PFU) per milliliter. Viral RNA was extracted from infected cells using an RNA extraction kit following the manufacturer' s protocol. The extracted RNA was used to six one-step RT-PCR reactions by using a series of overlapping primers covering the complete viral genome of JEV. The RT-PCR products were purified with gel extraction kit and subjected to DNA sequencing for viral genome alignment. Replicon RNA transfected cells were seeded in 12-well plates in various amounts. At 2,12,24,48,72 and 96 hpt, the medium was removed, and the cells were washed with PBS once, 200 µl lysis buffer was added to each well of 12-well plate, the transfected cells were reclaimed and stored at −80°C for subsequent luciferase assay. Triplicate wells were seeded for each time point. Luciferase activity was measured in a microplate reader by mixing 20 µl lysates with 50 µl substrate into one well of 96-well plate.
We first performed a systematic mutagenesis analysis of the six key residues (P113, L115, W121, F467, F351 and P585) in hydrophobic network for its biological function study using an infectious clone of JEV (Fig. 1A). The substitutions of each of these six hydrophobic residues with Arginine (R), Aspartic acid (D) and Serine (S) were designed to alter or disrupt the hydrophobic network. D and S represent amino acids with charged and uncharged polar side chains, while the side chain of R could provide a variety of interactions through its charged guanidinium head and the 3-carbon aliphatic chain. Equal amounts (10 µg) of wild type (WT) and mutant viral RNAs were transfected into BHK-21 cells, and viral protein expression, plaque morphology and virus production were compared between WT and the mutants. First, viral protein (envelope) expression in transfected cells was detected by immunofluorescence assay (IFA) (Fig. 2A). In comparison with WT (100% IFA positive cells observed at 72 hpt), all mutants either failed to produce or produced much fewer IFA positive cells. Among all the engineered mutations, the residues P113, W121, F467 and P585 played a comparably vital role in viral replication since most D/R/S substitutions dramatically decreased the numbers of IFA-positive cells, albeit Serine overall has a more modest effect among all three types of substitutions. Specifically, P113R, P113S and W121S produced much less IFA-positive (1%, 50% and 20% relative to WT, respectively) cells than WT at 72 hpt; P113D, F467D and P585S occasionally produced IFA-positive cells (less than 1%) at 120 hpt and no IFA-positive cells were observed in W121R/D, F467R/S and P585R/D at the same time points. Comparably, L115 and F351 sites were tolerant with D/R/S substitutions (Fig. 2A). In particular, R/D/S mutations at L115 still led to 100% IFA-positive cells at 120 hpt. Since L115 is less conserved among flavivirus NS5 (Fig. 1B) [29], it is conceivable that L115 substitution mutants had less impact on viral replication. Then the culture supernatants from transfected cells were used for plaque assay to compare plaque morphology alterations. All mutations affected plaque morphology with different extents. As shown in Fig. 2A, no plaques were formed for P113R/D, W121R/D, F467R/D/S and P585R/D; pin-point-sized plaques were generated for P113S, L115D, W121S, F351R/D/S; plaque size was increased for the other mutants but still smaller than WT. Finally, virus productions were also quantified by plaque assay at each time point post transfection. Consistent with the results of IFA and plaque morphology, only P113S, L115R/D/S, W121S, F351R/D/S and P585S mutant RNAs yielded viruses, although much lower than WT RNA at each time point and the other mutant RNAs did not yield any detectable viruses (Fig. 2B). Overall, the results indicated that hydrophobic residues in the interface of MTase and RdRP are important for viral replication. We next tested the relevance of the invariant GTR sequence in viral replication by engineering G263A, T264V, R265K and R265L mutations into JEV infectious clone. In most of NS5 structures, G263 adopts glycine-specific phi-psi angles to allow a 60° turn of the main chain [29], [32], and the G263A mutation was expected to alter the main chain direction, which in turn could modulate the MTase-RdRP interactions. T264V, R265K and R265L were supposed to alter hydrogen bonding network formed between MTase and RdRP but may retain the capability for other interactions [29]. At 72 hpt, only T264V and R265K mutants produced a small amount of IFA-positive cells with 10% and 1%, respectively (Fig. 3A). G263A produced around 5% IFA-positive cells at 120 hpt and no IFA-positive cells were observed in R265L mutant at that time. For plaque morphology, small plaques were detected for most GTR mutations except for R265L with no plaque observed (Fig. 3A). Similarly, virus production for GTR mutations were dramatically impaired comparing with WT, and no virus production was detected for R265L even after 120 hpt (Fig. 3B). The results indicated that the GTR sequence is also essential for viral replication. To test the functional conservation of the aforementioned MTase-RdRP interactions in flaviviruses, we performed similar analysis using a DENV-2 infectious clone. Each of the corresponding hydrophobic residues (P113, P115, W121, F465, F349, and P583) in the hydrophobic network was mutated to Aspartic acid and the same types of mutations were tested in the GTR residues 261-263. Similar results were observed in DENV-2 (Fig. 4A). For the mutations from hydrophobic network, no IFA positive cells were observed in P113D, W121D and F465D-transfected cells; F349D mutant produced less than 1% IFA-positive cells. For GTR mutants, no IFA-positive cells were found in G261A and T262V, only less than 5% IFA positive cells were observed for R263K and R263L. Plaque morphologies were compared by using supernatants from 144 hpt, and except G261A and R263L, all other mutations produced small plaques in contrast to WT. Furthermore, virus production was impaired by all mutations (Fig. 4B). P115D had much less effect on virus production at each time point post transfection, consistent with the data from JEV as P115 is less conserved among flaviviruses (Fig. 1B). Overall, the results demonstrated that both hydrophobic network and GTR sequence of MTase and RdRP interface of NS5 are essential for replication of both JEV and DENV-2. The supernatants from all JEV mutant RNA transfected BHK-21 cells were continuously passaged in BHK-21 cells (3-4 days per passage) for five rounds in triplicate to determine the stability of engineered mutation or to recover the possible adaptive viruses. The results from the sign of CPE in infected cells, plaque assay and JEV sequence specific RT-PCR (data not shown and representative plaque morphology changes were shown in Fig. 5A) consistently showed that no recovered viruses occurred from mutants of F467R/S, W121R/D, P585R/D and R265L, which again demonstrated that these residues of interface are essential for JEV replication. All of the recovered viruses were then subjected to sequencing of the complete NS5 region, and consistent mutations were identified for each passaged mutant viruses in triplicate. The sequencing results were summarized in Fig. 5B. Specifically, P113S, L115S, W121S, F351S, P585S and T264V still retained engineered mutations. L115D and F351D produced a mixture of D with N and D with Y, respectively, and P113R generated a wild type revertant or a P113G mutation. At the same time, secondary adaptive mutations of A336T, G806R, D170G+Q353L, P113G+L322F and L322F were identified for F351R, R265K, G263A, P113D and F467D mutants, respectively. In Fig. 5C, all the second site compensatory mutations (large spheres) were also summarized and indicated in the crystal structure of NS5 with the same color as its original mutations (small spheres). Interestingly, L322F is a common adaptive mutation for both P113D and F467D that reside in MTase and RdRP, respectively. This residue sits at the interface between the RdRP thumb and index finger and is not part of the MTase-RdRP interface. As a control experiment, wild type viruses were also passaged for five rounds according to the same protocols as other mutants viruses, and no mutations (data not shown) were found in NS5 region. Taken together, crosstalk may exist not only at the interface, but also among different components within the MTase or RdRP. Next, we focused on the biological study of the L322F adaptive mutation derived from P113D and F467D. To examine the role of adaptive mutations in rescuing viral replication for P113D and F467D, we generated P113G, L322F, P113G+L322F, and P113D+L322F variants of JEV genome-length RNAs. The same protocol was used to compare viral replication efficiency as mentioned above for JEV. Compared with P113D, single mutant P113G rescued IFA positive cells at 48 and 72 hpt with low efficiency; the addition of L322F further improved viral replication, generating more IFA positive cells (Fig. 6A). Interestingly, combining L322F with P113D could also rescue P113D viral replication as P113D+L322F could produce IFA positive cells at 48 and 72 hpt, although the P113D+L322F combination was not identified from recovered viruses. Similar results were observed for F467D adaptive mutation; an additional L322F mutation rescued the replication of F467D mutant as L322F+F467D produced IFA positive cells at the indicated time points post transfection compared with F467D mutation alone (Fig. 6B). Consistent with the results in IFA, the adaptive mutations, especially L322F were also able to recover the abilities of both plaque-forming and virus production of mutant RNAs (Fig. 6A–C). For viral production assay, L322F alone produced slightly higher amount of viruses comparing with WT at 24 and 48 hour post transfection (Fig. 6C), although L322F alone produced similar amounts of IFA positive cells as WT (Fig. 6A), indicating that L322F itself may slightly increase viral replication. Overall, the results confirmed that compensatory mutations L322F could rescue the replication defect of P113D mutation in MTase and F467D mutation in RdRP, respectively. To directly measure the effect of various mutations on viral RNA replication, we tested representative mutations (P113D and F467D) and their compensatory mutations in a transient replicon system of JEV. The JEV replicon (JEV-Rluc-Rep) [30] was constructed by replacing the structural genes with a Renilla luciferase sequence and a foot-and-mouth disease virus 2A sequence in its open reading frame (Fig. 7A). Previous studies showed that BHK-21 cells transfected with flavivirus Rluc-Rep RNA generates two distinct luciferase peaks which represent viral RNA translation and viral RNA replication, respectively [1]. The first peak at 2 hpt, was used to quantify input RNA translation and the second peak after 12 hpt was taken as a surrogate readout of RNA replication. Using this system, we transfected BHK-21 cells with equal amounts of WT and mutant replicon RNAs and then assayed their luciferase activities at various time points post transfection (Fig. 7B). All replicons yielded roughly equal levels of luciferase signal at 2 hpt, indicating that the translation was similar for the various replicon RNAs and mutations have minimum effect on viral translation. For the P113D replicon, only a background level of luciferase signal was detected from 24 to 96 hpt, suggesting that P113D replicon was non-replicative. In contrast, the L322F replicon replicated to a level slightly higher than that of the WT replicon at early time point (Fig. 7B) which are consistent with the results from the viral production assay (Fig. 6C). The compensatory mutations rescued the efficient replication of P113D replicon in the order of L322F+P113D<P113G<L322F+P113G (Fig. 7B). For the F467D replicon, a low level of luciferase signal was detected from 24 to 96 hpt, and the addition of L332F to the F467D replicon did improve replicon RNA replication (Fig. 7B). Overall, the data of the replicon assay provided further evidence that mutations of the critical residues in the interface between MTase and RdRP could impair viral RNA replication, and the compensatory mutations could restore the replication ability of these interface mutants. We have shown that two sets of conserved residues within the recently identified MTase-RdRP interface of flaviviruses NS5 play essential roles for flavivirus replication, providing the first functional validation of the structural work in JEV NS5 [29]. It is of particular interest that a common adaptive mutation of L322F which resides in the thumb-interacting region of the RdRP index finger (Fig. 5C, large red sphere) was identified from the recovered viruses from P113D (MTase) and F467D (RdRP) mutants. The tip of the index finger interacts with thumb domain to form the viral RdRP unique encircled active site [33]–[35]. In HCV NS5B, this so-called fingertip or Δ1-loop region has been suggested to modulate the enzymatic properties of the polymerase. Altering its interaction with the thumb via site-specific mutagenesis affected the de novo initiation process. In poliovirus polymerase 3Dpol, mutations in the same thumb-index interface led to changes in thermal stability of the protein [36]. Very interestingly, two crystal forms of HCV genotype 2a NS5B were obtained with different conformation at the index-thumb interface [37]. One structure adopts a canonical “tight” conformation, with extensive hydrophobic interactions at the thumb-index junction. In contrast, the other structure exhibits a “loose” conformation that only retains a portion of the thumb-index contacts. For RdRPs from family Flaviviridae in general, it has been suggested that the tight form of the polymerase allows only single-stranded RNA template to enter the active site to stabilize a de novo initiation complex, and the aforementioned thumb-index dynamics likely play important roles to accommodate lengthening of template-product duplex during the transition to the elongation phase [38]–[39]. Hence, adaptive mutation L322F likely modulates the early stages of RNA synthesis by altering the dynamics of thumb-index contacts, and therefore compensates the effect brought by MTase-RdRP interface mutation of F467D or P113D. In the full-length JEV NS5 crystal structure, F467 sits at the tip of the ring finger that contains the NTP binding motif F, and P113 from the MTase directly interacts with F467. While the ring finger is ordered and intact in the full-length JEV NS5 structure, it is largely disordered in the WNV and DENV-2 RdRP crystal structures and does not adopt the canonical conformation observed in other viral RdRPs [25], [29], [40]. Hypothetically, the mutation of F467D or P113D could greatly affect the interactions between MTase and RdRP and therefore change the dynamics of ring finger, resulting in the altering of polymerase catalytic activity. The L322F mutation, albeit via a different mechanism, may help compensate the effect brought by F467D or P113D to recover the viral genome replication. Taken together, we have provided evidences for the functional significance of the MTase-RdRP interface revealed by the full-length JEV NS5 crystal structure, and found novel sites within NS5 that could modulate viral replication through different routes. Although the structure-and-function-based interpretation of our data is largely focused on the RdRP module of NS5, we certainly cannot rule out the possibility that the presence and dynamics of the MTase-RdRP interface could regulate the MTase function. Further mechanistic dissection, in particular by in vitro polymerase and MTase assays, is necessary to gain in-depth understanding of the interactions between MTase and RdRP and regulation of JEV replication. Moreover, the interface overlaps with the suggested or putative NS3, importin β, and CRM1-mediated exportin binding sites, and therefore could play a role in a variety of events in the flavivirus life cycle [7], [19], [41]-[42]. Our data have consolidated the breakthrough of the recently reported full-length JEV NS5 crystal structure and may invoke investigations focusing on the MTase-RdRP interface with potentially versatile functions. | Flaviviruses, such as Japanese encephalitis virus (JEV) and dengue virus 1-4 (DENV1-4), are important arthropod-borne pathogens causing major public health threats worldwide. For example, JEV is the most common cause of viral encephalitis in eastern and southern Asia, affecting over 50,000 patients, and leads to 15,000 deaths annually. DENV is responsible for an estimated 50 million cases of dengue fever and over 450,000 cases of life-threatening dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS) each year. Currently, there is no effective antiviral therapeutics available for flaviviruses. Non-structural protein 5 (NS5) is a multifunctional, viral protein that contains a methyltransferase (MTase) and an RNA-dependent RNA polymerase (RdRP). Both MTase and RdRP are required for replication of the viral genome. A conserved interface between MTase and RdRP was first identified in the full-length NS5 crystal structure of JEV. In this study we used the infectious clones of both JEV and DENV-2 to perform functional mutagenesis analyses and demonstrated for the first time that the recently identified conserved interface between MTase and RdRP is critical for viral replication. In a replicon system, we also confirmed that mutations within the conserved interface greatly affect viral RNA replication during viral infection. Our functional validation of the conserved MTase-RdRP interface consolidated its potential as a novel target for anti-flavivirus drug development. | Abstract
Introduction
Materials and Methods
Results and Discussions | viruses
infectious diseases
infectious diseases of the nervous system
medicine and health sciences
japanese encephalitis
rna viruses
viral classification
virology
dengue fever
neglected tropical diseases
biology and life sciences
tropical diseases
microbiology
viral diseases
organisms
encephalitis | 2014 | The Interface between Methyltransferase and Polymerase of NS5 Is Essential for Flavivirus Replication | 7,220 | 373 |
Post Kala-azar Dermal Leishmaniasis (PKDL) develops in patients apparently cured of Visceral Leishmaniasis (VL), and is the strongest contender for being the disease reservoir. Therefore, existence of a few cases is sufficient to trigger an epidemic of VL in a given community, emphasizing the need for its active detection and in turn ensuring success of the current elimination program. This study explored the impact of active surveillance on the demographic profile of PKDL patients in West Bengal. Patients with PKDL were recruited through passive (2003-date, n = 100) and active surveillance (2015-date, n = 202), the former from outpatient departments of dermatology in medical colleges in West Bengal and the latter through an active door-to-door survey in four VL hyper-endemic districts of West Bengal. Passive surveillance indicated a male preponderance and a predominance of polymorphic lesions, whereas active surveillance indicated absence of any gender bias and more importantly, macular PKDL constituted almost 50% of the population burden. In terms of polymorphic vs. macular PKDL, the former appeared at a later age, their disease duration was longer and had a higher parasite burden. In the polymorphic variant, the lesional distribution was asymmetrical, comprised of papules/nodules/macules that were present mainly in sun-exposed areas whereas in macular cases, the hypopigmented patches were diffusely present all over the body. Active surveillance unraveled a disease component whose demographic profile showed important differences with PKDL cases who sought treatment in government hospitals. Detection of a higher proportion of macular cases indicates that this variant is not an uncommon presentation as conventionally stated in text books, and should be studied in greater detail to ensure success of the ongoing Leishmaniasis elimination programme.
Leishmaniasis are a group of neglected tropical diseases caused by the parasite Leishmania and demonstrates clinical pleomorphism with regard to the causative species, disease reservoirs, vectors as also host-immune responses. It can manifest as life-threatening and/or disfiguring lesions ranging from innocuous self-healing cutaneous lesions to fatal visceralization or a dermal dissemination [1]. Post Kala-azar Dermal Leishmaniasis (PKDL) usually presents in patients with a history of treated Visceral Leishmaniasis (VL) / kala-azar caused by L. donovani. It is unique to South Asia and East Africa (mainly Sudan) with approximately 5–10% of apparently cured VL patients developing PKDL in South Asia as against 50–60% in East Africa [2,3]. Unlike VL where patients have significant morbidity and present with prolonged fever, hepatosplenomegaly, weight loss and anemia, patients with PKDL manifest with innocuous hypopigmented macular lesions (macular PKDL) or a combination of macules/papules/nodules (polymorphic PKDL). It is possibly the most challenging variant of Leishmaniasis, especially in terms of its etiopathogenesis [4–6]. The clinical diagnosis of PKDL remains empirical and till now, was mostly dependent on self-reporting i. e. passive surveillance [7]. As PKDL is a disease of the poor, occurring mostly in remote rural villages with poor housing and little or no access to modern health-care facilities and importantly, is not life-threatening, the tendency to actively seek treatment is minimal [8]. In South Asia, the VL endemic zones include the Gangetic plains of India, Bangladesh, and Nepal. In 2005, the governments of Bangladesh, India, and Nepal agreed on a regional initiative, the Regional Kala-azar Elimination Programme (KAEP) to eliminate VL as a public health problem from the region by 2010. This was further extended to 2015, the target being to reduce the annual incidence of kala-azar to less than one new case in a population of 10,000 at a district (or sub-district) level [9]. This target was later extended to 2017, which too has been missed, and presently stands at 2020 [10]. PKDL cases harbor parasites within their dermal lesions, and as they are easily accessed by sandflies, play a major role in the transmission cycle, especially as VL is anthroponotic, emphasizing their inclusion as a component of the ongoing VL elimination programme [11]. PKDL is recognised as a constraint in the Leishmaniasis elimination effort, and accordingly, development of strategies for case finding, diagnosis, and treatment are among the major objectives of the KAEP [11]. However, the strategies are currently not based on solid scientific evidence, limitations being costly case finding, risk of misdiagnosis, and possibility of inadequate and unnecessary treatment with potentially toxic drugs [11]. From 2015 onwards, an ‘active case surveillance’ approach was adopted in the VL/kala-azar hyper endemic districts of West Bengal. This translated into a dramatic rise in the number of PKDL cases [12] and significant changes in the demographic profile of the disease warranting a deeper analysis of the various clinical and epidemiological aspects of PKDL. This study aimed to delineate the differences, if any, in the demographics of PKDL cases detected by active vs. passive surveillance over the last 15 years in West Bengal.
From 2003 to date, patients clinically diagnosed with PKDL were recruited through passive surveillance from the Dermatology outpatient departments of the School of Tropical Medicine/Calcutta Medical College/ Institute of Postgraduate Medical Education and Research, Kolkata, West Bengal. Additionally, from 2015 onwards, active field surveys were conducted in VL endemic districts of West Bengal (Malda, Dinajpur, Darjeeling and Birbhum) by a camp approach, wherein an initial house-to-house survey was conducted by first-line health workers [Kala-azar Technical Supervisors and ASHA workers (accredited social health activists) ] using standard case definitions and defined risk factors, for example, living in an endemic area and having an epidemiological link (past history of VL). The suspected cases were then examined at medical camps held at block hospitals. These camps were conducted by a team of 4–10 members comprising of medical officers, laboratory technicians, health supervisors, health workers and community based health volunteers. The suspected cases were assessed clinically, rk39 strip test and if positive, informed consent was taken for a 4 mm punch biopsy. In general, cases with hypopigmented macules were considered as macular PKDL, whereas cases with an assortment of papules, nodules, macules, and/or plaques were termed as polymorphic PKDL. The cases were included only when confirmed by internal transcribed spacer-1 (ITS-1) PCR [13] and/or Giemsa staining of dermal biopsies. None suffered from any co-infection or pre-existing disease. Owing to the teratogenic potential of miltefosine, and PKDL being a non fatal disease, pregnant women were not offered any treatment and treated after completion of gestation and lactation period. After confirmation by ITS-1 PCR, their parasite load was quantified by amplification of kinetoplastid DNA [12]. Cases were randomly allocated, as per the existing guidelines, to receive Miltefosine (for >25 kg b. w. 100 mg, p. o. x 12 weeks and 50 mg p. o. x 12 weeks for <25 kg) or Liposomal Amphotericin B, LAmB (5 mg/kg body IV, twice weekly, for 3 weeks) [14]. A defined number of Leishmania parasites sourced from an L. donovani strain (ranging from 10 to 1 × 105) was added to blood (180 μL) from a healthy control. Following extraction of DNA using a QIAmp DNA Mini kit (Qiagen, Hilden, Germany), real time -PCR was performed in an Applied Biosystem Step One Plus (Applied Biosystems, Foster City, CA, USA). A fragment of 116 bp of L. donovani kDNA was amplified by using a primer set (forward 5′-CCTATTTTACACCAACCCCCAGT-3′ and reverse 5′-GGGTAGGGGCGTTCTGCGAAA-3′). Template DNA (5 μL) was added to 19 μL reaction mixture containing SYBR Green qPCR Master mix (Roche, Basel, Switzerland) and 400 nM of each primer. For measurement of parasite load, a standard curve was generated as previously described [12]. Negative controls included DNA from a healthy donor (no amplification), and a reaction mixture with water instead of template DNA (no-template control [NTC]). The range of the kDNA qPCR assay had an efficiency of 5 orders of magnitude (1x105 to 1x10 parasite/μg of genomic DNA). The standard curve (ranging from 10 to 1 × 105) had a mean square error of 0. 007, correlation coefficient, r2 = 0. 995. The number of parasites was extrapolated from the standard curve and parasite load stated as the number per μg genomic DNA. As the parasite number when <10 reported a cycle threshold (Ct) value almost equivalent to NTC, it was accorded an arbitrary value of 1. Results were expressed as median (Interquartile range, IQR) and data analyzed between groups by Kruskal wallis test followed by Dunn’s multiple comparison test for non-parametric data; where 2 groups were present, the Mann-Whitney test was used using GraphPad Prism software (version 5. 0, GraphPad software Inc. , La Jolla, CA, USA), p< 0. 05 was considered significant. The study received approval from the Institutional Ethics Committee of School of Tropical Medicine, Kolkata, India and Institute of Post Graduate Medical Education and Research, Kolkata, India. All experiments were performed in accordance with relevant guidelines and regulations. Individuals or their legally acceptable representative (if age was <18 years) gave a written informed consent. The patients in this manuscript have given written informed consent (as outlined in the PLOS consent form) to publication of their case details.
Since 2003, irrespective of the surveillance, 436 suspected cases of PKDL were registered, and included 102 via passive and 334 following active surveillance (Fig 1). Following passive surveillance, 83. 3% gave a history of VL while 100/102 (98%) were confirmed as cases of PKDL (Fig 1); amongst these 100 confirmed cases, 15 (15%) reported no history of VL. Following active surveillance undertaken in the VL/kala-azar hyper endemic districts of Malda, Dinajpur, Darjeeling and Birbhum from 2015–2018, a latent disease burden of 334 suspected cases of PKDL were identified, with 91. 3% having a previous history of VL. Following their initial screening by KTS workers, they were physically examined at a ‘medical camp’ and 202 (60. 5%) were confirmed as PKDL by ITS-1 PCR (Fig 1). Of these confirmed cases, 16 (7. 9%) did not report any prior history of VL. There were 5 rK39 -ve cases (2 from passive surveillance and 3 from active surveillance), but were ITS-1 PCR +ve. The rK39 -ve/ ITS-1 -ve cases (n = 132), were followed up, and their skin lesions identified as pityriasis versicolor, vitiligo or pityriasis alba. During passive surveillance, cases (n = 100) were not only from West Bengal, but included the adjoining states of Bihar, Jharkhand and Uttar Pradesh as also Bangladesh; majority originated from Bihar (n = 54,54%), while the rest were from West Bengal (n = 35,35%), Jharkhand (n = 8,8%), Uttar Pradesh (n = 2,2%) and Bangladesh (n = 1,1%), Fig 2. Of the 54 PKDL cases from Bihar, more than 50% (n = 30,55. 6%) were from districts with high VL endemicity and antimonial resistance [15,16] e. g. Samastipur, Muzzafarpur, Vaishali, Araria, Saran, Purnia and Saharsa, while the rest were from areas of low endemicity namely Siwan, Madhubani, Begusarai, Darbhanga, Kishenganj, Gaya, Jehnanabad, Khagaria and Aurangabad [17]. In West Bengal, patients reported not only from VL-endemic districts (Malda, Murshidabad, Birbhum, Dinajpur, Darjeeling, North 24 Parganas, South 24 Parganas, Nadia and Burdwan), but also from non-VL endemic districts (Kolkata, Howrah, Midnapore and Alipurduar). Following 2015, active surveillance was initiated in four VL-endemic districts of Bengal namely Malda, North and South Dinajpur, Darjeeling and Birbhum that unraveled a huge disease burden (n = 202). These areas reported the maximum number of PKDL cases, as they share their western borders with the VL-endemic states of Bihar and Jharkhand, and eastern borders with Bangladesh. Malda was the leading district with the maximum number of cases (n = 123,60. 9%), followed by South Dinajpur (n = 56,27. 7%), North Dinajpur (n = 16,7. 9%), Darjeeling (n = 5,2. 5%) and Birbhum (n = 2,1%), Fig 2. From January 2015- till date, 16 active field surveys were conducted in these VL endemic districts, namely Malda (6 visits) where 4 blocks were included [Habibpur, Gazole, Old Malda and Bamongola]. In South Dinajpur, 7 field trips were conducted in Kushmandi, Banshihari, Harirampur and Tapan blocks. One field trip each was undertaken in North Dinajpur (2 blocks, Karandighi and Goalkhopar-II), Darjeeling (2 blocks, Khoribari and Phasidewa) and Birbhum (1 block, Bolpur). A male preponderance in PKDL reported during passive surveillance was in concordance with previous reports [4,7, 18,19] (Table 1, Fig 3A), whereas with active surveillance the scenario changed drastically as the ratio became 1: 1. 1, indicating absence of any gender bias (Table 1, Fig 3A). Following pooling of the data, the study demonstrated a male: female ratio of 1. 3: 1, with 57. 3% (n = 173) being males and 41. 1% (n = 129) females (Table 1, Fig 3A). Irrespective of the survey, the median age, disease duration and lag period, i. e. gap between manifestations of PKDL following completion of VL treatment were comparable (Table 1). Interestingly, the percentage of PKDL cases reporting a disease duration of less than 1 year increased following active surveillance as 40/202 (19. 8%) had a disease duration of <1 year vis-a-vis 8/100 (8%) cases from passive surveillance. Out of 302 PKDL cases, parasite load was measured in 229, and on a stratification based on surveillance, cases from the passive survey had a 4. 0 fold higher parasite burden than those recruited from active surveillance, p<0. 01 (Table 1). With regard to the type and distribution of lesions, striking differences emerged depending on the type of surveillance. Following passive surveillance, there was an overwhelming predominance of the polymorphic variant vis-a-vis the macular (Table 1) whereas active surveillance unravelled a huge population having macular lesions, that translated to the ratio of polymorphic: macular shift dramatically to 80: 122. Taken together, the cumulative scenario mimicked the active surveillance data, and was possibly a more accurate representation of the lesional distribution of PKDL, at least in West Bengal (Fig 3B). As the patients enrolled during passive surveillance were from VL-endemic states of India and Bangladesh, it was relevant to examine if differences observed were attributable to their geographical location. Accordingly, they were sub-stratified into (i) West Bengal, n = 35 (ii) Bihar, n = 54 and (iii) other regions (Jharkhand Uttar Pradesh and Bangladesh), n = 11. Irrespective of the geographic location, the demographic profiles in all three groups was comparable, and hence were analyzed as a single group (S1 Table). In terms of their lesional profile, significant differences emerged with regard to the median age, disease duration and lag period. Following passive surveillance, the median age (in years) for polymorphic PKDL was significantly higher than the macular group, and this difference persisted with active surveillance (Table 2). The gender bias present during passive surveillance was not evident following active monitoring. Additionally, upon stratification, in terms of their lesional profile, the significantly longer disease duration and lag period present in the polymorphic variant was absent following active case detection (Table 2). Furthermore, significant differences emerged regarding their parasite load as with passive surveillance, it was 3. 0 fold higher in the polymorphic variant as compared to the macular variant, and this trend remained even with active surveillance, being 3. 2 fold higher (Table 2, Fig 4A). Based on the disease duration at the time of reporting, they were categorised into 4 groups, namely 0–6 months, 6–12 months, 1–2 years and >2 years. For the initial 0–6 and 6–12 months, both variants showed a comparable median parasite load, being 36458 (730–642578) vs. 24365 (958–504512) and 57785 (450–154632) vs. 35825 (489–65582) parasites/μg of genomic DNA respectively (Fig 4B). However, when the duration increased to 1–2 years, a 2. 3 fold higher parasite load was obtained in the polymorphic vs. macular variant being 35854 (984–85524) vs. 15547 (458–45785), p<0. 05 (Fig 4B). This trend persisted when the duration increased to >2 years, with the difference being 1. 9 fold, 28544 (657–55284) vs. 14652 (285–24658) parasites/ μg of genomic DNA), p<0. 05 (Fig 4B). Irrespective of the type of surveillance, differences existed in the lesional distribution pattern of polymorphic vs. macular PKDL. In polymorphic cases, lesions were asymmetrically distributed, varied from 8–12 in number and appeared mainly in sun exposed areas (Fig 5A). However, for macular PKDL, the distribution was diffuse, patchy, mostly symmetrical, with macules being sometimes large and coalescent (Fig 5B). In some patients, these patches developed initially in one area (e. g. upper limbs) as minute pin-point lesions and gradually spread to other areas (e. g. trunk). Based on the anatomical distribution, majority had lesions all over the body i. e. face, trunk and/or limbs followed by lesions on the face and neck, and least being on the face and trunk. Five patients reported mucosal involvement, where four had papules/nodules present on the lips, tongue, buccal mucosa and one on the glans penis [20]. As PKDL is proposed to be a SAG driven phenomenon [21], we examined the treatment profile during VL in 257/302 patients. Irrespective of the type of surveillance, the treatment modality for VL in the majority was SAG followed by Miltefosine, LAmB and herbal medicines (Table 3). The disease duration and lag period of cases who received SAG was comparable, being 4. 5 (3–8) vs. 3. 2 (2–6. 5) years and 4 (3–7) vs. 3. 7 (2. 2–6. 0) years respectively.
Based on the limited resources invested in diagnosis, treatment, and control, and its strong association with poverty, Leishmaniasis is classified as one of the most neglected diseases. Disease burden estimates place leishmaniasis second in mortality and fourth in morbidity among all tropical diseases with over 1 billion people in endemic areas still at the risk of contracting one or the other forms of the disease [22,23]. VL is a major public health importance in India, Bangladesh and Nepal and additionally, for reasons still unknown, a certain population of apparently cured VL patients go on to develop PKDL, whose clinical manifestations can be confused with other dermal disorders like leprosy, vitiligo, etc. The South Asian elimination initiative of VL and the 2012 London Declaration on neglected tropical diseases (NTDs) have raised global awareness about Leishmaniases and translated substantially into increased funding especially for control [1]. However, most of the classical challenges of a NTD persist and include limited therapeutic options, suboptimal diagnostics, and poor community awareness. Historically the occurrence of VL in India has shown a cyclical pattern with case resurgence characteristically occurring every 15 years [24], but the cause remained poorly defined. Modelling data provided valuable information regarding the transmission of VL and suggested the potential causes for resurgence are asymptomatic individuals and in particular, PKDL cases [25]. Importantly, as PKDL is the only inter-epidemic reservoir of anthroponotic VL, with proven transmission potential [26], existence of even a few cases can trigger a new epidemic of VL, reiterating the need for the early diagnosis and prompt treatment of every case of PKDL [11]. The implementation of the ongoing elimination programme unravelled a different clinico-demographic profile of PKDL patients in West Bengal. Prior to 2015, passive surveillance was the sole mode of identification, with suspected cases having clinical symptoms similar to PKDL, a past history of VL, and they self-reported at the dermatology outpatient departments of different hospitals in Kolkata. These patients (n = 100, Fig 1) were not restricted to West Bengal, but included the VL-endemic neighbouring states of Bihar and Jharkhand as also Bangladesh. A substantial proportion were inhabitants of districts in Bihar with high VL endemicity (Fig 2), and for professional commitments, resided in Kolkata. However, in spite of differences in their geographical origin, their demographic profiles were comparable and hence this study can be regarded as representing the overall disease scenario for PKDL, at least in India (S1 Table). From 2015 onwards with implementation of the elimination drive, active field surveys began in the VL/kala-azar hyper-endemic districts of West Bengal (Fig 2), and proved to be a pivotal game changer, as an undetected disease burden (n = 202) was identified in only three years vis-à-vis 100 detected by passive surveillance over 15 years (Fig 1), akin to a study in Bangladesh where a 30 fold increase was reported [27]. The proportion of positive cases was higher through passive vs. active surveillance being 98. 0 vs. 60. 5%, Fig 1, attributable to the lower clinical proficiency of the KTS (Kala-azar Technical Supervisors) or local health workers [28], who have been trained to refer a suspected PKDL case based on a previous history of kala-azar and/or characteristic clinical features of PKDL [29]. Furthermore, majority of the negative cases (70%) detected by active surveillance presented with hypopigmented patches, which pose a diagnostic dilemma even for a dermatologist, especially as these patients reported a past history of VL (Fig 1). As the diagnosis of PKDL at the field level is primarily dependent on clinical features, this is a practical problem, but it is preferable to err on the side of a positive diagnosis than leave the case neglected and allow it to be a mobile disease reservoir [4]. Following active surveillance, a sharp increase in the number of patients with macular PKDL emphasized the presence of this hidden disease burden, a sadly neglected component of a neglected tropical disease. Hypopigmented macules, especially on the face, have been reported in PKDL patients in Nepal and Bangladesh [8,27,30]. Till date, studies in India involving patients with PKDL, have always considered the polymorphic variant as the predominant form, ranging from 80–90% of the disease burden [7,18,31,32]. Similarly, our passive surveillance recorded 77% with polymorphic lesions while only 23% presented with macular patches (Fig 3B). On the other hand, active surveillance reported an overwhelming increase in macular cases, the ratio of polymorphic: macular becoming almost 1: 1. In view of the cumulative scenario of PKDL mimicking the active surveillance data indicating a 50: 50 distribution, it is possibly a better representation of the lesional distribution of PKDL in India (Fig 3B). Macular cases are less likely to seek treatment than those with disfiguring nodular and polymorphic lesions [33], Fig 5A and 5B. An ongoing debate regarding the role of macular cases in disease transmission was resolved by Molina et al. , 2017 [26] who in a proof-of-concept experiment established that both maculopapular and nodular PKDL lesions played a definitive role in transmission, thus countering the conventional belief that macular and papular forms pose a lesser threat than nodular PKDL. Furthermore, quantification of parasite load substantiated that macular PKDL harbour a considerable disease burden [Table 2,12]. Facility-based studies from South Asia have consistently reported a higher incidence of VL in males than females, the scenario being similar for PKDL [4,7, 18,19], possibly attributable to differences in care-seeking behaviour, males being accorded preferential treatment, females covering their lesions and therefore ignoring them, along with easier accessibility for males to reach healthcare facilities [34] whereas data from active surveillance radically differed (Fig 3A). Another notable feature secondary to active surveillance was the decrease in disease duration, possibly secondary to the awareness raised by the elimination drive. Contrary to previous studies where the mean disease duration was more than 5 years [8,18,31], active surveillance translated into an increase in the proportion of patients with a disease duration of less than a year (Table 1). All other features related to the disease profile remained unchanged, namely median age and the time interval between cure from VL and onset of PKDL (Table 1). Although the most definitive diagnostic approach in PKDL would be parasite detection in skin smears, it has an unacceptably low detection rate ranging from 4–58%. In VL, molecular monitoring of parasites has been effective in detecting asymptomatic VL [35] and monitoring treatment efficacy [35] whereas studies in PKDL were limited, primarily due to the limited number of cases and logistic limitations for follow up. The ITS-1 PCR and LAMP (loop mediated isothermal amplification) have been successfully employed for detection of parasite DNA [12,13,36,37]. However, studies pertaining to monitoring treatment effectiveness via qPCR remains limited wherein the efficacy of Miltefosine was confirmed whereas LAmB demonstrated poor efficacy [12]. For macular PKDL, their hypopigmented lesions take a considerable time to recede and therefore, monitoring the treatment efficacy of macular PKDL by clinical features is challenging. Accordingly, establishment of a ‘point of cure’ is particularly relevant [12,38]. In the present study, the parasite burden was substantially higher in patients self reporting as compared to active surveillance (Table 1, Fig 4A), attributable to the reluctance of PKDL patients to actively seek treatment. This translates into them harboring the parasite for an unacceptably longer duration and being ‘mobile disease reservoirs’, and a stumbling block for the elimination efforts. Overall, irrespective of the type of surveillance, the parasite burden showed a wide IQR from 918 to 47738, comparable with previous reports [37], and the polymorphic variant consistently showed a significantly higher parasite load vis-a-vis the macular group (Fig 4A, Table 2), corroborating with other studies [36,39]. During the disease time span of 0–12 months, no significant differences in parasite kinetics were observed between the two variants (Fig 4B). However, as the disease duration increased to 1–2 years and >2 years, significant differences appeared between the polymorphic and macular variant (Fig 4B), suggesting that differences in the host-parasite immune response may play a role in development of these two variants, emphasizing the need to perform longitudinal surveys with sequential assessment in a cohort of PKDL cases. In a retrospective cohort study in Nepal [30], inadequate SAG treatment for VL was proposed to increase the risk of subsequently developing PKDL [21]. Irrespective of the mode of recruitment, majority of individuals in this study had received SAG as treatment for VL (Table 3), as unlike Bihar, West Bengal was considered as a SAG-sensitive zone. Further studies are warranted as other contributory factors e. g. inadequate SAG treatment and a lack of compliance can also increase the risk of PKDL. The appearance of PKDL has been reported with all anti-leishmanial drugs [40,41], and even in patients with no history of VL. Using an amastigote-macrophage model, it was demonstrated that PKDL isolates were more tolerant towards miltefosine as compared to VL isolates [42], and testing of miltefosine susceptibility was recommended. In terms of the immune response generated by these drugs, SAG is more parasiticidal in nature whereas Miltefosine exerts both parasiticidal and immunomodulatory properties [43]. In 2012, WHO published a roadmap on neglected tropical diseases, including the regional elimination of VL in the Indian subcontinent by 2020 [44]. With PKDL being the disease reservoir, this study delineates the significant changes in the demographic profile of PKDL patients in West Bengal triggered by active surveillance, and endorses the benefits of active case detection in strengthening the Kala-azar elimination drive in South Asia. Most importantly, the overwhelming increase in the number of macular cases indicates that PKDL can no longer be considered as a singular entity. Furthermore, monitoring the treatment efficacy for macular PKDL is not straightforward as repigmentation of macular lesions occurs well after parasite clearance has been achieved. Therefore, in spite of the causative parasite species being L. donovani, these two variants of PKDL present with varied clinical features, suggesting possible differences in their host-parasite interactions, necessitating an in-depth analysis of the macular variant, a neglected component of this neglected tropical disease. | Post Kala-azar Dermal Leishmaniasis (PKDL) is a dermal condition that occurs in South Asia in 10–20% of patients after apparent cure from Visceral Leishmaniasis (VL). It presents with macular or papulonodular lesions that harbor Leishmania parasites, thus playing a major role in the transmission cycle. Accordingly, development of strategies for case finding, diagnosis, and treatment are major foci of the Leishmaniasis elimination programme. The major limitations include a risk of misdiagnosis, costly case finding along with inadequate and/or unnecessary treatment. This study delineated changes in the epidemiological characteristics of PKDL cases in West Bengal arising out of two modes of surveillance. Traditionally, passive surveillance was the major mode of patient recruitment, but from 2015, active surveillance came into effect. This led to the scenario changing drastically, with important differences appearing in the demographic profile of PKDL. Active surveillance proved to be a game changer as it unravelled a huge disease burden of macular cases, which had so far remained undetected. Therefore, contrary to previous studies, our study has established that macular PKDL in West Bengal, and perhaps India, is not a rare, uncommon presentation, and therefore warrants a deeper analysis. | Abstract
Introduction
Materials and methods
Results
Discussion | kala-azar
medicine and health sciences
pathology and laboratory medicine
tropical diseases
geographical locations
india
parasitic diseases
signs and symptoms
neglected tropical diseases
public and occupational health
infectious diseases
zoonoses
epidemiology
lesions
protozoan infections
people and places
diagnostic medicine
asia
leishmaniasis
disease surveillance | 2019 | Active surveillance identified a neglected burden of macular cases of Post Kala-azar Dermal Leishmaniasis in West Bengal | 7,243 | 287 |
Wolbachia infections confer protection for their insect hosts against a range of pathogens including bacteria, viruses, nematodes and the malaria parasite. A single mechanism that might explain this broad-based pathogen protection is immune priming, in which the presence of the symbiont upregulates the basal immune response, preparing the insect to defend against subsequent pathogen infection. A study that compared natural Wolbachia infections in Drosophila melanogaster with the mosquito vector Aedes aegypti artificially transinfected with the same strains has suggested that innate immune priming may only occur in recent host-Wolbachia associations. This same study also revealed that while immune priming may play a role in viral protection it cannot explain the entirety of the effect. Here we assess whether the level of innate immune priming induced by different Wolbachia strains in A. aegypti is correlated with the degree of protection conferred against bacterial pathogens. We show that Wolbachia strains wMel and wMelPop, currently being tested for field release for dengue biocontrol, differ in their protective abilities. The wMelPop strain provides stronger, more broad-based protection than wMel, and this is likely explained by both the higher induction of immune gene expression and the strain-specific activation of particular genes. We also show that Wolbachia densities themselves decline during pathogen infection, likely as a result of the immune induction. This work shows a correlation between innate immune priming and bacterial protection phenotypes. The ability of the Toll pathway, melanisation and antimicrobial peptides to enhance viral protection or to provide the basis of malaria protection should be further explored in the context of this two-strain comparison. This work raises the questions of whether Wolbachia may improve the ability of wild mosquitoes to survive pathogen infection or alter the natural composition of gut flora, and thus have broader consequences for host fitness.
Wolbachia pipientis is a maternally inherited intracellular bacterium that is found in a wide range of arthropod species and filarial nematodes, with approximately 40% of insect species infected [1]. Wolbachia spreads rapidly through populations and to high frequencies by inducing a range of manipulations of host reproduction that benefit infected females. In insects, the most common manipulation is cytoplasmic incompatibility (CI) [2], [3]. Interestingly, some Wolbachia strains that cannot induce reproductive manipulations still spread through populations [4]. This would not be predicted unless there were other positive benefits for Wolbachia-infected insects. Despite numerous laboratory and semi-field based experiments examining a range of life history traits, few studies have identified significant fitness benefits of infection [5], [6], [7], [8]. Most reveal no effect [9] or weak negative effects [10], [11], [12]. It is possible, however, that there are benefits to Wolbachia infections that are only detectable under field conditions or in circumstances not yet tested in the laboratory. Recently, Wolbachia was found to either extend the lifespan and/or increase the survival of Drosophila infected with native viruses, a trait termed pathogen protection [13], [14], [15]. Subsequently, Wolbachia strains native to Drosophila have also been shown to confer pathogen protection against arboviruses, bacteria, filarial nematodes and the malaria parasite Plasmodium gallinaceum when stably transinfected into the mosquito Aedes aegypti [16], [17] [13], [17], [18], [19], [20], [21], [22]. This broad-based pathogen protection may offer a potential fitness advantage, assisting cytoplasmic incompatibility in the maintenance and spread of Wolbachia in wild populations. Understanding the true fitness effects of Wolbachia infections in mosquitoes is important as these symbiont-infected mosquitoes are being released into wild populations as part of a biocontrol strategy for reducing dengue virus transmission to humans [23]. While the mechanism of pathogen protection is not fully understood, several recent studies have shed some light on its basis. It was originally hypothesized that priming of the insect immune response might provide a single mechanistic explanation for symbiont-induced protection against viruses, bacteria, nematodes and malaria. Under an immune priming model, Wolbachia infections activate the basal immune response, better preparing insects against subsequent infection by pathogens. Three different A. aegypti: Wolbachia strain associations have been created thus far and in each case infection with the symbiont induces the host immune response [18], [19], [20]. The same is true for transient infections established in the mosquito Anopheles gambiae [22]. Immunity genes upregulated in these mosquitoes include members of the opsonisation, Toll and melanisation pathways [18], [21], [24]. Whether the expression of this limited set of insect immunity genes can confer protection against pathogens other than bacteria [25] is not clear, although the Toll pathway participates in dengue virus control [26] and the Imd, Toll, opsonisation and melanisation pathways assist in Plasmodium control [18], [27], [28], [29], [30]. In each case where Wolbachia-associated immunity gene activation has been reported, the host insects did not have histories of association with this symbiont. In Drosophila naturally infected with Wolbachia there is no activation of the immune response and no bacterial protection [31], [32]. There is, however, weak protection against dengue virus [24] as well as other native viruses [13], [14] indicating that innate immune priming cannot explain viral protection in this host. Interestingly, in A. aegypti transinfected with the same Wolbachia strains native to D. melanogaster, there is both innate immune priming and strong protection against dengue virus. The comparative study indicates that innate immune priming alone cannot fully explain pathogen blocking although it may be contributing to the strength of the effect in A. aegypti [24], [33]. This same study also revealed that Wolbachia strains differ in their level of immune induction in A. aegypti [24]. The wMelPop Wolbachia strain, known for causing life shortening and other fitness effects in its host, is present in more tissues and grows to higher densities [16], [20], [34], [35] and is associated with a greater immune response than the wMel strain, which is present in fewer tissues and grows to much more moderate densities [17]. While the transcriptional profiles of Wolbachia-infected A. aegypti predict that they should experience broad protection against bacterial infection, evidence of bacterial protection in this host comes from a single study demonstrating the ability of wMelPop to protect against systemic Erwinia carotorovra infection [21]. Here we expose A. aegypti stably transinfected with either the wMel or the wMelPop strain to infection with several bacterial pathogens using in previous infection studies in D. melanogaster. We characterised the response to two extracellular bacteria, E. carotovora [36] and the slow-killing but highly pathogenic Burkholderia cepacia [37], and two intracellular bacteria, Salmonella typhimurium [37], [38] and the slow-growing Mycobacterium marinum [37]. Following pathogen infection we then examined mosquito survival and corresponding changes in Wolbachia and pathogen densities. As a control, we also confirm that these Wolbachia strains provide no protection against these same pathogens in D. melanogaster. We studied that the protective effect of wMelPop and wMel in terms of both survival and delayed death rates. We examined the association between survivorship and pathogen load. Our result indicates either a direct effect of immune priming on the symbiont or an energetic tradeoff, with sick hosts affecting resources for Wolbachia' s growth and replication.
Approval for blood feeding by human volunteers for maintenance of the mosquito colony was granted by the Monash University Human Research Ethics Committee (2007001379). Volunteers provided written informed consent to participate. The w1118 fly line infected with wMel (w1118wMel) or wMelPop (w1118wMelPop) and their respective tetracycline-cured lines (w1118wMel. tet and w1118wMelPop. tet respectively) were used in this study [34], [39]. PCR using primers specific for the wMel and wMelPop IS5 repeat was used to confirm the tetracycline-cured lines to be free of Wolbachia [40]. Flies were reared on standard yellow corn meal medium at 25°C with 50% relative humidity and a 12: 12 hr light/dark cycle. Around fifty individuals were allowed to oviposit in bottles with 40 ml of fly food for two days. After eclosion, adults were transferred to and aged in vials at a density of ∼30 individuals per vial. Mosquito lines used in this study are laboratory lines artificially infected with wMel (MGYP2) or wMelPop-CLA (PGYP1) and their tetracycline-cured (PGYP1. tet and MGYP2. tet respectively) Wolbachia uninfected counterparts [16], [17]. Mosquitoes were reared under standard laboratory conditions (26±2°C, 12: 12 hr light/dark cycle, 75% relative humidity). Mosquito larvae were fed 0. 1 mg/larvae of TetraMin Tropical Tablets once a day at a density of 150 larvae per 3 liters of distilled water in trays. Adults were transferred to cages (measuring 30×30×30 cm) at emergence at 400 individuals per cage. Adults were supplied with a basic diet of 10% sucrose solution. E. carotovora strain 15 (Ecc15) and S. typhimurium strain TM11 were cultured in LB medium in a shaker at 37°C [36], [37]. B. cepacia clinical isolate AH1345 was cultured in brain heart infusion broth (Oxoid, Australia) at 37°C in a shaker [37], [38]. M. marinum was cultured at 29°C in the dark without shaking in Middlebrook 7H9 broth (Difco, Australia) supplemented with OADC [41]. For survival assay, female flies and mosquitoes aged 3–8 days were used. Insects were anesthetized with CO2 before being infected by either stabbing with a needle previously dipped into a bacterial culture or injected with 69 nl via an individually calibrated pulled glass needle attached to a Nanoject II injector (Drummond Scientific Company, Broomall). PBS mock stabbed or injected insects were used as control for the infection processes. For E. carotovora and M. marinum infection, bacterial cultures were pelleted (OD∼20). Flies were infected by injection in the abdomen and mosquitoes were infected by pricking the thorax [36], [42]. For B. cepacia infection, flies and mosquitoes were infected by pricking in the thorax from a bacterial culture of OD = 0. 1 measured spectrometrically at 600 nm [37]. For S. typhimurium infection, bacterial culture of OD of 0. 1 at 600 nm was injected into flies and infection in mosquitoes was achieved by pricking in the thorax [37], [38]. Survival data were collected over the entirety of the insect' s life, however, only the first 200 hours post infection were used for analysis (when over 90% of death had occurred) prior to the onset of shortening effects of wMelPop. Survival curves were analyzed using Kaplan-Meier analysis, and log-rank statistics (SPSS statistics version 19, SPSS Inc, an IBM Company) were corrected for false positives using q-value [43]. We used qPCR to quantify bacterial density as it is a more sensitive and specific way to estimate bacterial number than plating for bacterial growth, especially for bacteria that are difficult to culture [44]. Specific primers (Table 1) were designed for the bacterial 16S ribosomal RNA gene of each of the bacterial pathogens using Primer3 [45]. For Wolbachia previously published primers for the single copy ankyrin gene WD0550 were employed [46]. Bacterial gene copy numbers were expressed as a ratio by normalizing against copy numbers for the host rpS17 [47] gene (Table 1). To correct for potential differences in body size between different mosquito lines that would affect host rpS17 copy number, the change in bacterial density was expressed as the fold increase of 16S/rpS17 ratio post-infection to the 16S/rpS17 ratio immediately after infection (zero hour post-infection). Post-infection mosquitoes were collected at either 8 or 26 hours when ∼10% of the individuals had died. Five pairs of females were used for each bacterial strain. DNA was extracted from individual mosquitoes using DNeasy spin columns (QIAGEN, Australia) and qPCR was performed on LightCycler 480 (Roche Applied Science, Australia) using PlatinumSYBRGreen (Invitrogen Inc, Carlsbad, CA) according to manufacturer' s instructions. For each reaction a mastermix of 2 µl RNase-free water, 5 µl of SYBR Supermix and 0. 5 µl of each primer (5 µM) was added to 2 µl of DNA. The cycling protocol was as follows: 1 cycle Taq activation at 95°C for 2 minutes, 40 cycles of denaturation at 95°C for 5 s, annealing at 60°C for 5 s, extension at 72°C for 15 s, fluorescence acquisition 78°C, and 1 cycle of melt curve analysis from 68–95°C in 1°C steps. A standard curve was constructed using serially diluted DNA to calculate the amplification efficiency of each set of primers. The raw output data of crossing points (CP) was normalized by taking into consideration the differences in amplification efficiency of target and the reference genes using Q-Gene [48]. Scatter plot with median ± interquartile range were plotted. Treatment effects were then examined using Mann-Whitney U tests using Statistica 8. 0 (StatSoft, Inc.).
We tested whether w1118wMel and w1118wMelPop fly lines were protected against either extracellular or intracellular bacterial infection by comparing their survival to that of their tetracycline-cured counterparts. The pathogens varied in their virulence as measured by how quickly they killed flies. Almost all the flies infected with E. carotovora and S. typhimurium were dead within 24 hours, whereas those infected with B. cepacia and M. marinum survived for several days. There was no significant difference in survival, however, between w1118wMel and w1118wMel. tet or between w1118wMelPop and w1118wMelPop. tet for any of the pathogens tested (Figure 1A–H, Table S1A). We examined mosquitoes infected with wMel (MGYP2) or wMelPop-CLA (PGYP1) for protection against the four bacterial strains. After demonstrating no significant difference in survival between Wolbachia-infected and uninfected mosquitoes when injected with PBS (Table S1B), direct comparisons were then made between Wolbachia-infected vs uninfected mosquitoes in the presence of each of the pathogens. Infection with wMelPop-CLA provided protection against all four pathogens, but wMel protected only against E. carotovora and S. typhimurium (Fig. 2, Table S1B). For these two pathogens, the relative risk ratios (risk of dying for Wolbachia-uninfected relative to Wolbachia-infected individuals in the presence of the pathogen) were also greater for PGYP1 compared to MYGYP2 (Fig. 2, A vs E, C vs G) although only significantly so for S. typhimurium (Table S1B). The nature of the protection when present also varied. In response to E. carotovora (Fig. 2A & E), Wolbachia conferred both a delay in death (lines not parallel) and an increase in survival from 0% to 28% and 0% to 50% for wMel and wMelPop-CLA, respectively. The wMel strain only delayed death for S. typhimurium-infected mosquitoes (Fig. 2C) and the same was true for wMelPop-CLA mosquitoes infected with B. cepacia (Fig. 2F). The wMelPop-CLA strain delayed death and increased survival from 0 to 10% for S. typhimurium infected mosquitoes at 200 hours post-infection and from 38% to 79% for those infected with M. marinum at 287 hours post-infection (Fig. 2 G&H). Taken together these patterns demonstrate that, compared to wMel, wMelPop-CLA offers mosquitoes protection against a broader range of pathogens, greater strength of protection, and is more likely to provide increased survival rather than simply delaying death. To investigate if co infection with Wolbachia could limit pathogen replication, we used qPCR to quantify the change in the bacterial density during early infection. For the extracellular bacteria E. carotovora (Figure 3A) and B. cepacia (Figure 3B), both wMel and wMelPop-CLA infected mosquitoes were able to inhibit the bacteria, with pathogen densities significantly higher in Wolbachia-uninfected counterparts relative to infected. This difference in pathogen density appears to be correlated with increased survival and reduced death rate due to E. carotovora (Fig. 2 A&E) but less so for B. cepacia (Fig. 2B & F). In contrast, only wMelPop-CLA infection results in reduced densities of the two intracellular pathogens (Fig. 3 C&D). For both S. typhimurium and M. marinum, as for E. carotovora, reduction in the proliferation of intracellular bacteria correlates with significant delays in mosquito death and increases in survival (Fig. 2 C–D, G–H). The magnitude of the reduction in pathogen density due in association with Wolbachia was also more modest for intracellular bacteria (∼3–4 fold) than for extracellular infections (∼30–7000 fold). To investigate whether the presence of pathogenic bacteria could affect the replication and/or survival of Wolbachia, we used qPCR to quantify the change in Wolbachia density during the first 8 hours of infection. In most cases, Wolbachia densities were significantly reduced during the early hours of infection with a pathogen regardless of Wolbachia strain (Fig. 4). Fold reductions were similar across all pathogen x Wolbachia strain pairings, ranging from 1. 5–2. 7. Only wMelPop-CLA in response to S. typhimurium and wMel in response to M. marinum did not experience statistically significant reductions (Fig. 4 C & D), although the median Wolbachia densities demonstrate decreasing trends.
Our findings support previous studies indicating that native Wolbachia infections in D. melanogaster do not confer pathogen protection against bacteria. In the recently transinfected A. aegypti, in contrast, we demonstrate pathogen protection that varies by strain, with wMelPop-CLA exhibiting more effective protection than wMel against a broader range of bacteria. We also provide evidence that the expression of innate immunity genes induced by Wolbachia infection in mosquitoes likely explains these differences in protection. Future work will need to identify the potential role for innate immune priming as an enhancer of viral protection, assess whether bacterial protection is providing benefit for mosquitoes in the field. These findings may assist with Wolbachia strain selection for field release. | Wolbachia is a commonly occurring bacterium or symbiont that lives inside the cells of insects. Recently, Wolbachia was artificially introduced into the mosquito vector dengue virus that was naturally Wolbachia-free. Wolbachia limits the growth of a range of pathogens transmitted to humans, including viruses, bacteria and parasites inside the mosquito. This “pathogen protection” forms the basis of field trials to determine if releasing Wolbachia into wild mosquito populations could reduce dengue virus incidence in humans. The basis of pathogen protection is not fully understood. Previous work suggests that the symbiont may activate the basal immune response, preparing the insect to defend itself against subsequent pathogen infection. Here we infect mosquitoes harbouring Wolbachia with a range of bacterial pathogens as a means to understand the nature of protection. We show that different Wolbachia strains vary in their ability to limit pathogen growth and that this correlates with the degree to which the Wolbachia activates the host immune response. These findings may assist with Wolbachia strain selection for future open field release and raise the question whether Wolbachia might provide a fitness advantage to mosquitoes in the wild by limiting their death due to bacterial infection. | Abstract
Introduction
Materials and Methods
Results
Discussion | zoology
genetics
biology
microbiology | 2013 | Wolbachia-Associated Bacterial Protection in the Mosquito Aedes aegypti | 5,013 | 296 |
Profiling of DNA and histone modifications has recently allowed the establishment of reference epigenomes from several model organisms. This identified a major chromatin state for active genes that contains monoubiquitinated H2B (H2Bub), a mark linked to transcription elongation. However, assessment of dynamic chromatin changes during the reprogramming of gene expression in response to extrinsic or developmental signals has been more difficult. Here we used the major developmental switch that Arabidopsis thaliana plants undergo upon their initial perception of light, known as photomorphogenesis, as a paradigm to assess spatial and temporal dynamics of monoubiquitinated H2B (H2Bub) and its impact on transcriptional responses. The process involves rapid and extensive transcriptional reprogramming and represents a developmental window well suited to studying cell division–independent chromatin changes. Genome-wide H2Bub distribution was determined together with transcriptome profiles at three time points during early photomorphogenesis. This revealed de novo marking of 177 genes upon the first hour of illumination, illustrating the dynamic nature of H2Bub enrichment in a genomic context. Gene upregulation was associated with H2Bub enrichment, while H2Bub levels generally remained stable during gene downregulation. We further report that H2Bub influences the modulation of gene expression, as both gene up- and downregulation were globally weaker in hub1 mutant plants that lack H2Bub. H2Bub-dependent regulation notably impacted genes with fast and transient light induction, and several circadian clock components whose mRNA levels are tightly regulated by sharp oscillations. Based on these findings, we propose that H2B monoubiquitination is part of a transcription-coupled, chromatin-based mechanism to rapidly modulate gene expression.
To assess the contribution of chromatin state variations to development and phenotypic plasticity, evaluation of the role of histone post-translational modifications in the regulation of genome expression dynamics represents an important objective. A first step requires the establishment of a reference epigenome through the profiling of representative chromatin proteins and histone modifications in standard growth conditions [1]. This approach recently revealed simple organization principles based on 4 or 5 major chromatin states with distinct functional properties in Drosophila cells [2], in Caenorhabditis elegans embryos [3] and in the model plant species Arabidopsis thaliana [4]. One such chromatin signature associates with active genes and combines several histone modifications, notably histone H3 trimethylated on lysine 4 and/or lysine 36 (H3K4me3, H3K36me3), as well as monoubiquitinated histone H2B (H2Bub). These chromatin marks have the potential to influence transcriptional activity and, hypothetically at least, to be maintained through mitosis and/or meiosis [5], [6]. However, the assessment of dynamic chromatin changes has been more difficult because of the confounding effects of cell division and tissue specificity. Functional analyses in S. cerevisiae showed that a transcription-coupled cyclic process involves the monoubiquitination of histone H2B by the Bre1 ubiquitin ligase and subsequent deubiquitination by SAGA, a complex that combines the two histone-modifying activities of Ubiquitin protease 8 (Ubp8) and GCN5 acetyltransferase [7], [8], [9], [10]. The SAGA evolutionarily conserved complex acts as a transcriptional coactivator that promotes gene expression at a post-initiation step in metazoans [11], [12]. Indeed, H2Bub was found to facilitate the processivity of RNA Pol II through nucleosomes by affecting DNA accessibility, to help recruit the histone chaperone FACT (FAcilitates Chromatin Transcription) and to ensure nucleosome reassembly [13], [14], [15], [16], [17]. In both yeast and mammals, the Polymerase-associated factor 1 complex (Paf1C) serves as a platform for the monoubiquitination of H2B during transcription elongation, which in turn induces the trimethylation of histone H3 on lysines 4 and 79 by COMPASS/MLL complexes in a so-called trans-histone crosstalk [18], [19]. Accordingly, Paf1C and the Set1 and Set2 methyltransferases that catalyse H3K4me3 and H3K36me3 deposition, respectively, have been found to associate with the elongating form of RNA Pol II (reviewed in [20], [21], [22]). Altogether, an emerging picture is that transcriptional coactivators can increase RNA Polymerase II activity by modulating H2Bub homeostasis and coordinating several other histone modifications, thereby contributing to the selective regulation of cellular pathways [11], [12]. Nevertheless, much remains to be understood about the dynamic changes of histone modifications and their impact on gene expression in response to developmental or environmental signals. Chromatin-based regulatory processes play important roles during plant developmental transitions (reviewed in [23], [24], [25], [26]), and in particular in response to light signals for the establishment of photomorphogenic development [27], . When dark-grown (etiolated) seedlings emerge from the soil, the initial light perception event promotes a major developmental switch that orchestrates a massive reprogramming of gene expression through which heterotrophic seedlings become photosynthetically competent and can complete their life cycle [29], [30], [31]. This rapid and division-independent developmental window is therefore especially well suited for the study of chromatin state dynamics over a large repertoire of genes, many of which undergo pioneering rounds of transcription upon light perception. Evidence has indeed emerged that photomorphogenesis involves chromatin modifications. Profiling of the antagonistic histone H3 modifications K9ac/me3 and K27ac/me3 during de-etiolation showed that gene upregulation associates with histone acetylation, and reciprocally, that some light-repressed genes gain H3K27me3, a mark of Polycomb Group-mediated repressive activity [32]. Additionally, functional approaches using plant mutants for the evolutionarily conserved GCN5 and HD1 factors affected in the acetylation/deacetylation of several histone H3 and H4 residues further support a model in which histone modifications may contribute to maintain genes in a repressed state in darkness and subsequently modulate their activity upon illumination [33], [34], [35]. In the current study, the early events of photomorphogenesis were used as a paradigm to investigate spatial and temporal dynamics of chromatin states. We focused on H2Bub because of its link with transcriptional activation in yeast and metazoans. In Arabidopsis, canonical histone H2B proteins are monoubiquitinated on a lysine residue at positions 143 or 145 depending on the sequence [36] by the heterodimeric HUB1/HUB2 E3 ubiquitin ligase, a homolog of the budding yeast Bre1 protein [37], [38]. The hub1 mutants represent unique tools to assess the consequence of H2Bub loss in plants. In particular, null alleles of the genes encoding the histone H2B E2/E3 ubiquitin ligase represent the only Arabidopsis mutants with near-to-normal phenotypes in which modification of just one histone residue is abrogated [38], [39], [40], [41]. By integrating the light-induced transcriptional responses in wild-type and hub1-3 mutant seedlings with the changes in genome-wide distributions of H2Bub over a 6 h period of exposure to light, we here assess how H2Bub influences the rapid regulation of gene expression.
Five-day-old Arabidopsis seedlings were grown in complete darkness or exposed to light for 1 or 6 h before being harvested for RNA and chromatin extractions (Figure 1A). This temporal window allows plants to shift from a fast response mode at 1 h to slower, more selective, responses at 6 h but is not long enough for cell division to occur [42]. RT-qPCR analyses of known light-responsive genes confirmed that these two time points could differentiate between early and late regulated genes in our conditions (Figure S1). Except for an opening of the apical hook, no morphological change was visible after 1 h of illumination (Figure 1B). After 6 h, most plants had open cotyledons and had initiated greening. In agreement with previous studies [43], the photomorphogenic switch was irreversible after 6 h of illumination but not after 1 h, as tested by transferring the plants back to darkness for an additional 24 h period (Figure 1B). In these conditions, the hub1-3 mutant did not exhibit significant morphological defects, except for a frequent lack of apical hook in darkness (45% of the mutant compared with 11% of the wild-type seedlings). Like for other hub1 null alleles, no H2Bub is detectable in hub1-3 chromatin extracts by immunoblot analysis (). ChIP-chip analyses were carried out at the 3 time points, hereafter called Dark (D), 1 h and 6 h. In keeping with our previous findings using light-grown seedlings [4], H2Bub was located almost exclusively over expressed genes (Figure S3A and S3B). The distribution of H2Bub over genic elements is low on promoter regions and typically resembles a Gaussian curve that peaks in the central part of the transcribed region (Figure S3C and S3D). This property allowed us to assign specific criteria for defining genes marked by H2Bub based on the central 40% of a CDS overlapping H2Bub-enriched domains (see supplementary methods in Text S1). Altogether, this first analysis revealed a common set of 4023 genes marked in all three conditions (Table S1). It also allowed us to identify 256 genes defined as being marked only in darkness and 566 genes only upon light exposure, among which 396 at 1 h and 429 at 6 h (Figure 1C). The H2Bub profiles of three representative genes marked at 1 h and 6 h (TZP) or only at 6 h (RBFa and HCF107) are shown in Figure 1E. To investigate more quantitatively the potential variations of H2Bub levels on these genes, we then examined differential enrichment using a so-called TileMap approach based on Hidden Markov Modelling (HMM; see Text S1). This approach determined genomic domains with differential H2Bub enrichment between the three conditions, which were then mapped to genes and further compared to the list of H2Bub-marked genes defined previously. The combination of these two analyses showed that 328 genes gain H2Bub de novo upon illumination, among which 177 at 1 h and 272 at 6 h (Figure 1D, Figure S4 and Table S1). In contrast, H2Bub was lost from only 4 genes at 1 h and from 54 genes at 6 h. A selection of genes in each category was then validated by ChIP-qPCR performed with anti-H2Bub and anti-H3 antibodies. This confirmed that potential variations in nucleosome occupancy did not account for differential H2Bub enrichments (Figure S5). Altogether, these data illustrate the dynamic nature of H2Bub deposition over a specific set of genes. To assess whether H2Bub variations affect light-regulated genes, the epigenomic data were integrated with transcriptome analyses. Of the ∼20,000 genes represented on CATMA microarrays [44], 695 and 1537 genes were differentially expressed at 1 and 6 h relative to the dark point, respectively, indicating that light had rapid and major effects on gene expression (Table S2). Comparison with H2Bub marking in the three conditions further showed that 49% of the light-induced genes gain the H2Bub mark, a fraction that is far above the average for all H2Bub-marked genes considered together (around 10%; Figure 2A). More globally, a scatterplot correlating changes in RNA levels with changes in H2Bub enrichment showed that light-upregulated genes tend to gain H2Bub upon illumination (r = 0. 25;). In contrast, downregulation did not appreciably correlate with H2Bub (r = 0. 09) and only a minor fraction of the downregulated genes showed a loss of the H2Bub mark (Figure 2A right panel). The link between H2Bub gain and gene induction was therefore explored more quantitatively. First, mean H2Bub levels were calculated for genes in each category of light-regulation. This revealed that upregulated genes displayed increased H2Bub levels at 1 h and at 6 h compared to dark and compared to downregulated genes (Figure 2B). The distribution of H2Bub levels over light-regulated genes was then examined by plotting enrichment levels along gene length. To avoid confounding effects, the analysis was restricted to genes that were marked under the relevant condition (marked in dark for downregulation and at 6 h for upregulation). This showed that upregulated genes progressively gain H2Bub levels along their transcribed region (Figure 2C). In contrast to upregulation, these analyses identified no correlation between gene downregulation and H2Bub loss. First, the fraction of genes that lose the H2Bub mark within 6 h is very low (n = 55). Second, downregulated genes showed no concomitant decrease in H2Bub level (Figure 2B and 2D). Finally, we tested individually by ChIP-qPCR the genes with the best predicted H2Bub decrease and found that they exhibited only a slight reduction in H2Bub (Figure S5). We conclude from this data that gene upregulation is usually associated with local H2Bub enrichment, while H2Bub domains tend to persist during downregulation. Having determined the set of genes subjected to variations in H2Bub upon illumination, we further tested whether they are also subject to other chromatin changes. We first examined in silico whether the genes that gain/lose H2Bub display similar trends for acetylation or trimethylation of histone H3 on lysine 9 and 27 using the data from Charron et al. [32]. This revealed that genes that gain H2Bub during de-etiolation frequently gain H3K9ac and/or H3K27ac and that, reciprocally, H2Bub loss associates with H3K27ac loss and with H3K27me3 gain (Table S3). These observations are in good agreement with the proposed role for H3K9ac and H3K27ac in light-induced gene expression [27], [32], [34], and suggest that some genes may lose H2Bub and H3K27ac to acquire Polycomb-associated marks for repression by light. More specifically, H2Bub deposition is a prerequisite for a trans-histone crosstalk triggering the trimethylation of H3 at K4 and K79 in S. cerevisiae and mammals [21], [22]. We therefore monitored H3K4me3 enrichment along four light-induced genes (HCF173, TZP SPA1 and GI) that gain H2Bub upon illumination and that are marked by H3K4me3 at later stages of seedling development [4]. The HCF173, TZP and SPA1 genes display low levels of mRNA and of H2Bub before illumination. ChIP-qPCR analyses revealed that H3K4me3 was enriched on the 5′ part of the transcribed region of these three genes following 6 h of light exposure in both wild-type and hub1-3 seedlings (Figure 3). We concluded from these data that transcriptional activation can associate to H3K4me3 enrichment in the absence of H2Bub on these genes. We also probed H3K36me3 on the same genes. This mark is globally associated with H2Bub and H3K4me3 along the Arabidopsis genome [4] and might serve an equivalent function to H3K79me3, which is not detectable in plants [25], [45]. After 6 h of illumination, no clear gain of H3K36me3 was detected on these genes. Only on the SPA1 5′ region was a slight enrichment reproducibly detected in both wt and hub1-3 plants (Figure 3E), altogether suggesting that, in this context, H3K36me3 levels display weak variations. As shown above, for many light-responsive genes we observed that upregulation was associated with local H2Bub enrichment. To address the possible involvement of H2Bub in the modulation of gene expression, we conducted transcriptome analyses of wild-type and hub1-3 mutant seedlings during the early hours of de-etiolation. We first compared wild-type and hub1-3 seedlings directly at the dark time point. This showed that 715 genes were already significantly misregulated in the mutant prior to light exposure, ∼80% of them being downregulated (Table S2). Although many indirect effects might confound this analysis, such a proportion is in general agreement with the predicted role of H2Bub in promoting transcription. Strikingly, many of these genes correspond to light-repressed genes: 42% and 23% of the genes downregulated at 1 and 6 h in the wild-type, respectively, were already down in hub1-3 before light exposure (Figure 4A). In contrast, a smaller set of 151 genes were upregulated in hub1-3 mutant seedlings, possibly through direct effects or through regulatory cascades. Their overlap with the light-induced genes was low (8% and 3% at 1 and 6 h, respectively). The hub1-3 mutation therefore partially mimics the effect of light for gene downregulation but not for upregulation, suggesting that for many genes, HUB1 contributes to attain high expression levels in darkness. Nonetheless, altered gene expression patterns in hub1-3 seedlings were not sufficient to trigger constitutive photomorphogenic development beyond a tendency towards opened apical hooks (Figure 1B). Interestingly, comparison of the hub1-3 transcriptome in darkness with more than 4,000 other A. thaliana transcriptome patterns using Genevestigator [46] identified csn5 and other csn mutant profiles as being the most similar to hub1-3 downregulated genes (Figure S7). CSN5 (COP9-signalosome Subunit 5) is the plant homolog of human Jab1 and forms part of the highly conserved CSN complex. By regulating ubiquitin-ligase activity of Cullin4 [47], CSN5 has important roles during plant development, notably for cell cycle progression and for repression of photomorphogenesis in darkness [48], [49]. Given the lack of CSN5 misregulation in hub1-3 mutants in darkness, we can rule out a possible direct regulatory effect and propose instead that the CSN signalosome and the HUB1 pathways may interconnect to regulate common genes in plants, as was suggested in yeast [21]. We then examined the effect of the hub1-3 mutation on gene expression kinetics during de-etiolation. Overall, the majority of light-regulated genes still responded to light in the hub1-3 mutant (Table S2). Notwithstanding, clustering of gene expression data by Self Organizing Map partitioning (SOM; [50]) identified a group of 790 genes with a tendency for upregulation that was reduced in the hub1-3 mutant (class 1; Figure 4B and Figure S8), and a group of 295 downregulated genes with weaker repression in hub1-3 (class 4). Such defects were confirmed by RT-qPCR analysis of some individual genes (Figure 4C). Light-driven expression changes of these genes were reproducibly impacted, some of them dramatically (e. g. , LHCA1, CRB, GUN5). Considering this trend, we compared directly the expression changes of the light-induced and -repressed gene sets. At 6 h, both gene sets displayed reduced expression changes in the hub1-3 mutant, with roughly half of the genes being less responsive in hub1-3 than in wild-type seedlings (Figure 5A and 5B). Because H2Bub preferentially associates with long genes [4], each set was then dissected according to gene length. This revealed that for upregulation, but not for downregulation, the longest genes were the most sensitive to the lack of HUB1. Taken together, these data suggest that H2Bub is not required for determining on/off gene activation states but rather contributes to attain particular levels of expression. To better delineate the effect of H2Bub loss on the modulation of gene expression, we therefore focused on genes that undergo rapid changes in mRNA levels. The genes that are transiently induced at 1 h and subsequently downregulated at 6 h were selected (“Up&Down” genes; n = 162). In hub1-3 seedlings, many of these genes displayed a weak induction at 1 h followed by a weak downregulation at 6 h (Figure 5C). In several cases, reduced downregulation even resulted in higher mRNA steady state levels in hub1-3 mutants than in wild-type seedlings at 6 h. As exemplified for the AT1G07400 gene in this class, these analyses indicate that HUB1 positively influences rapid variations in gene expression (Figure 5D). Given the defective transcriptional responses to light in hub1-3 seedlings, we examined the phenotype of these plants upon prolonged illumination. This revealed that a fraction of the hub1-3 seedlings were overly light sensitive (Figure 6A). This phenomenon known as photobleaching has been observed in several photomorphogenic mutants such as det1-1 and pifs that overaccumulate chlorophyll precursors in darkness [51], [52]. Although less penetrant in hub1-3 than in det1-1 mutants, this sensitivity was significant when seedlings were grown for 3 days or more in darkness before transfer to light (Figure 6B). Many misregulated light-responsive genes in hub1-3 mutants in darkness and during photomorphogenesis might be responsible for the photobleaching. One gene possibly responsible for the light sensitivity is POR-A, an essential regulatory gene in the chlorophyll biosynthetic pathway [53] that is significantly underexpressed in hub1-3 seedlings prior to light exposure (Figure 4C). We further analyzed the response of light-regulated genes in 2-day-old seedlings, an early developmental stage at which hub1-3 mutant seedlings do not photobleach (Figure 6A). RT-qPCR analyses revealed that expression changes of all tested genes was affected in 2-day-old hub1-3 seedlings to an extent that was similar or even more severe than in 5-day-old plants (Figure 6C), thus confirming that the hub1-3 mutation has a primary effect on gene expression. To investigate which light-regulated genes and pathways might be subject to H2Bub-mediated transcriptional regulation, we first analyzed genes known to be involved in photoperception and light signal transduction events (Table S4). Most photoreceptor genes were marked by H2Bub but none were misregulated in hub1-3 mutants. This included Phytochrome A (phyA), encoding the major photoreceptor for far-red light signaling during photomorphogenesis, which is rapidly downregulated but retains H2Bub upon illumination (Table S4). In contrast to the photoreceptors, only a few photomorphogenic regulatory factors were detected as being marked by H2Bub, and their expression was also usually not affected in the hub1-3 mutant. We then conducted an unbiased search by selecting genes that (1) are induced by light, (2) concomitantly display an enrichment of H2Bub, and (3) whose upregulation is affected in hub1-3 plants (Table S5). These restrictive criteria identified 90 genes, many of which encode plastid-localized proteins (n = 35; Figure 7 and Table S6). Remarkably, the numerous structural components of the chloroplast photosynthetic machinery are poorly represented in this gene set (only four relevant genes; PsbP-1, GAPA, FNR1 and a putative violaxanthin de-epoxidase gene). Instead, many genes encode regulatory factors such as transcriptional/translational regulators (e. g. , HCF107, HCF152, HCF164, HCF173, SVR3, FUG1), proteins involved in stress responses, or factors with central functions in integration of photoperiod and circadian rhythms (TOC1, PRR7, GIGANTEA and TZP). Altogether, this list defines a suite of genes that are particularly impacted by H2B monoubiquitination during their upregulation by light.
This study has examined the spatial and temporal dynamics of H2Bub distribution along the Arabidopsis genome in relation to gene expression during the initial events of photomorphogenesis. We found that 177 genes gain H2Bub de novo within 1 h, highlighting the dynamic nature of H2B monoubiquitination in a natural genomic context. Because the study was conducted during a short temporal window, these dynamics occur largely in the absence of cell division. Whole plants with numerous cell identities were used, and therefore our observations do not reveal whether H2Bub can increasingly be deposited at some DNA loci or whether an increased number of cells acquire the mark on those loci during gene upregulation. Determining these aspects will require further investigations using homogeneous cell populations or single-cell analyses. As was expected from previous studies using mainly cell-based experiments and/or reporter genes, albeit not in a developmental context, we observed that gene activation associates with H2Bub enrichment. Moreover, inactivation of HUB1 globally and locally affected the upregulation of many genes. In line with their preferential marking by H2Bub, this defect was more pronounced for long genes. These data are therefore in agreement with a role for H2Bub during transcription elongation, long genes being more particularly sensitive to the kinetics of chromatin opening for RNA Polymerase II processivity [54]. In this respect, it is noteworthy that elevated expression of the Flowering Locus C (FLC) gene requires both H2B ubiquitination through HUB1/HUB2 and deubiquitination through the action of Ubiquitin-Protease 26, suggesting that H2Bub is subjected to transcription-coupled cycling in plants, as observed in other systems [55]. The process has not been characterized in plants so far, and re-examination of light-driven gene induction capacity in mutant plants lacking H2Bub deubiquitination might help to decipher this mechanistic aspect. We also observed that light-driven downregulation was frequently decreased in hub1-3 seedlings, and was in this case independent of gene length. The decreased kinetics of gene expression changes for both up- and downregulation notably affected genes with rapid and transient light induction. Given these findings, we propose that HUB1 and/or H2Bub can facilitate the fine-tuning of gene expression to rapidly attain appropriate levels of expression. In human, only a subset of genes were shown to be affected in cells lacking H2Bub, which seems at odds with the fact that H2Bub is present on most active genes [11], [56]. In plants too, only a few genes were found to be misregulated in Paf1c mutants [57] in which histone H2B monoubiquitination is supposed to be affected. Owing to a fine effect on gene expression dynamics, kinetic analyses rather than steady state comparisons may therefore be more appropriate for revealing genes impacted by H2Bub-associated pathways. In contrast to gene activation, we found no relationship between gene downregulation and H2Bub levels. This suggests that H2Bub domains are not simultaneously removed when genes are downregulated. The long delay for H2Bub decrease could further indicate that loss of H2Bub is mainly replication-dependent. This contrasts with other chromatin marks associated with active transcription such as H3K27ac and H3K4me3, which have been shown to decrease rapidly on the phyA gene during Arabidopsis de-etiolation [34]. This also contrasts with previous observations in human cells in which turning off of the p21 gene was associated with a concomitant decrease of H2Bub at this locus [56]. Differential efficiency of H2Bub ubiquitin proteases in plants or weak deubiquitination over light-regulated genes might account for this difference, as could the persistence of some residual transcriptional activity on some light-repressed genes. Even so, such a lasting effect resembles the persistence of H3K4me3 on the GAL10 reporter gene long after its inactivation and after the dissociation of RNA Pol II in S. cerevisiae [58]. It was proposed that H3K4me3 domains could serve as a short-term memory of previous elevated transcriptional activity, which might be important for genes that are rapidly switched on by environmental changes [58], [59]. Given the tight relationship between H3K4me3 and H2Bub in S. cerevisiae, our observations suggest that H2Bub may play a similar role in plants. In the context of photomorphogenesis, such temporary marking might allow establishment of an initial light-adapted expression state. Future studies aimed at determining the minimum time required for losing H2Bub after termination of the light stimulus as a function of gene activity and DNA replication will be of interest, as will experiments in which seedlings are compared for their capacity to respond to successive light stimuli. H2Bub is highly associated with H3K4me3 and H3K36me3 along the Arabidopsis genome [4], and several COMPASS-like complexes with Set-methyltransferase activity exist in plants [60], [61]. H2B monoubiquitination could therefore potentially contribute to a trans-histone crosstalk with histone H3 on active genes in plants as in other systems. Nonetheless, a requirement for H2Bub for subsequent H3K4me3 deposition has not been demonstrated mechanistically in plants. Although the genome-wide distribution of H3K4me3 in a hub1 mutant background has not been reported, the bulk of H3K4me3 is maintained in mutants lacking H2B monoubiquitination [39], [40], [55], [62] as well as in Paf1c mutants [57]. Here we observed that upregulation of HCF173, TZP and SPA1 can associate with H3K4me3 enrichment in the hub1-3 mutant background. H2Bub is therefore not a prerequisite for recruiting histone methyltransferase activities mediating H3K4 trimethylation on these two genes, and independent pathways are likely at play here. These observations are in agreement with a proposed model in which H3K4me3 occurs prior to H2B ubiquitination and deubiquitination, whereas H3K36me3 occurs afterward in plants [55]. They further suggest that the transcriptional defects linked to H2Bub loss are not globally mediated through secondary defects on these other two marks. Much remains to be investigated about the mechanistic role of H2B monoubiquitination during transcription in plants. In vitro evidence using human cell extracts suggests that histone H2B is ubiquitinated ahead of the transcribing polymerase, which is important for the pioneering round of transcription, but that its rate limiting function may lie in the reassembly of nucleosomes [15]. Recent genome-wide analyses in S. cerevisiae further showed that H2Bub-mediated nucleosome reassembly can elicit different functional outcomes on genes depending on its positional context in promoter (repressive) versus transcribed (activating) regions [63]. Different mechanisms may operate in plants, as H2Bub is absent from promoter regions ([4], this study). This discrepancy might either reflect fast and efficient H2Bub deubiquitination at promoters, or targeting of the H2B ubiquitination machinery to the transcribed regions only, eventually mediated by Polymerase-associated factors. Paf1c is a good candidate to determine this [57]. Consequently, comparison of the distribution of HUB1, Paf1c and the elongating form of RNA Pol II along genes in wild-type and hub1 mutant plants might allow these possibilities to be distinguished. Based on immunoblot and ChIP analyses, H2Bub is lost in hub1-3 plants. Consequently, hub1 mutants represent the best available tool to assess the impact of H2Bub deposition on gene expression in plants. Indeed, the Arabidopsis genome encodes 11 histone H2B genes [36] and therefore is not amenable to genetic strategies that allow specific abrogation of histone H2B monoubiquitination through targeted lysine mutation. Nonetheless, although we assume that most defects are due to the lack of H2Bub in these mutants, we cannot rule out that some might also result from other functions of HUB1. In particular, because H2Bub domains were stable on most light-downregulated genes, altered downregulation in hub1-3 might also be triggered through H2Bub-independent effects. Indeed, the absence of HUB1 and H2Bub might affect gene downregulation at the transcriptional level by directly/indirectly decreasing rates of transcription but also through post-transcriptional activities. For many genes, the decrease in mRNA levels during light-driven downregulation is rapid, and therefore mRNA turnover might be critical. It has long been known that transcription elongation is tightly coordinated with mRNA processing steps [64]. More precisely, Paf1c-dependent H2Bub deposition differentially affects the stability of short- and long-lived mRNAs in yeast [65]. The potential role of H2Bub and of the HUB complex on post-transcriptional events has not been investigated in plant systems, and our data therefore suggest that such analyses might be of interest to investigate mechanisms allowing the rapid modulation of gene expression. Our analyses identified a series of 90 genes impacted by H2Bub dynamics, i. e. , which are fully induced by light in a HUB1-dependent manner and which concomitantly gain H2Bub. This subset represented only ∼10% of the light-induced genes, so the determinants of this specificity remain to be investigated. In line with previous findings [11], they may correspond to genes particularly sensitive to H2Bub loss, eventually located in particular chromatin contexts or subjected to tight regulation. In the context of photomorphogenesis, we found that many encode regulatory components rather than structural elements of the photosynthetic machinery. This finding therefore suggests that several regulatory genes are particularly dependent on transcription-coupled chromatin-based regulatory processes for rapid modulation of expression during the photomorphogenic developmental switch. Weaker transcriptional responses might directly be responsible for the enhanced sensitivity to dark-to-light shifts of hub1-3 mutant seedlings. The identification of circadian clock components is also noteworthy, as rapid changes in RNA levels are critical for diurnal oscillations. Modulation of the transcripts of the central oscillator TOC1 requires diurnal cycles of histone H3 acetylation and de-acetylation [66], which suggests that much remains to be determined about the role of H2Bub and associated chromatin modifiers in circadian mRNA oscillations. More generally, it can be expected that contribution of HUB1 to the modulation of gene expression also impacts other rapid transcriptional responses to environmental cues, in agreement with the defective responses of hub1 plants during fungal infection [62]. This study therefore opens the way for future studies deciphering how specific sequences are targeted for H2Bub deposition through light signal transduction pathways and whether transcriptional activity might be memorized on these loci prior to DNA replication.
Besides hub1-1 (Ler ecotype), all Arabidopsis thaliana plants were in the Col-0 background. The hub1-3 T-DNA insertion mutant (GABI_276D08) was obtained from Gabi-Kat [67], and hub1-5, hub1-4, and hub2-2 corresponding to the lines SALK_044415, SALK_122512 and SALK_071289 were obtained from NASC [68]. The hub1-1 mutant [38] has been described previously as ang4-1 [69]. For de-etiolation experiments, seedlings were grown on MS medium without sugar as described in [70]. Some samples were further exposed to white light (100 µmol. m−2. s−1) for the indicated duration and seedlings were harvested concomitantly at 4pm (8zt) under a green safe light for RNA or ChIP extractions. For photobleaching assays, seeds were grown on MS medium supplemented with 1% sucrose in the same conditions. The proportion of green/bleached seedlings was recorded on 100 plants for each genotype 72 h after transfer to continuous light. Total RNA was isolated with the RNeasy Plant mini-kit (Qiagen). For RT-qPCR, 0. 5 µg of RNAs were DNase-treated using Amplification Grade DNaseI (Invitrogen) and cDNAs were synthesized using oligo (dT) and SuperScript III reverse transcriptase (Invitrogen). Quantitative PCR was performed using the LightCycler 480 SYBR Green I Master (Roche). Primer sequences are listed in Table S7. RNA levels were normalized as in [71] and against the two housekeeping genes At4g29130 and At2g36060. Wild-type and hub1-3 seedlings were grown as indicated in Figure 1A and RNA was extracted just before, 1 or 6 h after exposure to light. For each genotype, samples at 1 and 6 h were compared with their respective dark points separately. Two independent biological replicates were produced using different seed batches. The cDNA synthesis, amplification, labelling, hybridizations and scanning of the slides were performed as described in [72]. Microarray analysis was carried out on CATMA arrays containing gene-specific tags (GSTs) for 22,089 Arabidopsis thaliana genes [44], [73]. Additional descriptions and statistical analyses are given in Text S1. Data were validated by RT-qPCR using the primers listed in Table S7 with the same RNA samples and also on additional biological replicates for independent validations. Transcriptome data were deposited at GEO (http: //www. ncbi. nlm. nih. gov/geo/ [74]; Accession number GSE21922) and at CATdb (http: //urgv. evry. inra. fr/CATdb/ [75]; Project: AU10-03-Hub1) according to Minimum Information About a Microarray Experiment standards (MIAME). Chromatin extractions, immunoprecipitations, DNA amplification, labelling and hybridizations were performed as described previously [4] using antibodies recognizing H2Bub (Medimabs MM-0029, lot 298060417), H3K4me3 (Millipore 05-745, lot NG1717145), H3K36me3 (Abcam ab9050, lot 826245) and H3 (Millipore 07-690, lot DAM1832538). For each ChIP-chip experiment, two independent biological replicates were performed using different seed batches. Each replicate was analysed in dye-swap on Roche NimbleGen tiled arrays of 50–75 nt tiles, with 110 nt spacing on average, that are tiled across the entire genome sequence (TAIR7), without repeat masking and synthesized in triplicates of 711 320 tiles each on a single array (GEO accession GPL11005) as described in [4]. Computational analyses of the data are described in Text S1. Variations of H2Bub enrichment over relevant genes were validated by quantitative PCR as described above on the DNA samples used for ChIP-chip analyses and on additional independent biological replicates using the primers listed in Table S7. ChIP-chip data were deposited at GEO under accession number GSE36515 according to MIAME. | In eukaryotes, chromatin-based mechanisms overlay with DNA sequence information to determine the transcriptional output of the genome. Evaluating the role of chromatin state variations in the regulation of gene expression is therefore key to understanding their contribution to development. Several transcriptional coactivators contribute to the selective regulation of cellular pathways by coordinating histone H2B monoubiquitination (H2Bub) with other histone modifications. Although H2Bub is present on a large number of genes, its loss was shown to affect RNA levels for only a small subset of genes, and therefore its influence on gene expression is not well understood. Here we assessed the impact of H2Bub on expression changes during a rapid developmental transition that initiates upon exposure of plants to light. This revealed that H2Bub marking is highly dynamic in a genomic context. Furthermore, a large repertoire of light-responsive genes was impaired for rapid up- or downregulation, indicating that H2Bub is important for attaining appropriate expression levels. Regulatory factors and circadian clock components are well represented within the set of genes impacted by H2Bub dynamics for rapid changes in RNA levels, indicating that some genes whose mRNAs need tight and rapid control are particularly sensitive to chromatin-based mechanisms linked to H2Bub deposition. | Abstract
Introduction
Results
Discussion
Materials and Methods | plant biology
signaling in selected disciplines
histone modification
plant science
model organisms
developmental signaling
plant ecology
chromatin
arabidopsis thaliana
plant signaling
gene expression
plant genetics
biology
molecular biology
signal transduction
plant and algal models
molecular cell biology
plant-environment interactions | 2012 | Histone H2B Monoubiquitination Facilitates the Rapid Modulation of Gene Expression during Arabidopsis Photomorphogenesis | 9,364 | 294 |
The ability of Plasmodium falciparum–infected red blood cells (IRBCs) to bind to vascular endothelium, thus enabling sequestration in vital host organs, is an important pathogenic mechanism in malaria. Adhesion of P. falciparum IRBCs to platelets, which results in the formation of IRBC clumps, is another cytoadherence phenomenon that is associated with severe disease. Here, we have used in vitro cytoadherence assays to demonstrate, to our knowledge for the first time, that P. falciparum IRBCs use the 32-kDa human protein gC1qR/HABP1/p32 as a receptor to bind to human brain microvascular endothelial cells. In addition, we show that P. falciparum IRBCs can also bind to gC1qR/HABP1/p32 on platelets to form clumps. Our study has thus identified a novel host receptor that is used for both adhesion to vascular endothelium and platelet-mediated clumping. Given the association of adhesion to vascular endothelium and platelet-mediated clumping with severe disease, adhesion to gC1qR/HABP1/p32 by P. falciparum IRBCs may play an important role in malaria pathogenesis.
Malaria continues to be a major public health problem in many parts of the tropical world, with approximately 500 million malaria cases reported annually that result in 1–2 million deaths every year [1,2]. Deaths from malaria mainly occur in young children living in sub-Saharan Africa and are caused by infection with P. falciparum. One of the important virulence mechanisms associated with P. falciparum infection is the unique ability of P. falciparum trophozoites and schizonts to sequester in the vasculature of diverse host organs [3–7]. Sequestration of P. falciparum–infected red blood cells (IRBCs) in the microvasculature of the brain is associated with severe pathological outcome of cerebral malaria [3,5, 7]. P. falciparum IRBCs can also bind to platelets to form platelet-mediated clumps, a cytoadherence phenomenon that is associated with severe disease [8–10]. Adhesion of IRBCs to vascular endothelium is mediated by interaction of the P. falciparum erythrocyte membrane protein-1 (PfEMP-1) family of variant surface antigens with host receptors [11–13]. The endothelial receptors used by P. falciparum for adhesion include thrombospondin (TSP) [14], CD36 [15], intercellular adhesion molecule-1 (ICAM-1) [16], platelet/endothelial cell adhesion molecule (PECAM/CD31) [17], vascular cell adhesion molecule-1 (VCAM-1) [18], endothelial leukocyte adhesion molecule-1 (ELAM-1) [18], normal immunoglobulin (IgG) [19], chondroitin sulfate A (CSA) [20,21], and hyaluronic acid (HA) [22]. Expression of ICAM-1 is upregulated on cerebrovascular endothelium [5,23], and P. falciparum IRBCs co-localize with ICAM-1 in cerebral vessels of patients who die of cerebral malaria [23], suggesting that adhesion to ICAM-1 plays a key role in cerebral sequestration. Adhesion of P. falciparum IRBCs to host vascular endothelium under flow conditions involves three distinct events, namely, margination, rolling, and static arrest/tethering, which may require multiple receptor–ligand interactions [24–26]. Adhesion to endothelial cells under flow requires binding of P. falciparum IRBCs to ICAM-1 as well as to CD36 [25]. Expression of ICAM-1 on brain endothelium is upregulated during blood stage P. falciparum infection [5,23]. However, the expression of CD36 on brain endothelial cells is minimal [23]. Platelets, which have been shown to accumulate in brain microvasculature of patients who die of cerebral malaria, express CD36 on their surface and may act as bridges for adhesion of P. falciparum IRBCs with brain vascular endothelium [27–29]. Alternatively, other as yet unidentified endothelial receptors may play a role in adhesion of P. falciparum IRBCs to cerebral capillaries. In case of platelet-mediated clumping, the only receptor identified for binding of IRBCs to platelets thus far is CD36 [9]. However, in previous studies, antibodies to CD36 could not completely disrupt clumps formed by some P. falciparum field isolates [9], suggesting that alternative host receptors may participate in platelet-mediated clumping. Here, we report the identification of the 32-kDa human protein gC1qR/HABP1/p32 (referred to below as gC1qR/HABP1 for brevity) as a novel host receptor for cytoadherence by P. falciparum. gC1qR/HABP1 is a ubiquitously expressed membrane protein that was initially shown to bind to the globular “head” of complement component C1q [30] as well as to HA [31]. This receptor appears to bind to diverse ligands and has multiple functions [32,33]. It is expressed on diverse cell types, including endothelial cells [34], platelets [35], and dendritic cells [36], and is used as a cell surface receptor by microbial pathogens for pathogenic processes such as host cell entry [37,38] and suppression of immune function [39,40]. Given its localization on endothelial cells and platelets, we hypothesized that gC1qR/HABP1 may serve as a cytoadherence receptor for P. falciparum. Here, we demonstrate that gC1qR/HABP1 is expressed on human brain microvascular endothelial cells (HBMECs) and can be used by P. falciparum as a receptor for cytoadherence. In addition, we show that P. falciparum IRBCs can bind gC1qR/HABP1 on platelets to form platelet-mediated IRBC clumps. Given the association of both of these cytoadherence phenotypes with severe malaria, this study identifies a novel host receptor that may play an important role in malaria pathogenesis.
Recombinant human gC1qR/HABP1 was expressed in E. coli and purified to homogeneity (Figure S1). Recombinant gC1qR/HABP1 has the expected mobility on SDS-PAGE and has a purity of greater than 98%. Analysis by gel permeation chromatography reveals that the majority of gC1qR/HABP1 is trimeric, as predicted by the crystal structure ([41], Figure S1). Dimers and trimers of gC1qR/HABP1 purified by gel permeation chromatography migrate with the expected mobility for gC1qR/HABP1 monomers by SDS-PAGE (Figure S1). Recombinant gC1qR/HABP1 binds its known ligands, C1q (Figure S2) and HA (Figure S2), confirming that it is functional. P. falciparum laboratory strains as well as field isolates were tested for binding to recombinant gC1qR/HABP1 coated on plastic Petri plates (Table 1; Figure S3) and to CD36 and ICAM-1 expressed on the surface of stably transfected Chinese hamster ovary (CHO) cells (Table 1). Three of the eight P. falciparum field isolates tested bind gC1qR/HABP1 (Table 1). Of these, IGH-CR14 shows the most significant binding to gC1qR/HABP1 (Table 1) and was selected for further analysis. P. falciparum laboratory strain 3D7, which binds gC1qR/HABP1 (Table 1), was also used for further study. IGH-CR14 binds CD36 and ICAM-1 in addition to gC1qR/HABP1, whereas 3D7 binds CD36 and gC1qR/HABP1 but not ICAM-1 (Table 1). IGH-CR14 binds gC1qR/HABP1 monomers and trimers at similar levels (Figure S3). Soluble C1q blocks the binding of IGH-CR14 to gC1qR/HABP1, suggesting that binding sites on gC1qR/HABP1 used by IRBCs and C1q may be overlapping (Figure S4). HA has no effect on binding of IGH-CR14 to gC1qR/HABP1 (Figure S4). Polymerase chain reaction–based analysis of two polymorphic antigens, MSP-1 and MSP-2, using methods described previously [42] confirmed that both IGH-CR14 and 3D7 contain single distinct genotypes (unpublished data). However, both IGH-CR14 and 3D7 may contain multiple variants with distinct binding phenotypes as a result of antigenic variation. In order to test if P. falciparum IRBCs, which bind gC1qR/HABP1, can also bind other receptors like CD36 or ICAM-1, we selected IGH-CR14 and 3D7 for binding to gC1qR/HABP1, separated binders (IGH-CR14+ and 3D7+) from non-binders (IGH-CR14− and 3D7−), and tested them in binding assays. As expected, IGH-CR14+ and 3D7+ show increased binding to gC1qR/HABP1, whereas IGH-CR14− and 3D7− display reduced binding to gC1qR/HABP1 compared to IGH-CR14 and 3D7, respectively (Table 2). The gC1qR/HABP1 binders, IGH-CR14+ and 3D7+, do not bind ICAM-1 or CD36, whereas IGH-CR14− retains binding to ICAM-1 and CD36, and 3D7− retains binding to CD36 (Table 2). These findings indicate that binding of P. falciparum IGH-CR14 and 3D7 to gC1qR/HABP1 is not linked to binding to ICAM-1 or CD36. We have used mouse serum raised against recombinant gC1qR/HABP1 to detect gC1qR/HABP1 on the surface of human umbilical vein endothelial cells (HUVECs), immortalized HBMECs, and primary brain microvascular cells (PBMECs) by flow cytometry. Anti-gC1qR/HABP1 mouse serum recognizes a single band of the expected size (32 kDa) in whole cell lysates as well as in membrane preparations of HUVECs by western blotting (Figure S5). Moreover, anti-gC1qR/HABP1 mouse serum detects gC1qR/HABP1 on the surface of HUVECs, HBMECs, and PBMECs by flow cytometry (Table S1). Unlike ICAM-1, surface expression of gC1qR/HABP1 is not significantly upregulated on the surface of HUVECs, HBMECs, and PBMECs following treatment with TNF-α (Table S1). CD36 is not detected on the surface of HUVECs, HBMECs, and PBMECs before or after treatment with TNF-α (Table S1). In order to explore if P. falciparum IRBCs use gC1qR/HABP1 to bind endothelial cells, we tested IGH-CR14 and 3D7 for binding to HUVECs and HBMECs. We also tested whether selection of IGH-CR14 and 3D7 for binding to gC1qR/HABP1 results in enhanced binding to endothelial cells. IGH-CR14+ and 3D7+ show increased binding to both gC1qR/HABP1 and HUVECs compared to IGH-CR14 and 3D7 or the non-binders, IGH-CR14− and 3D7− (Table 2). The association of enhanced binding to gC1qR/HABP1 and HUVECs (Table 2) suggested that IGH-CR14+ and 3D7+ use gC1qR/HABP1 as a cell surface receptor to bind to HUVECs. In order to confirm that binding of IGH-CR14+ and 3D7+ to HUVECs was mediated by gC1qR/HABP1, we tested whether soluble gC1qR/HABP1, as well as anti-gC1qR/HABP1 mouse serum, can inhibit binding of IGH-CR14+ and 3D7+ to HUVECs. Soluble gC1qR/HABP1 blocks the binding of both IGH-CR14+ and 3D7+ to HUVECs in a dose-dependent manner, whereas bovine serum albumin (BSA) and recombinant ICAM1-Fc have no effect on binding (Figure 1). Anti-gC1qR/HABP1 mouse serum also blocks binding of both IGH-CR14+ and 3D7+ to HUVECs, whereas pre-immune mouse serum and antibodies directed against ICAM-1 or CD36 have no effect on binding (Figure 1). These findings demonstrated that binding of IGH-CR14+ and 3D7+ to HUVECs is mediated by gC1qR/HABP1. The gC1qR/HABP1 binder IGH-CR14+ also shows increased binding to HBMECs compared to IGH-CR14 and IGH-CR14− (Table 2). Binding of IGH-CR14+ to HBMECs is inhibited by soluble gC1qR/HABP1 but not by ICAM1-Fc or CD36-Fc (Figure 2). Binding of IGH-CR14+ to HBMECs is also inhibited by anti-gC1qR/HABP1 mouse serum but not by pre-immune mouse serum or monoclonal antibodies against ICAM-1 and CD36 (Figure 2). These findings demonstrate that IGH-CR14+ uses gC1qR/HABP1 as a receptor to bind to HBMECs. Mouse serum raised against gC1qR/HABP1 was used to detect expression of gC1qR/HABP1 on the surface of platelets by flow cytometry. P-selectin (CD62) was used as a marker for platelet activation. Whereas gC1qR/HABP1 is detected on the surface of both resting and activated platelets, P-selectin is only expressed on the surface of activated platelets (Table S2). Given the presence of gC1qR/HABP1 on the surface of platelets, we examined whether P. falciparum could use gC1qR/HABP1 as a receptor for platelet-mediated IRBC clumping. IGH-CR14, IGH-CR14+, and IGH-CR14− were tested for formation of clumps in the presence of platelet-rich plasma (PRP) and platelet-poor plasma (PPP). All three parasites form clumps in the presence of PRP, whereas no clumps are seen in the presence of PPP (Figure 3). Similarly, 3D7,3D7+, and 3D7− form clumps in the presence of PRP (Figure 3). The P. falciparum isolate JDP8, which binds ICAM-1 and does not bind gC1qR/HABP1 or CD36, does not form clumps in PRP or PPP. IGH-CR14, IGH-CR14−, 3D7, and 3D7− bind CD36 (Table 2), which is a known receptor for platelet-mediated clumping. IGH-CR14+ and 3D7+ do not bind CD36, but bind gC1qR/HABP1 (Table 2). Analysis of clumps formed by IGH-CR14+ using scanning and transmission electron microscopy confirmed the presence of platelets in the clumps (Figure 3), suggesting that IGH-CR14+ IRBCs use gC1qR/HABP1 as a receptor to form platelet-mediated clumps. In order to confirm the identity of the receptor used by IGH-CR14+ and 3D7+ for platelet-mediated clumping we tested the ability of soluble gC1qR/HABP1 and CD36-Fc, as well as antibodies directed against gC1qR/HABP1 and CD36, to inhibit clumping. Both CD36-Fc and anti-CD36 monoclonal antibodies block the clumping of IGH-CR14, IGH-CR14−, 3D7, and 3D7− (Figures 4 and 5). Soluble gC1qR/HABP1 and anti-gC1qR/HABP1 mouse serum does not inhibit clumping of these parasites (Figures 4 and 5). These findings indicate that IGH-CR14, IGH-CR14−, 3D7, and 3D7− primarily use CD36 on platelets to form clumps. Conversely, soluble gC1qR/HABP1 and anti-gC1qR/HABP1 mouse serum block clumping of IGH-CR14+ and 3D7+ parasites, whereas CD36-Fc and anti-CD36 monoclonal antibodies do not have any inhibitory effect on clumping of IGH-CR14+ and 3D7+ parasites (Figures 4 and 5). These studies confirm that both IGH-CR14+ and 3D7+ use gC1qR/HABP1 as a receptor for platelet-mediated clumping.
Adhesion of P. falciparum IRBCs to endothelial receptors, which enables sequestration in host organs, and binding to platelets, which produces IRBC clumps, are important pathogenic mechanisms in malaria [4–10]. Here, we report the identification of the 32-kDa human protein gC1qR/HABP1 as a novel cytoadherence receptor for adhesion of P. falciparum IRBCs to both endothelial cells and platelets. gC1qR/HABP1 is synthesized as a 282–amino acid pre-pro protein, which contains a 73–amino acid long N-terminal mitochondrial targeting sequence [43,44]. gC1qR/HABP1 is found in mitochondria and also on the surface of mammalian cells. There are other examples of proteins that have mitochondrial localization sequences and are found in other cellular locations in addition to mitochondria [45]. For example, mitochondrial aspartate aminotransferase (ApsAT) is found in the mitochondria as well as on the plasma membrane of adipocytes, where it is involved in binding and uptake of fatty acids [45]. Another mammalian protein, Slit3, whose homolog in Drosophila is involved in developmental regulation, is predominantly a mitochondrial protein having an N-terminal mitochondrial targeting sequence, but is also shown to be expressed on epithelial cell surfaces [46]. The mechanisms by which these proteins translocate to the cell surface and to the mitochondria are not known. Given the presence of the mitochondrial targeting sequence and absence of a transmembrane domain or consensus glcophosphatidyl inositol (GPI) –anchoring signature sequence, the surface localization of gC1qR/HABP1 is intriguing. The localization of gC1qR/HABP1 on the surface of diverse human cells has been demonstrated unequivocally both here and previously [31–34,47,48]. We have demonstrated here that mouse serum raised against recombinant gC1qR/HABP1 specifically reacts with a protein of the expected size for gC1qR/HABP1 (32 kDa) in whole cell lysates as well as in membrane fractions of HUVECs by western blotting (Figure S5). Anti-gC1qR/HABP1 mouse serum detects the presence of gC1qR/HABP1 on the surface of HUVECs, HBMECs, and PBMECs by flow cytometry (Table S1). The observation that gC1qR/HABP1 is expressed on the surface of microvascular endothelial cells suggests the possibility that it may be used as a receptor for cytoadherence by P. falciparum IRBCs. Given that the expression profile of the cytoadherence receptors gC1qR/HABP1, ICAM-1, and CD36 on HUVECs, HBMECs, and PBMECs is similar, we used HUVECs and HBMECs for adhesion assays with P. falciparum IRBCs. We have demonstrated that P. falciparum laboratory strains as well as field isolates can bind to recombinant gC1qR/HABP1 (Table 1; Figure S3). Selection of P. falciparum IGH-CR14 and 3D7 for binding to gC1qR/HABP1 allowed separation of gC1qR/HABP1 binders IGH-CR14+ and 3D7+, and non-binders IGH-CR14− and 3D7−. Selection of IGH-CR14 and 3D7 for binding to gC1qR/HABP1 resulted in increased binding of IRBCs to HUVECs and HBMECs (Table 2), suggesting that these parasites use gC1qR/HABP1 to bind to human endothelial cells. Indeed, recombinant gC1qR/HABP1, as well as anti-gC1qR/HABP1 mouse serum, blocked the binding of IGH-CR14+ to HUVECs and HBMECs (Figures 1 and 2), confirming that these parasites use gC1qR/HABP1 as a receptor for adhesion of IRBCs to endothelial cells. The demonstration that P. falciparum IRBCs can use gC1qR/HABP1 as a receptor to bind to microvascular endothelial cells suggests that adhesion to gC1qR/HABP1 may play a role in parasite sequestration in vivo. Another distinct cytoadherence phenotype that is associated with severe malaria is platelet-mediated clumping of IRBCs. CD36, which is expressed on both resting and activated platelets, has been identified as a receptor for platelet-mediated clumping [9]. However, in a previous study, clumps formed by some P. falciparum field isolates could not be disrupted completely by anti-CD36 antibodies [9], suggesting that other unidentified receptors on platelets might also mediate clumping of IRBCs. Previous studies have suggested that gC1qR/HABP1 is expressed on the surface of activated platelets [35]. Here, we have demonstrated that gC1qR/HABP1 is also expressed on resting platelets (Table S2). Expression of gC1qR/HABP1 on the surface increases upon activation of platelets (Table S2). IGH-CR14+ and 3D7+, which bind to gC1qR/HABP1 but not to CD36, formed clumps in the presence of platelets. Formation of clumps by IGH-CR14+ and 3D7+ was inhibited by soluble gC1qR/HABP1 and anti-gC1qR/HABP1 antibodies (Figures 4 and 5). These observations demonstrate that P. falciparum IRBCs can use gC1qR/HABP1 as an alternative receptor to bind to platelets and form clumps. The parasite ligands that mediate adhesion of IRBCs to gC1qR/HABP1 remain to be identified. Previous studies have demonstrated that the PfEMP-1 family of variant surface antigens encoded by var genes mediates interactions with a diverse range of host receptors to enable adhesion to host endothelium and sequestration in host organs [4,6, 12,13]. It is likely that PfEMP-1 may also mediate adhesion to gC1qR/HABP1. Identification of var genes that are differentially transcribed in gC1qR/HABP1 binding parasites may enable the identification of the PfEMP-1 variant that is responsible for adhesion to gC1qR/HABP1. In summary, we have shown that P. falciparum IRBCs use gC1qR/HABP1 as a receptor to bind vascular endothelium and platelets. The observation that P. falciparum can use gC1qR/HABP1 as a receptor to bind HBMECs, a cell line derived from brain microvascular endothelial cells, raises the possibility that adhesion of IRBCs to this novel receptor may be important for sequestration in brain microvasculature and cerebral malaria. The contribution of IRBC adhesion to gC1qR/HABP1 to platelet-mediated clumping and severe disease also needs to be examined. Comparison of the cytoadherence phenotypes of P. falciparum isolates collected from patients with mild and severe malaria may allow us to test whether adhesion to gC1qR/HABP1 is associated with an increased risk of severe malaria.
All chemicals used in the study were from Sigma (http: //www. sigmaaldrich. com/) unless otherwise indicated. Indian P. falciparum field isolates were collected from P. falciparum–infected individuals in Calcutta (Cal3770, Cal3813, Cal3875), Rajasthan (Raj68, Raj86, Raj116), Jagdalpur, Madhya Pradesh (JDP8), and Rourkela, Orissa (IGH-CR14), and cryopreserved in the Malaria Parasite Bank at the National Institute of Malaria Research, Delhi, India. P. falciparum field isolates and laboratory strains (A4, ITG-ICAM, and 3D7) were cultured in RPMI 1640 (Invitrogen, http: //www. invitrogen. com/) supplemented either with 10% heat-inactivated O+ pooled human sera or 5% Albumax I (Invitrogen) using O+ RBCs in an environment containing 5% O2,5% CO2, and 90% N2. Cultures were synchronized by sorbitol treatment as previously described [49]. To select parasites for binding to gC1qR/HABP1, synchronized P. falciparum 3D7 and IGH-CR14 cultures in trophozoite/schizont stages were incubated for 1 h in bacteriological Petri plates coated with recombinant gC1qR/HABP1 as described below for adhesion assays. Unbound parasites were collected using a pipette and separated from bound parasites. Both bound and unbound parasites were cultured and subjected to selection for binding to gC1qR/HABP1 two more times to separate binders (3D7+ and IGH-CR14+) and non-binders (3D7− and IGH-CR14−). Glycosaminoglycan biosynthesis–defective mutant Chinese hamster ovary cells (CHO-745) stably transfected to express human CD36 (CHO-CD36) and ICAM-1 (CHO-ICAM-1) on their surface [50] were kindly provided by Artur Scherf, Institut Pasteur, Paris, France. CHO cells were cultured in RPMI 1640 with 10% heat-inactivated fetal bovine serum (FBS). HUVECs were cultured in EBM2 bullet kit media (Cambrex Biosciences, http: //www. cambrex. com/) on gelatin-coated flasks according to instructions provided by the supplier. Immortalized HBMECs were cultured as previously described [51]. PBMECs were cultured in EGM-2 media provided by the supplier (ScienCell Research Laboratories, http: //www. sciencellonline. com/). A DNA fragment encoding mature human gC1qR/HABP1 (amino acids 74–282) was cloned in pET30b vector (Invitrogen) using the NdeI and BamHI restriction enzyme cloning sites. Recombinant gC1qR/HABP1 was expressed in E. coli BL21 (DE3) by induction with isopropyl-1-thio-β-galactosidase (IPTG) and purified from supernatants of lysed cells by ammonium sulfate fractionation followed by ion-exchange chromatography using UnoQ (Bio-Rad, http: //www. bio-rad. com/) as described previously [39]. Binding of recombinant gC1qR/HABP1 to its ligands, C1q and HA, was tested in solid phase binding assays as follows. gC1qR/HABP1 and HA were biotinylated with sulfo-NHS-LC-biotin and biotin-LC-hydrazide, respectively, as described by the manufacturer (Pierce Biotechnology, http: //www. piercenet. com/). Ninety-six-well plates were coated at 4 °C overnight with human C1q (250 ng per well). After blocking with 2% non-fat milk, the wells were incubated with varying concentrations of biotinylated gC1qR/HABP1. Bound biotin-gC1qR/HABP1 was detected with streptavidin-horse radish peroxidase (HRPO) using o-phenylene diamine dihydrochloride (OPD) as substrate. In order to test the binding of gC1qR/HABP1 to HA, ELISA plate wells were coated with recombinant gC1qR/HABP1 (250 ng per well) and incubated with different concentrations of biotin-HA. Bound biotin-HA was detected using streptavidin-HRPO and OPD. Ten microliters of purified gC1qR/HABP1 (100 μg/ml) was spotted on bacteriological Petri plates (Becton Dickinson, http: //www. bd. com/), allowed to adsorb overnight at 4 °C in a humidified chamber and used for binding assays with parasite cultures as previously described for binding to soluble CD36 and ICAM-1 [52]. BSA was spotted as control. Trophozoite-schizont stage parasite cultures at ∼1% hematocrit and ∼5% parasitemia were incubated with gC1qR/HABP1-coated Petri plates to allow binding. Bound cells were fixed with 2% glutaraldehyde, stained with 5% Giemsa stain, and scored using a Nikon TE200 microscope with a 100× objective. The total number of IRBCs and uninfected RBCs (URBCs) were counted from seven randomly selected distinct fields in duplicate spots from two independent experiments. The number of URBCs bound to gC1qR/HABP1 and the number of IRBCs bound to BSA spots was subtracted from the number of IRBCs bound to gC1qR/HABP1 to get the number of specific binding events. Fewer than five URBCs bound gC1qR/HABP1 per mm2, and fewer than five IRBCs bound BSA per mm2. CHO-745, CHO-CD36, and CHO-ICAM1 cells were grown in spots in tissue culture plates and tested for binding to P. falciparum trophozoite-schizont stage cultures using methods described earlier [52]. Bound IRBCs were fixed with 2% glutaraldehyde and detected by Giemsa staining. The number of IRBCs and URBCs bound to ∼200 CHO cells was scored in duplicate spots in two independent experiments. The number of URBCs bound to CHO-CD36 or CHO-ICAM1 was subtracted from bound IRBCs in each case. The number of IRBCs bound to CHO-745 was further subtracted from the number of IRBCs bound to CHO-CD36 and CHO-ICAM1 to obtain the number of specific binding events. Fewer than three URBCs bound to 100 CHO-CD36 or CHO-ICAM1 cells, and fewer than two IRBCs bound to 100 CHO-745 cells in all experiments. Specific binding of ten IRBCs or more per 100 CHO-ICAM1 or CHO-CD36 cell was therefore considered significant. Flow cytometry was used to study the expression of gC1qR/HABP1, ICAM-1, and CD36 on the surface of HUVECs, HBMECs, and PBMECs before and after treatment with TNF-α (eBioscience, http: //www. ebioscience. com/) and on the surface of resting and activated platelets. HUVECs, HBMECs, and PBMECs were cultured as described above and were treated with TNF-α (15 ng/ml) for 24 h before analysis. Mouse serum raised against gC1qR/HABP1 (diluted 1: 100), anti-ICAM-1 monoclonal antibody 15. 2 (1 μg per 106 cells; Serotec, http: //www. serotec. co. uk/), and anti-CD36 monoclonal antibody SMΦ (1 μg per 106 cells, Serotec) were used for detection of receptors by flow cytometry using the BD FACSCalibur System (Becton Dickinson). Resting platelets were isolated from whole blood collected in citrate-phosphate-dextrose (CPD) as follows. PRP was separated by centrifugation of whole blood at 300g for 5 min at room temperature (RT). PRP was incubated in equal volume of CCAT buffer (7. 7 mM citric acid, 95 mM trisodium citrate, 150 mM glucose, 5 mM adenosine, 3 mM theophylline) for 10 min at RT. Platelets were collected by centrifugation of PRP at 1,000g for 10 min at RT, washed once in CCAT buffer and resuspended in modified Hank' s balanced salt buffer (136 mM NaCl, 5. 3 mM KCl, 0. 4 mM MgSO4. 7H2O, 0. 3 mM NaH2PO4. 2H2O, 0. 77 mM Na2HPO4,0. 44 mM KH2PO4,0. 5 mM MgCl2. 6H2O, 5. 5 mM glucose, 0. 4 mM NaHCO3), and stored at RT until use. Resting platelets resuspended in RPMI 1640 were activated by treatment with thrombin (100 units/ml; Sigma Chemicals, http: //www. sigmaaldrich. com/) for 30 min at 37 °C. Resting and activated platelets were fixed with 2% p-formaldehyde and 0. 2% glutaraldehyde in phosphate buffered saline for 30 min at 4 °C. Mouse antiserum raised against gC1qR/HABP1 (diluted 1: 100) and anti-P-selectin monoclonal IgG antibody CTB201 (1 μg per 106 platelets; Santa Cruz Biotechnology, http: //www. scbt. com/) were used for detection of receptors by flow cytometry using the BD FACSCalibur System (Becton Dickinson). HUVECs and HBMECs were grown on gelatin-coated plates and used for binding assays with IRBCs following the same procedure used in case of CHO-CD36 and CHO-ICAM1 cells described above. For inhibition assays, either parasite cultures were pre-incubated with ICAM1-Fc (R&D Systems, http: //www. rndsystems. com/), gC1qR/HABP1, and BSA, or HUVECs were pre-incubated with anti-gC1qR/HABP1 mouse serum or monoclonal antibodies directed against CD36 (clone SMΦ, Serotec) and ICAM-1 (clone 15. 2, Serotec). Binding in the presence of proteins or serum was expressed as percent of binding in absence of any protein or serum. Platelet-mediated clumping assays were performed in the presence of PRP and PPP according to the method described previously [9]. IRBCs were labeled with acridine orange and the percentage of IRBCs present in clumps was determined by scoring ∼3,000 IRBCs at 20× magnification using a fluorescence microscope to determine frequency of clumping. A clump consists of three or more IRBCs as described previously [9]. All the clumps observed had fewer than 50 IRBCs. Parasite cultures were pre-incubated with soluble gC1qR/HABP1 or CD36-Fc (R&D Systems) for 10 min prior to adding PRP to test their ability to inhibit clumping. Antibodies directed against host proteins were added to PRP prior to incubation with parasite cultures to test their ability to block clumping. Cells were fixed in 2. 5% glutaraldehyde in 0. 1 M phosphate buffer (pH 7. 2) and processed according to methods described previously [9]. Samples were analyzed on a Morgagni 268D transmission electron microscope (FEI Philips, http: //www. fei. com/) and LEO 435 VP scanning electron microscope (Leo Electron Microscopy, http: //www. smt. zeiss. com/nts). | Adhesion of Plasmodium falciparum–infected red blood cells (IRBCs) to the endothelium lining the capillaries of vital host organs can obstruct blood circulation and is an important pathogenic mechanism in malaria. Adhesion of P. falciparum IRBCs to platelets results in the formation of IRBC clumps that can also obstruct blood flow and is implicated in severe malaria. Here, we have identified a novel cytoadherence receptor that is found on both endothelial cells and platelets. We demonstrate, for the first time to our knowledge, that P. falciparum IRBCs use the 32-kDa human protein gC1qR/HABP1/p32 as a receptor to bind to human endothelial cells, including brain microvascular endothelial cells. In addition, we show that P. falciparum IRBCs can bind to gC1qR/HABP1/p32 on platelets to form clumps. Our study has thus identified a novel host receptor that is used for both adhesion to vascular endothelium and platelet-mediated clumping. Given the association of these cytoadherence phenomena with severe disease, our study opens the door to investigations on the role of adhesion of P. falciparum IRBCs to gC1qR/HABP1/p32 in malaria pathogenesis. | Abstract
Introduction
Results
Discussion
Materials and Methods | cell biology
infectious diseases
plasmodium | 2007 | Plasmodium falciparum Uses gC1qR/HABP1/p32 as a Receptor to Bind to Vascular Endothelium and for Platelet-Mediated Clumping | 9,357 | 338 |
The reservoir and mode of transmission of Mycobacterium ulcerans, the causative agent of Buruli ulcer, remain unknown. Ecological, genetic and epidemiological information nonetheless suggests that M. ulcerans may reside in aquatic protozoa. We experimentally infected Acanthamoeba polyphaga with M. ulcerans and found that the bacilli were phagocytised, not digested and remained viable for the duration of the experiment. Furthermore, we collected 13 water, 90 biofilm and 45 detritus samples in both Buruli ulcer endemic and non-endemic communities in Ghana, from which we cultivated amoeboid protozoa and mycobacteria. M. ulcerans was not isolated, but other mycobacteria were as frequently isolated from intracellular as from extracellular sources, suggesting that they commonly infect amoebae in nature. We screened the samples as well as the amoeba cultures for the M. ulcerans markers IS2404, IS2606 and KR-B. IS2404 was detected in 2% of the environmental samples and in 4% of the amoeba cultures. The IS2404 positive amoeba cultures included up to 5 different protozoan species, and originated both from Buruli ulcer endemic and non-endemic communities. This is the first report of experimental infection of amoebae with M. ulcerans and of the detection of the marker IS2404 in amoeba cultures isolated from the environment. We conclude that amoeba are potential natural hosts for M. ulcerans, yet remain sceptical about their implication in the transmission of M. ulcerans to humans and their importance in the epidemiology of Buruli ulcer.
Most mycobacteria are environmental opportunistic species that only occasionally infect humans [1]. Only few mycobacterial species are known to be obligate parasites, such as Mycobacterium tuberculosis and M. leprae, the causative agents of tuberculosis and leprosy respectively. The third most common mycobacterial disease, Buruli ulcer (BU), is caused by M. ulcerans, an environmental opportunistic mycobacterium. BU occurs mainly in rural areas of West and Central Africa with over 40,000 cases reported between 2002 and 2010 [2]. To the present day, the main reservoir of M. ulcerans and its transmission from the environment to humans remain unknown. Epidemiological data from Africa suggest that proximity to slow-flowing or stagnant water best explains the distribution pattern of BU [3]. Nevertheless, it is unlikely that M. ulcerans occurs free-living in these waters, because (i) M. ulcerans evolved recently from the generalist, more rapid-growing environmental M. marinum via lateral gene transfer and reductive evolution to become adapted to a more protected niche [4], and (ii) since M. ulcerans is sensitive to several antibiotics, such as streptomycin and rifampicin, it is unlikely that the bacteria occur free living in an environment where Streptomyces griseus and Amycolatopsis rifamycinica, producers of respectively streptomycin and rifampicin, thrive [5], [6]. In order to survive without protection of a host a or biofilm, M. ulcerans would have developed natural resistance against these antibiotics as is the case for most opportunistic non tuberculous mycobacteria [6]. Recently in Australia, M. ulcerans has been found at high prevalence in two species of possum, thus suggesting a role for these mammals as reservoirs of M. ulcerans [7]. Although in Africa M. ulcerans has never been found in such high numbers in any particular element of the environment, low levels of M. ulcerans DNA were detected in many biotic components of aquatic ecosystems, such as plants, snail, fish and insects [3], indicating that M. ulcerans is ubiquitous in these ecosystems. The most explored hypothesis of M. ulcerans transmission in Africa argues that microphagous arthropods, e. g the heteropteran waterbugs, feed on M. ulcerans in water or biofilms, which are in turn consumed by predatory insects that may occasionally bite humans [8]. M. ulcerans bacilli are thought to be concentrated along this food chain, resulting in a sufficient infectious dose for humans [8]–[12]. Although the exact role of insects in the transmission of M. ulcerans remains to be proven [13], a series of experiments do support this hypothesis [9]–[12] and recent extensive fieldwork found a relatively high prevalence of M. ulcerans DNA in several waterbug species in a BU endemic area [14]. In addition, the only successful cultivation of M. ulcerans from an environmental source was from an aquatic hemipteran [15]. Nonetheless, these transmission hypotheses do not exclude an important role for other host species as M. ulcerans reservoirs. Recently, it has been postulated that amoebae might represent hosts for M. ulcerans and that they could be involved in the transmission from the environment to humans [3], [16], [17]. A study in Benin found that the detection frequency of free-living amoebae in water bodies in BU endemic villages was higher than in non-BU endemic villages [16]. M. ulcerans has been shown to form a biofilm on aquatic plants [18], and amoebae are often the main predators in biofilm communities. As a response to amoebal predation, many bacteria have acquired resistance to digestion in amoebal food vacuoles [19]. Also many mycobacterium species have been shown to survive and even thrive intracellularly in protozoa [19], [20]. Because of their hydrophobic cell wall, mycobacteria tend to attach to surfaces and are easily phagocytised by protozoa [21] and macrophages [22] and can even actively promote their entry into phagocytes [23]. Mycobacteria can use nutrients of protozoa as a food source and the intracellular life offers protection against harmful and fluctuating environmental influences, as protozoan cysts are remarkably resistant to extremes of temperature, drought and all kinds of biocides [24], [25]. However, few studies have investigated whether mycobacteria infect amoebae in their natural environment. Natural resistance to amoebae can have important consequences, as bacteria that infect and evade digestion in amoebae might use the same tools to enter and resist destruction within macrophages [19], [26], which have similar properties, as has been shown for Legionella pneumophila [27], an environmental opportunistic bacterium with amoebae as main reservoir. An intracellular stay in amoebae even enhances the virulence of L. pneumophila against mammalian cells [28]. This feature is not restricted to Legionella; M. avium passaged through Acanthamoeba castellanii is also more virulent towards a mouse model [29]. Even though M. ulcerans was long considered an extracellular pathogen, there is increasing evidence that the bacterium exhibits an important intracellular phase in neutrophils and macrophages during human infection (reviewed in [30]). So far only one study, published in 1978, has shown that M. ulcerans can be phagocytized and retained in an Acanthamoeba, yet this study provided only limited results [31]. In the present study we experimentally determined the capacity of M. ulcerans to infect A. polyphaga. We then further investigated the role of free-living amoebae as hosts for mycobacteria in their natural environment, including the potential of these protozoa as reservoirs for M. ulcerans.
The M. ulcerans strains used were ITM 030216 (Benin), ITM 980912 (China), ITM 5114 (Mexico) and ITM 842 (Surinam) from the collection of the Institute of Tropical Medicine (ITM), Antwerp, Belgium. A one week old, and therefore starving, axenic culture of A. polyphaga CCAP 1501/15 in PYG712 broth was adjusted to 105 cells/mL and 1 mL transferred into each of the wells of a 24-well tissue culture plate. The amoebae monolayers were seeded with suspensions of M. ulcerans in triplicate at an approximate multiplicity of infection of 1 (M. ulcerans to A. polyphaga) and plates were incubated at 30°C. Three hours after infection, 20 µg/ml kanamycin was added to prevent extracellular growth of M. ulcerans. At times 3 h, 3 d, 7 d and 14 d after infection, the supernatant was aspirated and discarded. At each time point the medium of the unused wells was replaced by fresh PYG712 broth containing 20 µg/mL kanamycin. To estimate the number of intracellular colony forming units (CFU) present inside amoebae at each time point the monolayer was suspended in 0. 1% SDS in order to lyse the amoebae. The lysed amoebae suspension was transferred into a tube containing glass beads and vortexed. This suspension, as well as two 10-fold dilutions were inoculated on Löwenstein-Jensen (LJ) medium. The tubes were then incubated at 30°C and read after 6 weeks. To localize intracellular bacteria, the amoeba monolayer was suspended in PBS and then processed for electron microscopy and for acid-fast staining using the Ziehl-Neelsen (ZN) method. For transmission electron microscopy the processing of samples was carried out taking into account that bacteria, including mycobacteria, have specific requirements for adequate preservation as discussed elsewhere [32]. Briefly, a volume of 0. 3 mL of the suspended monolayer was pelleted and the pellet was pre-fixed with 4% formaldehyde-1. 25% glutaraldehyde-10 mM CaCl2 for 24 h, then fixed in 1% OsO4–10 mM CaCl2 for 16 to 24 h, and then post-fixed in aqueous 1% uranyl acetate for 1 h. Further processing for electron microscopy was carried out with ethanol dehydration and Epon embedding. Ultrathin sections were double-stained with uranyl acetate (saturated aqueous solution) for 5 min followed by lead citrate for 3 min.
Using light microscopy, AFB were observed co-localizing with amoebae after co-incubation of each M. ulcerans strain with A. polyphaga for 3 hours (Figure 2). Electron microscopy was used to confirm the intracellular localization (Figure 3). The bacilli were seen in phagocytic vacuoles with the phagosomal membrane tightly opposed to the bacillary surface (tight phagosomes) (Figure 3D) or, less frequently, inside “spacious vacuoles” (Figure 3C). The phagosomes contained single (Figure 3C and D) or groups of bacilli (not shown). About 45% of the bacilli in the electron micrographs looked normal according to the parameters previously defined [47], including the presence of an asymmetric profile of the cytoplasmic membrane with the outer layer thicker and denser than the inner layer (Figure 3B). As is typical of the ultrastructure of normal mycobacterial cell envelopes [48], an electron-transparent layer of the cell wall of M. ulcerans was observed (Figure 3B). No electron-transparent zone [49] was seen around the intracellular bacilli (Figure 3). Figure 4 shows that viable M. ulcerans persisted within A. polyphaga for the duration of the experiment (two weeks) although their numbers decreased with 1 to 2 log. 181 amoeba cultures were obtained from 134 out of 148 collected samples (90. 5%). The isolation frequency of amoebae did not differ significantly between BU endemic and non-endemic sites (p = 0. 954, χ21 = 0. 004). Different habitats yielded different frequencies of amoeba isolation (p = 0. 044, χ23 = 8. 084), with the highest detection frequency in detritus (97. 8% vs. 76. 9% in water, and 88. 9% in biofilms). There was no significant difference between the sampled water bodies (estimated effect: 0. 03; 95% CI: −0. 19 to 0. 24). Because 15 of the 181 amoeba cultures did not survive transportation and/or storage, mycobacteria were only searched for in the remaining 166 amoeba cultures (isolated from 124 different samples). IS2404 was detected by real-time PCR in 3 out of 148 samples, after extracting the DNA directly from these samples. In only one of them IS2606 and KR-B were also detected, strongly suggesting the presence of M. ulcerans in that sample (Table 1). The Δ CT (IS2606 - IS2404) value of 1. 96 approaches the known fold difference in copy numbers between IS2404 and IS2606 for M. ulcerans, i. e. 2. 3 [44]. However, the high CT values in all 3 of the positive assays (Table 1) imply the presence of less than a genome in the 1 µL DNA extract added to the PCR-mixture so that the Δ CT' s cannot be considered as true representations of the relative copy numbers of the repeated sequences. Therefore in the two IS2404 positive yet IS2606 and KR-B negative samples the presence of M. ulcerans cannot be denied nor confirmed. Out of the 166 amoeba cultures tested (originating from 124 different samples), seven were positive for IS2404 (Table 1). Again, given the high CT values (Table 1), also here less than a genome was present in the 1 µL of DNA extract added to the PCR mixture. The IS2404 positive amoeba cultures were isolated from BU endemic as well as BU non-endemic communities and from different microbial habitats. None of the IS2404-containing amoeba cultures tested positive for IS2606 or KR-B. However, because of the low mycobacterial DNA content neither the absence or presence of M. ulcerans can be confirmed. None of IS2404 positive amoeba cultures were isolated from samples that had already been found positive for IS2404 in DNA extracted directly from the samples. The following amoebae were identified among the IS2404 positive cultures: Vahlkampfia avara (99% identical with the V. avara sequence in Genbank), a close relative of V. inornata (92% identical with the V. inornata sequence in Genbank), A. lenticulata (T5 genotype), Acanthamoeba sp. T11 genotype and Acanthamoeba spp. T4 genotype. One of the IS2404 positive agar plates that supposedly supported an amoeba culture did not contain amoebae at the time of IS2404 detection. Identification of the IS2404 negative amoebae will be detailed in a subsequent study by Amissah et al. (in preparation). The geographical origins of the IS2404 positive samples and amoeba cultures did not show any distribution pattern. We detected IS2404 in at least 1 sample and/or amoeba culture from all sampled localities, except Bebuso. As described in the methods section, subsamples were made to cultivate extracellular and intracellular mycobacteria. Twenty-six of the 148 samples were excluded from further analysis due to contamination of one or both of the subsamples. From 15 samples (12. 2%) only intracellular mycobacteria were isolated, from 17 samples (13. 9%) only extracellular mycobacteria, and from 32 samples (26. 2%) both intra- and extracellular mycobacteria were isolated. Details are given in Table 2. In general the difference between the isolation frequency of extracellular and intracellular mycobacteria was not significant (χ21 = 0. 17, p = 0. 89). To assess whether the intracellular life style was more frequent in certain sites or certain habitats, we determined the relative isolation frequency of intracellular mycobacteria (i. e. the number of samples from which intracellular mycobacteria were cultivated divided by the total number of samples from which we cultivated mycobacteria –intracellular and/or extracellular), and related this to BU endemicity, sampling sites and habitat type. The relative isolation frequency of intracellular mycobacteria did not differ between BU endemic and non-BU endemic areas (0. 77 vs. 0. 68; p = 0. 86, χ21 = 0. 03). The type of habitat, however, did have a significant effect on the relative occurrence of intracellular mycobacteria (p = 0. 002; χ23 = 15. 1): intracellular mycobacteria were more frequently isolated from detritus samples (relative isolation frequency of 0. 95) than from biofilm samples (relative isolation frequency of 0. 63; p = 0. 01). Based on a 821 to 837 bp portion of their 16S-rRNA gene sequence, 76 isolated mycobacteria (of intra- and extracellular origin) could be identified to the species level, with their sequence >99% identical to reference strains of which the sequence is present in GenBank. For 27 isolates, 16S rRNA-DNA sequence based identification was not possible due to the presence of a mixture of different species in the culture. An overview of the identified mycobacterial isolates is shown in Table 3 and Table S1. Species diversity did not show a marked difference between any type of isolation source (Table 3, S1). Mycobacterial 16S-rRNA-DNA was detected in 29 amoeba cultures (17. 5%), isolated from 25 out of 124 samples (20. 2%). Mycobacterial presence was confirmed by microscopy in 13 of these positive cultures; 1 to 100 AFB were detected per 100 fields, which approximates to orders of 103 to 105 bacilli per culture of amoebae. No AFB were observed co-localising with the amoebae, however.
Amoebae are good candidates to be a reservoir of the elusive M. ulcerans, but this relationship has not yet been thoroughly investigated. Here, we study the potential for amoebae to host M. ulcerans both experimentally as by sampling an aquatic environment. Our results show that M. ulcerans can indeed be phagocytosed in vitro by A. polyphaga and that viable bacilli persist for at least 2 weeks. We observed both tight and spacious phagosomal vacuoles containing M. ulcerans in infected A. polyphaga with transmission electron microscopy, as has been described for M. avium-infected A. castellanii [29], [47] and for macrophages infected with mycobacteria [50], including M. ulcerans [51]. The observed reduction in the number of viable bacilli is probably due to bacilli that are expelled by the amoebae after a phase of intracellular multiplication, as has been reported for in vitro mycobacteria-infected macrophages [52], [53]. The kanamycin in the medium therefore probably killed released bacteria and resulted in an underestimation of the capacity of M. ulcerans to grow inside the protozoan cells. Compared to a noninfected A. polyphaga monolayer, infection with the M. ulcerans strains did not result in a higher loss of cells (data not shown) indicating that the infection did not affect A. polyphaga viability. By analysing samples from an aquatic environment in BU endemic and nearby non-endemic communities in southern Ghana, we found several mycobacterium species intracellularly in eukaryotic micro-organisms. Most of the mycobacterium species we identified are potentially pathogenic to humans [54]–[57]. We did not isolate M. ulcerans, even not by successively passaging IS2404 positive specimens and amoeba cultures in mouse footpads, the method that has led to the only successful isolation of M. ulcerans from the environment [15] (data not shown). We isolated mycobacteria as frequently from an intracellular source as free-living, suggesting that it is quite common for several species of mycobacteria to infect micro-organisms in natural circumstances. The intracellular lifestyle was found significantly more frequent in detritus samples compared to water and biofilm samples. This could be due to the low oxygen levels in this organic debris. For several environmental bacteria (including M. avium) it has been shown that oxygen depletion (and other conditions that typically dominate in animal intestines) triggers the invasion of and enhances the survival within host cells [58], [59]. We detected the marker IS2404 in 1 water and 2 biofilm samples collected in a BU endemic and a nearby non-endemic community in southern Ghana. In addition, we detected the same marker in 6 amoeba cultures obtained from other samples. This is the first report of the detection of the marker IS2404, suggestive of M. ulcerans presence, in amoeba cultures isolated from the environment. It is noticeable that we tripled our detection frequency of IS2404 by searching in the amoeba cultures in addition to the original samples. We could not observe AFB in the smears of these amoeba cultures, but one must take into account that M. ulcerans and other IS2404 containing mycobacteria grow very slowly and thus were probably present in very low quantities on the amoeba cultures. On LJ-medium, M. ulcerans colonies only appear after an average of 10 weeks in primary culture from clinical specimens [60]. From environmental sources, M. ulcerans was only isolated once despite numerous attempts [15]. Other mycobacteria were also quite frequently detected in amoeba cultures (in 17. 5%), by a PCR assay targeting their 16S-rRNA gene. For these, mycobacterial presence could be confirmed by microscopy in almost half of the positive cultures. However, the AFB were not observed inside or attached to the amoebae. The fact that in our study mycobacteria could still be detected after multiple subcultures of the amoeba cultures, suggests that the mycobacteria were multiplying extracellularly on the agar plates. IS2404 was in fact also detected on one agar plate on which the amoebae did not survive subculturing. Similarly, in a co-culture study of M. avium and A. polyphaga, M. avium was shown to persist and multiply both intracellularly and extracellularly as a saprophyte on the excrement of A. polyphaga, and mycobacterial growth was most extensive extracellularly [47]. The successful uptake and persistence of M. ulcerans inside A. polyphaga in vitro and the higher detection frequency of IS2404 in amoeba cultures as opposed to the crude samples from the environment suggest that amoebae may act as a host for M. ulcerans in natural circumstances. However, our data do not reveal a significant role for protozoa in the distribution patterns of BU disease in humans, so we remain sceptical about their involvement in the direct transmission of M. ulcerans to humans. If a protozoan were to be principally responsible for the observed distribution pattern of BU in humans, one would expect either a particular species with a limited distribution to harbour M. ulcerans, or otherwise several species that only do so in areas where BU actually occurs in humans. In this study, however, we detected IS2404 as frequently in amoeba cultures isolated from BU endemic as from non-BU endemic communities. Moreover, 5 different protozoan species from two divergent families were identified in the IS2404 positive amoeba cultures, some of which are known to be cosmopolitan. On the other hand, we cannot completely rule out that some or all of the IS2404 we detected originated from different mycobacterial species than M. ulcerans. More environmental research is needed in Africa if we want to understand the distribution of BU, and to prevent its transmission from the environment to humans. Environmental research of M. ulcerans has been severely hampered by the difficulties of detecting the pathogen in the environment. Our results indicate that perhaps amoeba cultures can serve for improved detection of M. ulcerans in environmental samples. Co-cultivation with an existing amoeba culture is a technique to selectively isolate amoebae-resistant bacteria that are difficult to grow from the environment [61] and has already been proven successful in the identification of new pathogens and their distribution patterns [62]. | Buruli ulcer (BU) is a devastating skin disease caused by Mycobacterium ulcerans, an environmental bacterium that is probably linked to slow-running water. It is unlikely to occur free-living, but even though M. ulcerans DNA has been detected in quite a few different organisms (with most studies focusing on insects), it is still not clear what its real reservoir is. Amoeboid protozoa, inhabitants of biofilms in slow flowing water, are good candidates since all previously tested mycobacteria are resistant to the digestion by these macrophage-like organisms. In this paper we demonstrate that M. ulcerans can indeed infect Acanthamoeba polyphaga in the lab, and remain viable intracellularly. We also collected water, biofilm and detritus samples in BU endemic and non-endemic regions in Ghana. We found that several mycobacteria species commonly occur intracellularly in protozoa in these environments. Amoebae were isolated from almost all samples, and an M. ulcerans marker (IS2404) was detected in 4% of the amoeba cultures. We conclude that amoebae are potential hosts for M. ulcerans. However, because these IS2404 positive amoebae originated from both BU endemic and non-endemic areas, we remain sceptical about their implication in the transmission of M. ulcerans to humans. | Abstract
Introduction
Materials and Methods
Results
Discussion | medicine
public health and epidemiology
microbiology
bacterial diseases
emerging infectious diseases
neglected tropical diseases
applied microbiology
infectious diseases
buruli ulcer
mycobacterium
epidemiology
biology
microbial ecology
protozoology | 2012 | Amoebae as Potential Environmental Hosts for Mycobacterium ulcerans and Other Mycobacteria, but Doubtful Actors in Buruli Ulcer Epidemiology | 5,887 | 328 |
Detrimental inflammation of the lungs is a hallmark of severe influenza virus infections. Endothelial cells are the source of cytokine amplification, although mechanisms underlying this process are unknown. Here, using combined pharmacological and gene-deletion approaches, we show that plasminogen controls lung inflammation and pathogenesis of infections with influenza A/PR/8/34, highly pathogenic H5N1 and 2009 pandemic H1N1 viruses. Reduction of virus replication was not responsible for the observed effect. However, pharmacological depletion of fibrinogen, the main target of plasminogen reversed disease resistance of plasminogen-deficient mice or mice treated with an inhibitor of plasminogen-mediated fibrinolysis. Therefore, plasminogen contributes to the deleterious inflammation of the lungs and local fibrin clot formation may be implicated in host defense against influenza virus infections. Our studies suggest that the hemostatic system might be explored for novel treatments against influenza.
Influenza A viruses (IAV) are an important cause of outbreaks of respiratory tract infections and are responsible for significant morbidity and mortality in the human population [1]. Upon infection with IAV, innate and adaptive immune responses are induced that restrict viral replication and that afford protection against infection with these viruses. However, excessive inflammation, particularly in the lower respiratory tract, may result in alveolar damage limiting respiratory capacity and deteriorate the clinical outcome of IAV infections [2], [3]. Dys-regulation of cytokine production in the lungs is thus often associated with a fatal outcome of IAV [4]. The sites of virus replication in the respiratory tract represent complex microenvironments, in which extracellular proteases are present abundantly [5], [6]. Some of these proteases can play a role in innate immune responses since they are important mediators of inflammatory processes [7] and influence virus replication [8], [9]. To date, however, the elucidation of host proteases contributing to pathogenesis of IAV infections in vivo has been hampered by the lack of experimental models. One of the proteases of interest is plasmin, which is a serine protease involved in fibrinolysis, the biological process of dissolving fibrin polymers into soluble fragments. Plasmin is generated through cleavage of the proenzyme plasminogen, produced in the liver and present in the blood. Specific binding and conversion of plasminogen into plasmin by IAV may afford the virus an alternative protease for cleavage of its hemagglutinin molecule [10], [11]. This is an essential step in the virus replication cycle and this may contribute to the pathogenesis of IAV infection [12], [13]. In addition, plasminogen/plasmin plays a central role in fibrinolysis-mediated inflammation [14] and there is evidence of fibrinolysis activation during IAV infections [15]. Thus, plasminogen could contribute to the pathogenesis of IAV infections by promoting virus replication or by inducing a fibrinolysis-dependent harmful inflammatory response in the respiratory tract. At present it is unknown whether one or both of these two mechanisms of plasminogen activity contribute to pathogenesis of IAV infections in vivo. In the present study we address this research question and using plasminogen-deficient mice (PLG-KO) and pharmacological approaches the role of plasminogen during IAV infections was investigated. Our findings show that plasminogen plays an important role in lung inflammation upon IAV infections, mainly through fibrinolysis. Therefore, targeting host factors, such as the fibrinolytic molecule plasminogen may be of interest for the development of new therapeutics against IAV infections.
To explore the role of plasminogen in IAV pathogenesis, we investigated the consequence of plasminogen-deficiency. Plasminogen +/− mice were intercrossed to generate wild-type (WT) and plasminogen −/− (PLG-KO) mice, which were infected with IAV A/PR/8/34 (H1N1; 50,000 or 500 PFU) and weight loss and survival rates were monitored. As shown in Figure 1A, compared to WT mice, PLG-KO mice were significantly more resistant to IAV-induced weight loss and death. In PLG-KO mice substantial protection was also observed against infection with 2009 pandemic virus A/Netherlands/602/09 (30,000 PFU, Figure 1B) and highly pathogenic H5N1 virus A/chicken/Ivory-Coast/1787/2006 (10 EID50 H5N1, Figure 1C). Of note, the latter was not adapted to replicate in mammals, which could explain the delay in weight loss observed upon infection, as also observed by others [16]. Thus, we concluded that without plasminogen, pathogenesis of IAV infections was dampened and mortality reduced in a subtype-independent manner. To gain further insight into the role of plasminogen in virus replication, A549 cells were infected with IAV in the absence or presence of plasminogen. Interestingly, plasminogen supported the replication of IAV A/PR/8/34 but not that of A/Netherlands/602/09 (Figure 2A). In contrast, trypsin supported replication of both viruses while no replication was observed in absence of proteases. Since plasminogen promotes IAV replication through HA cleavage [11], plasminogen-mediated HA cleavage of both viruses was compared (Figure 2B). In absence of proteases (−), HA0 precursor protein was detected in A549 cells infected with either virus. In presence of plasminogen (PLG), an additional band, corresponding to HA2 [11] was detected at 25 kDa in A/PR/8/34, but not in A/Netherlands/602/09 infected cells. In presence of trypsin (Try), HA2 was detected in cells infected with either virus. Similar levels of tubulin were detected, which was included as control cellular protein. Thus, plasminogen promotes cleavage of HA of IAV A/PR/8/34 but not that of A/Netherlands/602/09, which correlated with differences in replicative capacity of these viruses in presence of plasminogen. On day 2 post-inoculation with IAV A/PR/8/34, mean lung virus titer of PLG-KO mice was significantly lower than that of WT mice (Figure 2C). This difference was not observed for IAV A/Netherlands/602/09. For both viruses, and at the other days post-infection, no significant differences in lung virus titers were observed between PLG-KO and WT mice. Thus, in vivo, plasminogen promoted early virus replication of IAV A/PR/8/34 but not of A/Netherlands/602/09. Since the absence of plasminogen protected mice against both viruses, the deleterious effect of plasminogen was most likely independent of virus replication in the lungs. To assess possible other contributions of plasminogen to the pathogenesis of IAV infections, inflammation of the lungs and viral dissemination were examined after infection of mice with IAV A/PR/8/34 or A/Netherlands/602/09. At day 3 post-infection, extensive alveolar damage and marked cellular infiltrates were observed in lungs of WT mice in contrast to those of PLG-KO mice (HE) after A/PR/8/34 virus infection (Figure 3A, left panel). This difference was also observed upon infection with A/Netherlands/602/09 virus, at day 5 (Figure 3A, right panel) but not at day 3 post-inoculation (data not shown). For all conditions, in WT and PLG-KO mice, similar numbers of IAV-infected cells were detected by immunohistochemistry (IHC). Also, no lesions were observed in Mock-infected mice (data not shown). Thus, plasminogen-deficiency protected mice against inflammation induced by A/PR/8/34 and A/Netherlands/602/09 viruses, showing that plasminogen plays a deleterious role in lung inflammation, independent of virus replication in the lungs. To investigate the difference in pulmonary inflammation between PLG-KO and WT mice, cytokine levels in the bronchoalveolar lavages (BALs) were assessed by ELISA (Figure 3B) or a luminex-based cytokine detection assays (Figure 4A) at various time point post-infection. Upon inoculation of A/PR/8/34 virus, both in PLG-KO and WT mice, BAL cytokine levels increased 2 and 5 days post-inoculation. However, in BAL of PLG-KO mice cytokine levels were considerably and significantly lower than in those of WT littermates (see scale differences for Figure 4A), which correlated with reduced IAV-induced lung inflammation in absence of plasminogen. Upon A/Netherlands/602/09 virus infection, release of cytokines in the BAL was also significantly higher in WT mice compared to PLG-KO mice at day 5 but not at day 2 post-inoculation (Figure 3B, right panel). Thus in concordance with the histological analysis, plasminogen promoted lung inflammation of IAV A/PR/8/34 and A/Netherlands/602/09 viruses, showing that the effect is most likely independent of virus replication in the lungs. Furthermore, in PLG-KO mice the virus failed to disseminate to extra pulmonary organs unlike in WT mice, upon intranasal infection with A/PR/8/34 virus (500 PFU) (Figure 4B). Especially high virus titers were detected in the liver, the source of plasminogen. Collectively, these results suggest that plasminogen plays an important role in promoting the inflammatory response and virus dissemination to extra-pulmonary organs during IAV-infection. Since degradation of fibrin is one of the main functions of plasminogen/plasmin, we hypothesized that the host fibrinolytic system plays a role in the pathogenesis of IAV infection. First, we investigated whether IAV infection induced fibrinolysis. To this end, mice were inoculated with IAV A/PR/8/34 and at various time points post-inoculation, the level of fibrinolysis markers in BALs was assessed by ELISA (Figure 5A). Plasminogen and active plasmin levels were barely detectable in the BAL of uninfected mice but their levels significantly increased during the course of infection. Levels of fibrinogen also significantly increased at day 4 post-infection and then dropped at days 5 and 6, suggesting a recruitment of fibrinogen to the lungs and a rapid consumption of the molecule and fibrinolysis. Finally, levels of FDP and D-dimers, degradation products of fibrinolysis, significantly increased upon infection, reaching 45 and 13 ng/ml respectively on day 6 post-inoculation. Similar results were also obtained upon infection with influenza virus A/Netherlands/602/09 (Figure 5A). As expected, in the BAL of infected PLG-KO mice, used as negative control, fibrinolysis markers were barely detectable. Thus, IAV infection induced fibrinolysis. These results were confirmed by Western blot analysis using an antibody directed against the mouse Aα chain of fibrinogen (Figure 5B), which recognizes purified mouse fibrinogen at a molecular weight of 66 kDa (data not shown). Compared to uninfected mice (−), fibrinogen was readily detectable 2–6 days post-inoculation in the lungs of infected mice. In the tissues, no marked fibrinogen consumption was detected but during the course of IAV infection, additional smaller bands corresponding to FDP were observed in mouse lungs. These findings confirmed that fibrinolysis took place during IAV infections in vivo. To simulate the depletion of fibrin (and therefore fibrinolysis), mice were treated with the snake venom Ancrod, a thrombin-like protease that cleaves the Aα chain of fibrinogen, enhancing its degradation and severely reducing its plasma levels (Figure 5C). Treatment with Ancrod significantly increased IAV-induced weight loss and mortality compared to vehicle-treated mice, but had no effect on uninfected control mice (Figure 6A). This increased mortality was also associated with an increase in inflammation of the lungs, as detected by elevated cytokine levels in the BAL (Figure 6B, WT). Of particular interest, the level of interferon-gamma was barely detectable in untreated mice but severely increased upon ancrod treatment. Thus, degradation of fibrin (ogen) contributed to inflammation and increased pathogenicity of IAV infection. Next, we investigated whether Ancrod treatment could reverse the protective effect of plasminogen-deficiency in terms of inflammation and mortality rate. Again, PLG-KO mice were protected from lung inflammation (p<0,05, between WT versus PLG-KO), as judged from cytokine responses (Figure 6B) and from IAV-induced mortality (Figure 6C). Interestingly, Ancrod-treatment reversed the protection observed in the absence of plasminogen and cytokine responses and mortality rates were similar to those of Ancrod treated WT mice (Figure 6B and C, p>0. 05, between WT-treated and PLG-KO-treated ancrod). Ancrod had no effect in uninfected mice (Figure S1). Thus, fibrinolysis contributes to inflammation and pathogenesis of IAV infections, which is mediated by plasminogen. To further confirm if the deleterious role of plasminogen is caused by fibrinolysis, we tested the outcome of infection of mice after treatment with Ancrod and/or 6-aminohexanoic acid (6-AHA). Indeed, 6-AHA is a lysine analogue that binds to the lysine binding sites of plasminogen, inhibiting plasminogen-binding to fibrin (ogen) and plasmin-mediated fibrinolysis [17]. First, 6-AHA treated mice inoculated with 5,000 or 500 PFU of A/PR/8/34 were significantly more resistant to infection than untreated mice (Figure 7A) and this protection correlated with reduced inflammation in 6-AHA treated animals (Figure S2). Also, lung virus titers were significantly lower in 6-AHA-treated mice compared to untreated mice, at day 2 but not at days 3 or 5 post-infection (Figure 7B). Thus, inhibition of plasminogen fibrinolytic activity protected mice from developing pneumonitis and severe disease. Furthermore, Ancrod-treatment of 6-AHA treated mice over-rode the protective effect of 6-AHA, again resulting in IAV-induced mortality (Figure 7A, lower panel). Administration of Ancrod and/or 6-AHA had no effect in uninfected mice (Figure S3). Thus, the protective effect of 6-AHA was reversed by Ancrod-mediated fibrinogen degradation, demonstrating that plasminogen contributed to pathogenesis of IAV infection through fibrinolysis activation. Preventing deleterious inflammation after IAV infection could be a promising new strategy to treat IAV infections. Therefore, we investigated whether blocking the fibrinbolytic activity of plasminogen by 6-AHA administration at a later time point post-inoculation was still protective. WT mice were inoculated with IAV A/PR/8/34 and treated or not with 6-AHA, two days later. As shown in Figure 7C, treatment with 6-AHA improved the outcome of infection and prevented mortality. 6-AHA treatment also protected mice from infection with A/Netherlands/602/09 and highly pathogenic H5N1 viruses (Figure 7C, lower panels). Thus, blocking plasminogen-mediated fibrinolysis protected mice against infections with various and highly pathogenic IAVs.
The present study showed for the first time that fibrinolysis plays a central role in the inflammatory response and the pathogenesis of IAV infections. Consistently, evidence is accumulating that the fibrinolytic molecule plasminogen and plasmin are critical host factors for immune cell infiltration and cytokine production upon injury [18]–[20]. The absence of plasminogen blunts inflammation in response to several inflammatory stimuli and suppresses development of lesions [21]–[23]. In our study, absence of plasminogen also considerably reduced the extent of lung inflammation upon IAV infection. Since severe inflammation contributes to the pathogenicity of IAV infections of humans [2], [4], most likely the proinflammatory properties of plasminogen play a role in the pathogenesis of these infections. IAV have the capacity to bind plasminogen and convert it into its active form plasmin through viral or cellular proteins like annexin-2 [11], [12]. However, the extent of plasminogen activation is strain-dependent [11], which may explain differences in pathogenicity of IAV strains. Mechanistically, the mode of action of plasminogen-driven lung inflammation was through fibrinolysis. Indeed, degradation of fibrinogen by Ancrod treatment increased pathogenicity of IAV infection and compensated the protective effect in PLG-KO mice or in mice in which plasminogen fibrinolytic activity was blocked by 6-AHA treatment. Consistently, Keller et al showed an activation of the fibrinolytic system during non-pathogenic IAV infection in mice [15]. Remarkably, in humans increased production of D-dimer, a marker of fibrinolysis was found to be a risk factor for fatal outcome of H5N1 and pandemic H1N1 virus infections [24], [25]. Furthermore, IAV infections have been associated with bleeding medical disorders [26], [27]. Thus, as for bacteria [28], the dysregulation of hemostasis by virus infections may cause serious complications. Consistent with our results, it was recently demonstrated that endothelial cells are central orchestrators of cytokine amplification during IAV infections [29]. Interestingly, plasminogen-dependent inflammation appears early after infection with influenza virus A/PR/8/34, of which virus replication is promoted by plasminogen. In contrast, replication of influenza virus A/Netherlands/602/09 is independent of plasminogen and control of plasminogen activity has a delayed impact on inflammation and disease. Thus, the capability of plasminogen to cleave HA and promote virus replication may also contribute to lung inflammation for some IAV strains. Possibly, a sustained high degree of inflammation is deleterious for the host. Collectively, we propose a model (Figure 8) in which plasminogen-mediated fibrinolysis increases FDP production and vascular permeability allowing increase recruitment of inflammatory cells at the site of infection. As a positive feedback loop, plasminogen mediated virus replication may also further contribute to lung inflammation. Fibrinolysis may also allow systemic haematogenous spread of virus. Consistently, we and others detected IAV replication in extrapulmonary organs in plasminogen-competent mice [30]. Since plasminogen is omnipresent in the blood, it may provide certain IAV an alternative mechanism of HA cleavage in extra-pulmonary organs [10], [11]. For example, the plasminogen-binding property of the neuraminidase of A/WSN/33 strain is a determinant of its neurotropism and pathogenicity in mice [12], [13]. Interestingly, particular high virus titers were found in the liver, which is the primary source of plasminogen. This may explain why IAV can replicate in hepatocarcinoma liver HEPG-2 cells in the absence of exogenous proteases (Figure S4). Whether plasminogen-dependent IAV replication contributes to damage of the liver or other extra-pulmonary organs, as observed in Reye' s syndrome or other postinfluenza complications [31] requires further investigation. Interestingly, differences in virus replication were not at the basis of plasminogen-dependent differences in pathogenesis of IAV infection although it also can contribute to exacerbation of inflammation. Indeed, A/Netherland/602/09 virus replication in the lungs was not affected by plasminogen deficiency, while infected PLG-KO mice were protected from infection. This is consistent with a recent report showing that presence of critical residues in HA, necessary for cleavage by plasmin is strain-dependent [32]. In addition, the HA of A/chicken/Ivory-Coast/1787/2006 contains a polybasic site, which is cleaved by furin-type proteases. This suggests that plasminogen plays a minor role in replication of this virus, while plasminogen deficiency still protected from infection with this virus. Alternative proteases may thus play a more dominant role in HA cleavage and virus replication in vivo than plasminogen [33]–[36]. For the clinical management of influenza patients, a limited number of antiviral drugs are available. The use of these currently available drugs is compromised by the emergence of virus strains that developed resistance to these drugs. Therefore, intervention strategies that aim at preventing deleterious inflammatory responses after IAV infection are of interest and do not suffer from resistance to antiviral drugs. Specifically, blocking protease activity may be an efficient way to achieve this, as previously suggested [37]–[39]. Our results are consistent with these studies but differ in term of mechanism of action. Indeed, our results suggest a more predominant role for proteases in lung hemostasis compared to virus replication and HA cleavage. In summary, our findings reveal a previously unrecognized role for fibrinolysis and plasminogen in the pathogenesis of IAV infections. Thus, targeting plasminogen, its conversion into plasmin or regulating fibrinolysis may be a venue for the development of novel intervention strategies for the treatment of severe IAV infections.
Experiments were performed according to recommendations of the “National Commission of Animal Experiment (CNEA) ” and the “National Committee on the Ethic Reflexion of Animal Experiments (CNREEA) ”. The protocol was approved by the committee of animal experiments of the University Claude Bernard Lyon I (Permit number: BH2008-13). All animal experiments were also carried out under the authority of license issued by “la direction des services Vétérinaires” (accreditation number 78–114). All efforts were made to minimize suffering. Viruses, cells, and reagents used, were: IAV A/Netherlands/602/09 [40], A/chicken/Ivory-Coast/1787/2006 [41], A/PR/8/34 (American Type Culture Collection, ATCC), A549 cells (ATCC), Madin-Darby Canine Kidney cells (MDCK, ATCC), trypsin (Becton Dickinson), plasminogen and 6-AHA (Sigma), Ancrod (NIBSC), 23-Plex Mouse Cytokine Assay (Bio-Rad), ELISA kits for mouse -IL-6, -KC, -–RANTES, -IFN-α -IFN-γ (R&D Systems), -plasminogen (Mybiosource), -active plasmin (Kordia), -D-dimer, -fibrinogen and -FDP (Genway), antibodies anti-HA (Santa Cruz), anti-tubulin (Sigma), anti-NP (ATCC), anti-fibrinogen (Genway). Blood fibrinogen and lung proteins were extracted as described [42], [43] and proteins were analyzed by western blot [44]. A549 experiments were performed as described previously [11]. Mice with a disrupted PLG gene (PLG-KO) and their WT littermates were bred as described previously [45]. Briefly, PLG heterozygous mice (C57BL/6 and 25% 129Sv) were crossed and WT and PLG-KO mice offspring were genotyped by polymerase chain reaction, which was performed, as previously described [46] using primers amplifying the WT PLG gene (5′ACTGCTGCCCACTGTTTGGAG 3′ and 5′ GATAACCTTGTAGAATTCAGGTC3′) or the inactivated PLG gene (5′ATGAACTGCAGGACGAGGCAG3′ and 5′ GCGAACAGTTCGGCTGGCGC 3′). Most of the experiments were performed using 5–6 weeks old mice. Also, males and females were used in the experiments. Groups between WT and PLG KO mice were homogenized for these different parameters. Except when PLG-WT and PLG-KO mice were used, experiments were performed with six-week-old C57BL/6 female mice (Charles River Laboratories). Mice were anesthetized with ketamine (42,5 mg/kg) and inoculated by the intranasal route with the indicated IAV in a volume of 25 µl. Upon inoculation, survival rates and loss of body weight was scored daily, as previously described [47]. For weight loss curves, the last measured value was carried forward until the end of the observation period. Alternatively, mice were sacrificed at various pre-fixed time points post-inoculation to perform bronchoalveolar lavages (BAL) or to sample organs. Virus titers in organs were determined by classical plaque assay using MDCK cells [47]. ELISA and luminex assays were performed according to the instructions of the manufacturer and virus titers were assessed as described [48]. Lungs histology and immunohistochemistry were performed as described [49]. Treatment with 6-AHA was injected intraperitoneally (30 mg per mouse in 200 µl of physiological serum) every 6 hours for 4 days. Ancrod was injected (1. 75 unit per mouse) intraperitoneally two days before infection for 7 days at 10 hours intervals. Kaplan-Meier test was used for statistical analysis of survival rates and Mann–Whitney' s test was used for lung virus titers and ELISA results, p values<0. 05, were considered statistically significant. Two-tails analysis was performed. The number (n) of animals per experimental group is mentioned in the figure legends. Experiments were stratified in terms of weight, gender and age of the mice. | Influenza viruses, including H5N1 bird influenza viruses continue to form a major threat for public health. Available antiviral drugs for the treatment of influenza are effective to a limited extent and the emergence of resistant viruses may further undermine their use. The symptoms associated with influenza are caused by replication of the virus in the respiratory tract and the host immune response. Here, we report that a molecule of the fibrinolytic system, plasminogen, contributes to inflammation caused by influenza. Inhibiting the action of plasminogen protected mice from severe influenza infections, including those caused by H5N1 and H1N1 pandemic 2009 viruses and may be a promising novel strategy to treat influenza. | Abstract
Introduction
Results
Discussion
Materials and Methods | immunopathology
medicine
pulmonology
virology
immunology
biology
microbiology
host-pathogen interaction
immunomodulation
respiratory medicine
hematology | 2013 | Plasminogen Controls Inflammation and Pathogenesis of Influenza Virus Infections via Fibrinolysis | 6,220 | 153 |
The phylum Apicomplexa comprises a group of obligate intracellular parasites that alternate between intracellular replicating stages and actively motile extracellular forms that move through tissue. Parasite cytosolic Ca2+ signalling activates motility, but how this is switched off after invasion is complete to allow for replication to begin is not understood. Here, we show that the cyclic adenosine monophosphate (cAMP) -dependent protein kinase A catalytic subunit 1 (PKAc1) of Toxoplasma is responsible for suppression of Ca2+ signalling upon host cell invasion. We demonstrate that PKAc1 is sequestered to the parasite periphery by dual acylation of PKA regulatory subunit 1 (PKAr1). Upon genetic depletion of PKAc1 we show that newly invaded parasites exit host cells shortly thereafter, in a perforin-like protein 1 (PLP-1) -dependent fashion. Furthermore, we demonstrate that loss of PKAc1 prevents rapid down-regulation of cytosolic [Ca2+] levels shortly after invasion. We also provide evidence that loss of PKAc1 sensitises parasites to cyclic GMP (cGMP) -induced Ca2+ signalling, thus demonstrating a functional link between cAMP and these other signalling modalities. Together, this work provides a new paradigm in understanding how Toxoplasma and related apicomplexan parasites regulate infectivity.
The phylum Apicomplexa comprises a large group of obligate intracellular parasites that cause many important human and livestock diseases and includes Plasmodium spp. (malaria), Cryptosporidium spp. (severe diarrhoea), and Toxoplasma gondii (toxoplasmosis). Toxoplasma is transmitted to humans by eating undercooked meat harbouring cyst forms or by consuming soil, vegetables, or water contaminated with oocysts shed from an infected cat. In healthy individuals, acute toxoplasmosis manifests with mild flu-like symptoms and is self-resolving; however, infection can cause life-threatening illness in the immunocompromised. Furthermore, infection during pregnancy can cause abortion early in gestation or severe neurological developmental abnormalities in the foetus. Vertical transmission is also considered the cause of the relatively high rates of progressive blindness in some countries [1,2]. Toxoplasma, like all apicomplexan parasites, critically requires the ability to move through tissue and invade host cells for survival and proliferation. Parasite movement is powered by a distinctive form of cellular locomotion referred to as ‘gliding motility’, which is powered by the glideosome, an actomyosin-dependent motor located underneath the plasma membrane [3,4]. Gliding motility is activated when transmembrane adhesins are released from the microneme organelles at the apical tip of the parasite and deposited onto the parasite surface. The current model then posits that the glideosome drives parasite motility when dynamic actin filaments attach to the cytoplasmic tails of transmembrane adhesins via the glideosome associated connector (GAC) [5]. This then allows Myosin A (MyoA), which is anchored in the parasite pellicle, to engage and drag actin-adhesin complexes rearward through the plane of the membrane, to the basal end, thus resulting in forward directional movement of the parasite through tissue and into host cells [3]. Parasite egress and motility are under tight control of intracellular signal transduction pathways so as to only activate at an appropriate time and switch off upon host cell invasion. Both microneme and glideosome activity are controlled by the cytosolic concentration of Ca2+ ([Ca2+]cyt), cyclic GMP (cGMP), and inositol signalling pathways [6–8]. The current model suggests that cGMP signalling is activated by extracellular signals and results in stimulation of inositol phosphate metabolism and subsequent release of Ca2+ from intracellular stores, as well as influx of Ca2+ from the external environment [9,10]. A rise in [Ca2+]cyt is temporally linked to the activation of egress from host cells and subsequent bursts of extracellular motility [11–13]. A rise in [Ca2+]cyt then triggers Ca2+-dependent protein kinases (CDPKs) and a calcineurin phosphatase, which then likely phosphorylate/dephosphorylate substrates, triggering downstream events, including activation of the glideosome and the release of adhesins from the micronemes [9,14–21]. To transition from immotile replicating tachyzoites to actively motile and invasive forms, parasites must sense the extracellular environment to regulate motility. High extracellular [K+], as encountered in the host cell, inactivates motility, whilst a drop in this cation causes a rise in [Ca2+]cyt and activation of locomotion [22]. More recently, it has been shown that extracellular pH ([H+]) also regulates motility. Here, low extracellular pH potently activates motility, whilst also being able to overcome the suppressive effect of high extracellular [K+] [23]. During intracellular growth, a lowering of the pH of the vacuolar space can activate Ca2+-dependent egress whilst also promoting an acidic environment for the activation of perforin-like protein 1 (PLP-1) to elicit membrane damage and tachyzoite egress [23]. Despite advances in our understanding of how Toxoplasma and other apicomplexan parasites activate motility, it remains unclear how these important pathogens sense environmental cues and how Ca2+ signalling is switched off. In other eukaryotic systems, G-protein coupled receptors (GPCRs) are essential for environmental sensing and act by receiving extracellular cues and transducing these signals across the plasma membrane. GPCRs physically associate with signalling proteins on the cytoplasmic side of the plasma membrane, causing changes in their activity upon extracellular cues. GPCRs commonly activate cyclic adenosine monophosphate (cAMP) signalling pathways. This occurs by direct coupling of GPCRs to adenyl cyclases (ACs), which are activated to produce cAMP upon encountering environmental stimuli. cAMP then activates a tetrameric protein kinase A (PKA) complex by binding to the regulatory subunit (PKAr), promoting its dissociation from the catalytic subunit (PKAc) and thus promoting downstream phosphorylation. Whilst genome sequencing has shown a complete lack of GPCRs in Apicomplexa, previous work has suggested that PKA, and thus cAMP signalling, is important for parasite invasion in several stages of the Plasmodium life cycle, in part through the phosphorylation of the cytoplasmic tail of the parasite adhesin apical membrane antigen 1 (AMA1) [24–26] and may be linked to Ca2+ signalling [15]. Interestingly, one of three PKA paralogues in Toxoplasma negatively regulates bradyzoite differentiation [27]. Here, we show that protein kinase A catalytic subunit 1 (PKAc1) acts as a negative regulator of tachyzoite egress immediately following invasion and is required for the transition of invasive tachyzoites into intracellular replicative forms. Whilst PKAc1-depleted tachyzoites appear to invade normally and form a parasitophorous vacuole (PV), they rapidly egress from host cells shortly thereafter. We show that in the absence of PKAc1, newly invaded tachyzoites cannot rapidly quench [Ca2+]cyt, leading to a persistent raised level. Furthermore, we provide evidence that PKAc1 dampens cGMP signalling to supress [Ca2+]cyt. Together, our work pinpoints a critical role of PKAc1, and thus cAMP signalling, in negatively regulating [Ca2+]cyt upon parasite invasion. Our work therefore provides evidence of how Toxoplasma, and potentially other apicomplexan parasites, switch off motility upon successful invasion of a host cell.
We wished to understand if cAMP/PKA signalling participates in relaying environmental cues to modulate motility. Toxoplasma contains three annotated PKA catalytic subunit paralogues (toxodb. org) [28]. Previous work has shown that PKAc3 (TGME49_286470) negatively regulates the differentiation of bradyzoite forms during tachyzoite growth [27], whilst PKAc2 (TGME49_228420) appears not expressed in tachyzoite stages (toxodb. org). PKAc1 (TGME49_226030) is the most widely conserved Toxoplasma paralogue across the Apicomplexa and thus was chosen for study. To understand the function of PKAc1, we genetically introduced an epitope tag at the 5′ end of the protein, whilst simultaneously putting the gene under conditional genetic control using the tetracycline-off (tet-off) system (S1 Fig) (see below for description). To assess the localisation of PKAc1, we performed an immunofluorescence assay (IFA) using anti-haemagglutinin (HA) epitope antibodies and observed a peripheral staining pattern, as marked by inner membrane complex 1 (IMC1) protein antibodies [29] (Fig 1Ai). In some tachyzoites, we noticed internal staining, which colocalised with the apical IMC sub-compartment protein 1 (ISP1) marker [30], suggesting association with the forming daughter cells during replication (Fig 1Aii). Across eukaryotes, PKA typically localises to defined cellular compartments via association with A-kinase-associated proteins (AKAPs) and/or by binding to GPCRs. To determine if this was true in Toxoplasma, we performed immunoprecipitation of HA-tagged PKAc1 and compared eluates to parental controls either by in-solution trypsin digestion (Fig 1B) or in-gel digests from SYRPO Ruby-stained gels, followed by mass spectrometry to detect associated proteins (Fig 1C). In both cases we only confidently detected one additional protein, which we determined to be TGME49_242070, a protein with several predicted cAMP-binding domains. The Toxoplasma genome (toxoDB. org) predicts the presence of two regulatory PKA subunits (the other being TGME49_311300), and thus this work experimentally verifies that TGME49_242070 is most likely the regulatory PKA subunit for PKAc1 –thus we named this orthologue PKAr1. PKAr subunits in other eukaryotes contain an N-terminal docking domain that is responsible for associating the PKA complex with AKAP. Toxoplasma PKAr1 appears to lack this domain and instead contains a glycine at position 2 and two cysteines at positions 5 and 7, typical of sites of acylation. Acylation is an important mechanism for imparting membrane affinity on a range of proteins involved in signalling and motility in Toxoplasma [31–34]. To investigate the localisation of PKAr1 without disturbing this potential motif, we introduced a Ty epitope tag 15 amino acids downstream from the starting methionine. Western blot of Ty-PKAr1 revealed that there were two major species of this protein close to the expected size, suggesting that dual acylation could be imparting changes to the migration of the protein (Fig 1D). There are also two weak bands at about 100 kD, but the identity or significance of these is not known. Colocalisation of Ty-PKAr1 by IFA with IMC1 also demonstrated a largely peripheral localisation and further, we also observed localisation to the nascent IMC (as marked by ISP1) during internal daughter cell budding (Fig 1Ei and 1Eii), similar to that observed with PKAc1. PKAr1 was recently described to be part of the Toxoplasma ‘palmitome’, thus further supporting the evidence that this protein is acylated [35]. To investigate whether the N-terminal region of PKAr1 was responsible for sequestration to the periphery by myristoylation of glycine 2 (G2) and palmitoylation of cysteine 5 (C5) and/or 7 (C7), we fused the first 15 amino acids of PKAr1 to green fluorescent protein (GFP) and used IFA to monitor localisation and western blot to assess any changes in migration pattern. We found that, indeed, PKAr1 (1–15) was sufficient to target GFP to the parasite periphery (Fig 1Fi) and to internal budding daughter cells (Fig 1Fii). Furthermore, mutation of the putative myristoylation site to an alanine (G2A) resulted in a severe abrogation of peripheral localisation and instead caused an accumulation of the GFP fusion protein in the cytoplasm and nucleus (Fig 1Fiii). Mutation of the two putative palmitoylation acceptor cysteines to alanines (C5A, C7A) also resulted in abrogation of peripheral localisation (Fig 1Fiv), as did mutation of all three residues simultaneously (G2A, C5A, C7A) (Fig 1Fv). Protein acylation leads to changes in hydrophobicity and propensity to bind detergents, leading to change in apparent molecular weight on SDS-PAGE. To further implicate myristoylation and palmitoylation in PKAr1 membrane affinity, we monitored changes in protein migration of PKAr1 (1–15) WT-GFP and point mutants by western blot (Fig 1G). Here, we observed a single molecular weight species in PKAr1 (1–15) WT-GFP, which, upon mutation of the glycine at position 2 to alanine (PKAr1 (1–15) G2A-GFP), resulted in an additional higher molecular weight species, consistent with myristoylation causing faster migration (Fig 1G). Mutation of both putative myristoylation and palmitoylation sites (PKAr (1–15) G2A C5A C7A-GFP resulted in complete loss of faster migrating species (Fig 1G). Mutation of the putative palmitoylation sites (PKAr (1–15) C5A C7A-GFP) does not appear to affect SDS-PAGE migration as compared to the wild-type sequence, suggesting that myristoylation is able to occur in the absence of palmitoylation sites (Fig 1G). Together, these data are consistent with PKAc1 localising to the parasite periphery, likely the IMC membranes, via its association with PKAr1, which relies on dual acylation to derive membrane binding. To investigate the function of PKAc1 in Toxoplasma tachyzoites, we generated a PKAc1 conditional knockdown (cKD) line by replacing the endogenous promoter with the tet-off tetO7sag4 (T7S4) promoter (S1 Fig) (Fig 2A) [36,37]. To monitor the regulation of PKAc1 in the PKAc1 cKD parasites, we tracked protein production by the N-terminally appended HA epitope tag using HA antibodies. Treatment of parasite cultures with anhydrotetracycline (ATc) for 48 hours led to a marked decrease in PKAc1 levels (Fig 2B), and IFA showed loss of detectable levels of this protein (Fig 2C). To monitor growth of PKAc1 cKD parasites across the complete lytic cycle, we then performed plaque assays on confluent monolayers of human foreskin fibroblasts (HFFs). ATc treatment of PKAc1 cKD, but not parental, parasites resulted in a drastic reduction in plaque size, consistent with this kinase having an important role in one or more steps of the Toxoplasma lytic cycle, and furthermore is congruent with its deleterious clustered regularly interspaced short palindromic repeats (CRISPR) screen growth score [38] (Fig 2D). During our preliminary analysis, we noticed that in vitro cultures of PKAc1-depleted tachyzoites growing in host cells had an unusual presentation. After 24 hours of ATc treatment, PKAc1 cKD cultures contained very few intracellular parasites (compared to untreated cultures). Concurrently, host cells appeared to have been lysed, some detaching from the surface plastic (Fig 2E and S2 Fig). We hypothesised that the unusual presentation of host cells in cultures infected with PKAc1-deficient tachyzoites was due to aberrant invasion. We therefore performed two-colour invasion assay to enumerate invasion efficiency on a population level. After a 10-minute invasion period we observed a stark loss of intracellular parasites (Fig 3A). To get a better understanding of the role of PKAc1 during invasion we undertook live cell imaging. In ATc-treated or untreated parental lines and untreated PKAc1 cKD tachyzoites we saw typical invasion, in which parasites attached to host cell monolayers, activated motility, and penetrated the host cell through a tight constriction, typical of the formation of the moving junction (Fig 3Bi, Bii and 3Biii; S1, S2 and S3 Movies). Following invasion, tachyzoites remained stationary and little movement of the host cell was observed (invading tachyzoites marked with white arrowhead and dashed white lines mark outline of host cell) (Fig 3Bi, 3Bii and 3Biii; S1, S2 and S3 Movies). Upon depletion of PKAc1, shortly after the apparent completion of invasion, we saw that the host cell began to detach from the glass slide surface, often with the concomitant reactivation of parasite motility (Fig 3Ci and 3Cii, S4 and S5 Movies). We observed that tachyzoites could directly exit host cells (Fig 3Cii, S4 Movie) and in some cases move within the confines of the damaged host cell (Fig 3Ci, S5 Movie). We then quantified the timing of host cell collapse in relation to invasion over a larger population of cells. Over a period of 10 minutes of filming we saw no parental (+ATc) or PKAc1 cKD (−ATc) egress (Fig 3D). In comparison, some PKAc1-deficient tachyzoites exited as quickly as about 30 seconds post-invasion, whereas others did not exit within the 10-minute filming period. Over the population, host cell collapse occurred at an average of about 200 seconds post-invasion (Fig 3D). We then quantified host cell damage on a population level, as a function of parasite concentration, by monitoring host cell integrity using crystal violet staining. To do this, ATc treated and untreated parental and PKAc1 cKD tachyzoites were serially diluted and allowed to invade host cells for 3 hours, after which time host cell integrity was analysed by crystal violet staining (Fig 3E). The number of intact host cells decreased when incubated with increasing concentration of PKAc1-depleted tachyzoites. In contrast, PKAc1 cKD parasites expressing PKAc1 and parental parasites treated with ATc caused no such damage (Fig 3E). Together, these results suggest that PKAc1 is required for productive invasion and/or the suppression of motility once cell entry is complete. The exit of host cells by PKAc1-depleted tachyzoites shortly after invasion could be caused either by a defect in the formation or sealing of the parasitophorous vacuole membrane (PVM) during invasion, or by parasites activating egress shortly after internalisation is complete. To test if PKAc1 has a role in invasion, we measured the speed of invasion (S3A Fig) and the ability of invading tachyzoites to form a moving junction, as marked by rhoptry neck 4 (RON4) protein staining (S3B Fig). In both cases we saw no difference between parental and PKAc1-depleted lines (representative images shown of RON4 staining). We then determined whether PKAc1 plays a role in PVM formation. Secretion of rhoptries is required for PVM formation and can be measured by inhibiting invasion using cytochalasin D (CytD) and staining for the presence of ‘empty vacuoles’ (evacuoles) [39], using rhoptry protein 1 (ROP1) as a marker. Furthermore, we used microneme protein 8 (MIC8) cKD as a positive control, as this protein is known to be important in evacuole production [40]. Here, we observed that PKAc1-depleted tachyzoites were able to produce evacuoles at the same rate as parental lines, suggesting that this kinase plays no role in the secretion of rhoptry contents (S3C Fig). We also measured formation of the PVM on invading and invaded parasites using a mouse embryonic fibroblast (MEF) line that expresses membrane-bound tandem dimeric tomato (tdTomato) red fluorescent protein. [41]. In static images, in which PKAc1 and parental lines were expressing cytosolic GFP, we observed clear formation of an intense tdTomato+ membrane around the body of tachyzoites (S3Di Fig), as clearly demonstrated by plotting intensity values across a cross section (S3Dii Fig). Furthermore, when performing live cell imaging on tachyzoites invading tdTomato-expressing MEFs, we saw formation of a tdTomato+ membrane surrounding invading parasites, which persisted until collapse of the host cell (see S3E Fig and S6 Movie for representative time-lapse), further suggesting the formation of a PVM even in the absence of PKAc1. We also assayed whether PKAc1 was involved in motility and host cell attachment, as well as microneme secretion (S4 Fig). To assay tachyzoite motility, we performed live cell imaging and quantitated total motile and nonmotile fractions as well as segregating based on the different types of motility when put in intracellular (IC) or extracellular (EC) buffer. Whilst we found an increase in the fraction of motile tachyzoites in EC buffer, we saw no changes upon repression of PKAc1 expression (S4Ai and S4Aii Fig). We also measured tachyzoite host cell attachment using standard assays and again found no differences of PKAc1-deficient parasites, as compared to controls (S4B Fig). Furthermore, we assayed secretion of micronemal proteins into the supernatant using standard conditions (in Ringer’s buffer), using both quantitative western blot and quantitative proteomics. In both cases, we found that PKAc1-depleted parasites had no difference to parental and no ATc controls (S4C, S4D and S4E Fig). Note that when performing quantitative proteomics on supernatants, we did see more peptides from proteins known or predicted to be found inside the parasite. Whilst we do not know the reasons behind this finding, we suggest that loss of PKAc1 may make the plasma membrane more fragile and prone to lysis (S4E Fig). Together, these results suggest that PKAc1 has no detectable role in invasion, motility, or host cell adhesion. We next determined whether PKAc1 was required for programmed host cell egress. Toxoplasma PLP-1 is critical to induce PVM and host cell membrane breakdown required for parasite egress [42,43]. Furthermore, vacuolar acidification is required for activating PLP-1 [23] to elicit membrane damage. We therefore wondered whether early egress of PKAc1-deficient tachyzoites requires acidification of the vacuolar space and the cytolytic activity of PLP-1. Furthermore, given that PLP-1 has no role in invasion [43], testing for dependency on PLP-1 and vacuolar acidification of the premature egress phenotype of PKAc1-depleted parasites would support a role for this kinase as a negative regulator of egress post-invasion. To test this, we first genetically deleted PLP-1 in our PKAc1 cKD line (S5 Fig). Western blot of PKAc1 cKD/Δplp-1 parasite lysates using PLP-1 antibodies showed a loss of signal verifying gene disruption (Fig 4A). To determine if PLP-1 is required for early egress of PKAc1-deficient parasites, we grew PKAc1 cKD and PKAc1 cKD/Δplp-1 tachyzoites in the presence of ATc for 24 hours and first monitored the morphology of parasite-infected host cells under the microscope. As previously shown in Fig 2, loss of PKAc1 results in collapsed host cells that appear to readily detach from the substrate, with no intracellular parasites evident (Fig 4B, black arrows). Conversely, host cells infected with parasites also lacking PLP-1 (PKAc1 cKD/Δplp-1 +ATc) appeared to have a reversal of this effect, containing many late-stage vacuoles full of tachyzoites, similar to parental and non-ATc-treated lines (Fig 4B, white arrows and S2 Fig). This suggests that loss of this cytolytic protein at least partially rescues genetic depletion of PKAc1. To quantify the role of PLP-1 in host cell destruction post-invasion in PKAc1-deficient tachyzoites, we performed an end-point invasion assay. As compared to the loss of PKAc1 alone (as reproduced from above), the additional deletion of PLP1 resulted in the presence of significantly more intracellular parasites at 10 minutes post-invasion (Fig 4C). As a control, we confirmed that loss of PLP-1 alone has no significant effect on invasion [43]. Previously, it has been shown that the inhibitor of H+ ATPase transporters N, N′-Dicyclohexylcarbodiimide (DCCD) can prevent vacuolar acidification and inhibit PLP-1 activity. We therefore also tested whether treatment of DCCD could reverse the loss of PKAc1, thus further supporting a role for this kinase in supressing egress post-invasion. Using a standard invasion assay, we observed that whilst DCCD treatment can inhibit invasion, it also mildly rescues the proportion of observable intracellular parasites of PKAc1-depleted tachyzoites after 10 minutes of invasion (Fig 4D). Furthermore, there was no difference in the amount of intracellular tachyzoites with and without PKAc1 that were both treated with DCCD (Fig 4D). This suggests that a H+ ATPase activity could play a role in vacuolar acidification, which drives re-exit of PKAc1-depleted Toxoplasma from host cells. It is important to note that DCCD treatment is not very specific and is likely inhibiting other H+ ATPases (e. g. , mitochondrial), which is affecting the rate of invasion. To confirm these findings and further dissect the PKAc1-depletion phenotype dependency on PLP-1 and acidification, we performed live cell imaging on PKAc1 cKD/Δplp-1 tachyzoites and quantitated how long tachyzoites remained intracellular following invasion. As compared to PKAc1-deficient parasites, the additional loss of PLP1 saw the vast majority of parasites remain intracellular over the 10 minutes of filming (Fig 4E). Tracking of individual PKAc1-depleted tachyzoites treated with 50 μM of DCCD showed that almost all parasites remain intracellular after invasion over the 10-minute filming period, suggesting that this compound is able to reverse the loss of PKAc1 of those parasites that are able to invade. Together, these data suggest that PKAc1 has a role in negatively regulating host cell egress and demonstrate that PKAc1-dependent egress from host cells occurs in a PLP-1-dependent fashion. Tachyzoite [Ca2+]cyt has been temporally and functionally linked to parasite egress, motility, and invasion [11,12]. We hypothesised that PKAc1 may function to directly or indirectly negatively regulate [Ca2+]cyt, such that loss of this kinase sustains parasite motility and enables host cell egress to occur. Previously, we have used the genetically encoded biosensor, GFP-Calmodulin-M13-peptide-6 (GCaMP6), to monitor [Ca2+]cyt in intracellular, egressing, and extracellular motile tachyzoites [11,44]. To first assess the utility of GCaMP6 in monitoring Ca2+ levels during invasion, we stably introduced the GCaMP6/mCherry-expressing plasmid at the uprt locus into the parental line (Δku80: TATi) and quantitated fluorescence levels (+ATc treatment) over a 10-minute period, acquiring five z-projections approximately every 1. 3 seconds (Fig 5, S7 Movie). We observed that invasion of the parental line coincides with a rapid quenching of GCaMP6 fluorescence, dropping to 20%–40% of maximum levels in a period of 10–20 seconds post-invasion (t = 0 = beginning of invasion; blue arrow = completion of invasion; red arrow = time to reach 35% of maximum fluorescence) (Fig 5Ai and 5Aii, S7 Movie). We found that this pattern was seen in all observable invasion events in the parental line (Fig 5Aiii; for individual traces, see S6 Fig). We then introduced the GCaMP6/mCherry-expressing plasmid into PKAc1 cKD parasites and measured the dynamics in fluorescence as a readout for changes in [Ca2+]cyt. In the absence of ATc, PKAc1 cKD tachyzoites were able to quench [Ca2+]cyt shortly after invasion, similar to the parental line (Fig 5Bi, 5Bii and 5Biii, and S8 Movie; individual traces are shown in S7 Fig). However, depletion of PKAc1 by ATc treatment resulted in drastic changes in [Ca2+]cyt dynamics. Here, we noticed that PKAc1-depleted tachyzoites could not rapidly dampen GCaMP6 fluorescence after invasion was complete (green arrow = time of host cell collapse) (Fig 5Ci and 5Cii). Instead, as viewed by individual traces of PKAc1-depleted tachyzoites, [Ca2+]cyt varied greatly during this time (Fig 5Ciii and S8 Fig). In order to quantitate and graphically represent cytosolic Ca2+ dynamics over time, we chose, based on parental controls, a value of 35% of maximum fluorescence to baseline after the completion of invasion (Fig 5Aii, 5Aiii, 5Bii and 5Biii). By measuring the time taken for tachyzoites to reach 35% of maximum, we could observe that PKAc1-depleted parasites took significantly longer than untreated or parental controls to reduce [Ca2+]cyt, averaging about 150 seconds to reach 35%, as compared to controls, which typically took 20–40 seconds (Fig 5D). We then measured GCaMP6 fluorescence level at 100 seconds, a time when all tachyzoites of control samples have completed invasion and dampened [Ca2+]cyt (Fig 5A and 5B). Tachyzoites lacking PKAc1 expression were then split into those that egressed before and after t = 100, as well as those that did not egress at all. In doing so, we could see that those PKAc1-deficient parasites that did egress had significantly higher GCaMP6 fluorescence at this time point, whilst those that did not had levels equivalent to controls (Fig 5E). Overall, this work suggests that PKAc1 plays an important role in the rapid reduction of [Ca2+]cyt that normally takes place shortly after invasion is complete. We wondered whether [Ca2+]cyt in PKAc1-deficient parasites was influenced by PLP-1-dependent egress and exposure to the extracellular environment. To test this, we stably introduced GCaMP6/mCherry-expressing plasmid into the uprt locus of PKAc1 cKD/Δplp1 tachyzoites (Fig 6). In the absence of ATc, PKAc1 cKD/Δplp1 tachyzoites invaded host cells in a typical fashion, which was followed by rapid dampening of [Ca2+]cyt, indistinguishable from parental controls (Fig 6Ai and 6Aii), which was highly consistent across the population (Fig 6Aiii). Upon depletion of PKAc1 with ATc, we saw that GCaMP6 levels did not dampen in a typical fashion, demonstrating that the cytolytic activity of PLP-1 and thus early exposure to the extracellular environment play little role in the loss of [Ca2+]cyt dampening in PKAc1-deficient parasites (Fig 6Bi, 6Bii and 6Biii). To further explore the independence of [Ca2+]cyt from membrane disruption by PLP-1, we treated parasites with the H+ ATPase inhibitor DCCD (Fig 6C). After tracking many invading tachyzoites, we saw that DCCD treatment of PKAc1-deficient parasites largely phenocopied PLP-1 deletion in terms of loss of egress, but sustained high [Ca2+]cyt (Fig 6Ci, 6Cii and 6Ciii), thus further supporting the notion that regulation of [Ca2+]cyt is not influenced by cytolytic activity of PLP-1. We also quantitated these effects by graphing time taken of each treatment to reach 35% of maximum fluorescence (Fig 6D), as well as fluorescence levels at t = 100 seconds (Fig 6E). As compared to PKAc1 depletion alone (redisplayed here from Fig 5), this showed that PLP-1-deficient and DCCD-treated PKAc1-deficient tachyzoites maintain a higher [Ca2+]cyt level post-invasion (Fig 6D and 6E). Together, these data provide strong evidence that PKAc1 negatively regulates [Ca2+]cyt post-invasion, independent of PLP-1-dependent PVM and host cell lysis. To further dissect the role of PKAc1 in negatively regulating [Ca2+]cyt, we investigated how the extracellular environment affects cytosolic concentrations of Ca2+. First of all, we tested if PKAc1 controls Ca2+ transport across the plasma membrane. To test this, we loaded extracellular tachyzoites with the Ca2+-responsive dye Fura-2 and performed calibrated measurements on parasite populations using fluorometry. We found that, when suspended in a Ca2+-free buffer, PKAc1-depleted parasites had a statistically higher [Ca2+]cyt, containing nearly 2-fold higher levels (about 58 nM compared with about 33 nM) (Fig 7A). PKAc1-depleted parasites were also observed to have higher [Ca2+]cyt than PKAc1-expressing parasites when suspended in salines containing [Ca2+] of 1 μM, 100 μM, and 1 mM; however, in these cases, statistical significance was not reached (Fig 7A). We also developed a fluorescence-activated cell sorting (FACS) -based assay to measure [Ca2+]cyt using GCaMP6/mCherry-expressing parasites. This allowed us to further probe the relationship between environmental cues and the role of PKAc1 in negatively regulating [Ca2+]cyt. To do this, we resuspended tachyzoites in buffers mimicking the extracellular (EC) and intracellular (IC) environments and tested the role of PKAc1 in regulating [Ca2+]cyt in these conditions (Fig 7B). Here, we could show that PKAc1-deficient parasites, even in IC buffer, had a significantly higher GCaMP6/mCherry signal, signifying a higher [Ca2+]cyt than either PKAc1 cKD −ATc or parental controls (Fig 7B), suggesting that this cAMP-dependent protein kinase is critical for supressing [Ca2+]cyt in intracellular conditions. We then used live cell imaging to specifically interrogate any change in [Ca2+]cyt in motile parasites in the presence and absence of PKAc1. Quantitating both maximum and minimum fluorescent values in motile parasites, we could show that PKAc1-deficient parasites only had statistically significant differences in minimum fluorescent levels in IC buffer (S11 Fig). Overall, this analysis suggests that PKAc1 plays an important role in regulating the resting level of [Ca2+]cyt in conditions that mimic both the extracellular and intracellular environments. Previous work has shown that cGMP signalling positively regulates Ca2+ signalling in both Toxoplasma and Plasmodium spp. [11,45–47]. We therefore wondered whether PKAc1 controls [Ca2+]cyt by regulating cGMP signalling. Furthermore, it has recently been shown that a putative cyclic nucleotide phosphodiesterase (PDE) in Toxoplasma has several PKAc1-dependent phosphorylation sites and a specific inhibitor ‘Compound 1’ of cGMP-dependent Protein kinase G (PKG) blocks premature egress in PKAc1 mutant parasites [48]. Therefore, it stands to reason that PKAc1 could act to control [Ca2+]cyt by negatively regulating cGMP signalling. To test this, we developed a FACS-based assay to dynamically monitor [Ca2+]cyt when cGMP signalling is activated, using the PDE inhibitor 5-benzyl-3-isopropyl-1H-pyrazolo[4,3-d]pyrimidin-7 (6H) -one (BIPPO) [49]. We first measured basal GCaMP6 fluorescence in PKAc1 cKD and parental lines and then subjected them to a titration of [BIPPO] and measured fluorescence over time. In doing so, we were able to generate response curves over time and at a set time point (t = 50 seconds) (Fig 8A and 8B). Whilst we were able to show that the parental strain with or without ATc and PKAc1 cKD −ATc behaved similarly to a BIPPO titration, we found that PKAc1 depletion led to a greater rise in fluorescent levels at a given concentration. When subtracting background (to remove any effect of high resting [Ca2+]cyt in PKAc1-depleted parasites) and plotted as a dose–response curve at t = 50 seconds over three independent experiments, it can be seen that loss of PKAc1 expression leads to a greater response (Fig 8Ci). Furthermore, we could show that this phenomenon is specific to use of BIPPO, as this is not seen when the same experiments were performed using the Ca2+ ionophore A23187 (Fig 8Cii). Overall, this work provides a functional link between PKAc1 and cGMP signalling to negatively regulate [Ca2+]cyt.
Critical to the establishment and perpetuation of infection by apicomplexan parasites is their ability to sense and respond to a changing host environment. Environmental cues allow for parasites to undergo differentiation when encountering a new host or to activate motility and egress at an optimal time during host cell infection cycles. Parasites must also switch off motility when invasion of a new host cell is complete in order to start replication. Several environmental factors have been identified that activate differentiation and motility in Plasmodium spp. and include blood lysophosphatidylcholine and mosquito xanthurenic acid, which activate sexual stage commitment and differentiation, respectively [50,51], whereas low extracellular [K+] [52,53] stimulates microneme secretion in blood and liver stages, and low pH (high [H+]) activates cell traversal in liver stages [54]. Toxoplasma too activates motility upon a drop in extracellular [K+] or pH, suggesting that sensing of some environmental cues is conserved across the Apicomplexa [22,23]. Despite the identification of these environmental cues, it is still not understood how apicomplexan parasites transduce these signals across the plasma membrane to induce Ca2+-dependent motility and differentiation. cAMP/PKA signalling is commonly used to relay environmental changes and regulate cellular responses in eukaryotes. This typically occurs when GPCRs receive extracellular signals and activate ACs to produce cAMP or, alternatively, prevent its breakdown by inhibiting cAMP-specific 3′ 5′-PDEs. Despite apicomplexans having a striking paucity of GPCRs, we wanted to see if Toxoplasma uses cAMP/PKA signalling to relay environmental cues. Here, we identified a PKA orthologue—PKAc1—that localises to the parasite periphery, and performed immunoprecipitation of PKAc1 to identify interacting proteins. Using both gel- and whole eluate–based methods of protein identification, we could only robustly identify a likely regulatory subunit, PKAr1, suggesting that, unlike PKA in other eukaryotic systems, Toxoplasma PKAc1 does not stably interact with ACs, PDEs, or GPCRs. Furthermore, we demonstrate that PKAr1 likely derives membrane affinity through myristoylation and palmitoylation at its N-terminus, rather than through interactions with AKAP proteins, as is the case in metazoans. Together, this work suggests that PKAc1/PKAr1 either interacts with other proteins in much less stable complexes or that cAMP-dependent signalling in Toxoplasma is somewhat different from that in metazoans. In this regard, it is also interesting to note that some ACs (as well as guanylate cyclases) in Apicomplexa appear to be directly fused to ion transporter-like domains, suggesting a novel mechanism employed by this group of parasites to allow for sensing changes in extracellular ion concentrations and relaying these across the membrane to initiate intracellular signalling cascades [26,55]. Controlled depletion of PKAc1 highlighted the importance of this kinase during the Toxoplasma lytic cycle and revealed a unique phenotype, whereby tachyzoites rapidly exit the host cell shortly after invasion. In P. falciparum, PKA phosphorylates serine 610 (S610) on the cytoplasmic tail of the adhesin AMA1, and genetic mutation of this phospho-site leads to defects in invasion [25]. Given these findings in P. falciparum and the role that AMA1 plays in formation of the moving junction, we expected to find defects in invasion or sealing of the PVM in Toxoplasma PKAc1 mutants. However, using available assays and monitoring the formation of the PVM in real time using live cell imaging, we could not detect any apparent defects and instead could show that egress from host cells occurs well after invasion is complete, sometimes 200 seconds after. Supporting this notion, we demonstrated that loss of PLP-1 reverses loss of PKAc1 during host cell invasion, suggesting that aberrant activation of this cytolytic protein in this mutant is responsible for the phenotype. In saying this, the assays that we have available to us to monitor PVM biogenesis are not sensitive enough to determine if there is also a defect in sealing, and thus this remains a possibility that cannot be ruled out. It is also interesting to note that the cytoplasmic tail of AMA1 in Toxoplasma contains an aspartic acid at residue 557, the residue which is equivalent to S610 in P. falciparum [25]. Aspartic acids share physicochemical properties with phosphorylated serine residues, and thus Toxoplasma may have lost, or never gained, the capacity to regulate AMA1 tail function by PKA phosphorylation. It will be interesting in the future to compare the similarities and differences across Apicomplexa as to the function of cAMP signalling. Our data suggest that PKAc1, upon invasion, negatively regulates cytosolic [Ca2+], allowing for the rapid suppression of the Ca2+ signal to promote a stable intracellular infection. This suggests that PKAc1-dependent signalling is important for sensing the host cell environment and suppressing [Ca2+]cyt, to turn off motility. [K+] is considered to be at a relatively high concentration inside host cells and lower in the extracellular environment, and previous reports have shown that [K+] acts as a major regulator of motility by somehow supressing [Ca2+]cyt. Using GCaMP6 as a [Ca2+]cyt biosensor, we were also able to show that loss of PKAc1 in buffers mimicking the intracellular [K+] also resulted in aberrant [Ca2+]cyt. Together with our live cell imaging on GCaMP6-expressing parasites, our data suggest an important role of PKAc1 in negatively regulating [Ca2+]cyt upon completion of invasion and sensing of the host cell environment. Secretion of PLP-1 from the micronemes and its activation by PV acidification have both been shown to be important for host cell egress [23,42]. We have demonstrated that early egress by PKAc1-deficient tachyzoites is dependent on PLP-1 and PV acidification, but at this stage we have not been able to determine if PKAc1 directly regulates PLP-1 secretion and activity, or if they operate in different pathways that are both required for egress to take place. Given that PKAc1 negatively regulates [Ca2+]cyt in a buffer that mimics the intracellular environment and that raised [Ca2+]cyt levels are known to induce release of PLP-1 from the micronemes, it seems plausible that this is the reasonable mechanism that allows control of egress by cAMP signalling. It was recently described that PKAc1 in Toxoplasma is also responsible for negatively regulating host cell egress [48]. Here, Jia and colleagues have used a conditional dominant negative PKAr1, which cannot respond to cAMP, to show that PKAc1 is important for negatively regulating early egress from host cells. Unfortunately, we were unable to robustly look at the role of PKAc1 during egress to confirm this work, because of the time it takes to deplete PKAc1 levels and the stochastic nature of host cell egress in standard Toxoplasma culturing conditions. However, our work is wholly consistent and complementary with their findings and furthermore also describes the role of PKAc1 in negatively regulating basal [Ca2+]cyt levels, which appears especially important when encountering a host cell environment. Indeed, our findings provide a mechanism as to why dominant negative PKAr from Jia and colleagues can induce early egress and provide evidence as to why there is a preponderance of Ca2+-regulated proteins found to be more phosphorylated in the absence of PKAc1 activity [48]. Jia and colleagues also suggest that PKA and cGMP signalling are connected by identifying PKAc1-dependent phosphorylation sites on a putative PDE. We further this finding by showing that loss of PKAc1 sensitises parasites to cGMP-induced [Ca2+]cyt, thus providing a functional link between these three signalling modalities. This work and the work of Jia and colleagues [48] suggest that cAMP signalling could be a mechanism by which Toxoplasma (and potentially other apicomplexan parasites) transduce extracellular signals. If this model is correct, then we would expect that cAMP levels would have a reciprocal relationship with [Ca2+]cyt, being high in intracellular parasites and low in extracellular parasites. Indeed, using available genetically encoded cAMP biosensors, this question is now eminently addressable and will be important to understand the role of cAMP in regulating motility. Interestingly, a recent study has demonstrated that another PKA orthologue, PKAc3, negatively regulates bradyzoite differentiation in Toxoplasma. Together with the results presented here and by Jia and colleagues, this is suggestive that a drop in cytosolic cAMP could have one of two outcomes: either egress from a host cell or differentiation into latent forms. Clearly, understanding more detail on the function of cAMP will be important if we wish to determine how Toxoplasma and other apicomplexan parasites appropriately respond to environmental cues to either exit the host cell or initiate differentiation.
Detailed methodologies and oligonucleotides used for the construction of plasmids are available in the Supporting information (S1 Text and S2 Table, respectively). T. gondii tachyzoites were cultured under standard conditions. Briefly, HFFs (American Tissue culture collection [ATCC]), Vero, and immortalised MEFs expressing membrane-bound tdTomato [41] (a kind gift from C. Allison, WEHI) were grown in DME supplemented with 10% heat-inactivated Cosmic Calf Serum (Hyclone). When infecting HFFs with tachyzoites, media was exchanged to DME with 1% FCS. All cells were grown in humidified incubators at 37 °C/10% CO2. Transfection proceeded using either a Gene Pulser II (BioRad) or a Nucleofector 4D system (Lonza). Gene Pulser II transfection took place using standard procedures, using 50 μg of purified DNA, and the DNA was linearised if seeking homologous integration [56,57]. Nucleofection proceeded by using 2 × 106 tachyzoites and 5 μg of DNA in 20 μL of buffer P3 (Lonza), which was pulsed using program F1-115. Plaque assays were performed by inoculating 100–500 tachyzoites onto confluent monolayers and leaving the cells undisturbed for 7–8 days. Monolayers were then fixed in 80% ethanol and stained with crystal violet (Sigma). IFAs of Toxoplasma-infected host cells were undertaken using standard procedures detailed in S1 Text. Images were captured on an AP DeltaVision Elite microscope (GE Healthcare) equipped with a CoolSnap2 CCD detector and captured with SoftWorx software (GE Healthcare). Images were assembled using ImageJ and Adobe Illustrator software. Ty (mAb BB2), HA antibodies (3F10) (Roche), glydeosome-associated protein 45 (GAP45) [58], and ISP1 [30] were all used at 1: 1,000 in 3% BSA/PBS. Secondary Alexa Fluor conjugated antibodies (Molecular Probes) were all used at 1: 1,000. Images were adjusted in ImageJ and imported into Adobe Illustrator CC 2017. A detailed protocol for assaying invasion is available in S1 Text. The invasion assay proceeded largely as previously described [59]. Briefly, parasite lines were treated ±ATc and resuspended in IC buffer, to prevent invasion, and spun down onto the HFF monolayer in an eight-well chamber slide (Ibidi). IC buffer was exchanged for DME/1% FCS/10 mM HEPES, pH 7. 5, and incubated for 10 minutes at 37 °C. Samples were then fixed, stained with anti-SAG1 mAb DG52 (1: 3,000 dilution), permeabilised, and stained with rabbit anti-GAP45 (1: 2,000) [58], followed by secondary antibodies, then imaging on a Zeiss Live Cell Observer and quantitation by manual counting. For invasion assays using GFP-expressing parasites, GAP45 staining was not carried out. Invasion assays with DCCD were carried out as described above except that parasites were allowed to invade in D1/HEPES medium containing 50 μM DCCD (Sigma). Graphs were generated and statistical tests performed using Prism 7 software. A detailed protocol for attachment assays is described in S1 Text. Tachyzoites were resuspended in DME/1% FCS/10 mM HEPES, pH 7. 5, at a concentration of 1×107 tachyzoites/mL. A total of 200 μL of tachyzoites was added to each well of an eight-well chamber slide (Ibidi) that contained HFFs fixed with 2. 5% formaldehyde/0. 02% glutaraldehyde. Parasites were allowed to attach for 30 minutes at 37 °C before fixing again and performing IFAs as described in S1 Text. Graphs were generated and statistical tests performed using Prism 7 software. Egressed parasites from three T150 flasks for each of Δku80: TATi (parental) and Δku80: TATi: PKAc1 cKD parasites were resuspended in 1 mL lysis buffer (100 mM HEPES, pH 7. 5; 1% NP-40; 300 mM NaCl; 1 mM MgCl2; 25 U/mL Benzonase [Novagen], and protease inhibitors without EDTA [Roche]) and incubated on ice for 30 minutes, before centrifuging for 15 minutes at 14,000 rpm and 4 °C to pellet cellular debris. The lysates were added to 100 μL of anti-HA agarose beads (50% slurry [Sigma]) that had been washed twice with 1 mL PBS/0. 5% NP-40) and incubated for 2. 5 hours at 4 °C, with rotation. The beads were then pelleted by centrifugation at 10,000g for 15 seconds and the supernatants removed. The beads were washed five times with ice-cold 1 mL PBS/0. 5% NP-40. Analysis was performed by running eluates on a NuPAGE 4%–12% nonreducing Bis-Tris Gel (Invitrogen) and stained with SyproRuby according to the manufacturer’s instructions (Invitrogen), followed by Page Blue staining (Fermentas). The two prominent bands from the Δku80: TATi: PKAc1 cKD IP were cut out and identified by mass spectrometry. Whole eluates were also subjected to mass spectrometry without prior protein gel separation and quantitated using a custom label-free quantification pipeline. For more details on procedures and stable isotope labelling by amino acids in cell culture (SILAC) labelling used for quantifying protein release from micronemes, please refer to S1 Text. Eight-well cell culture–treated chamber slides (Ibidi) were seeded with either HFFs or ROSA MEFs and grown as outlined above. Toxoplasma tachyzoites were treated overnight with 1 μg/mL of ATc and harvested as stated above. Tachyzoites were then resuspended in Endo buffer and spun down onto the host cells, then taken to the imaging station. Tachyzoites and host cells were then imaged using a Leica SP8 confocal, equipped with resonant scanner. For quantitative fluorescent measurements, GCaMP6 was excited with a 488-nm krypton/argon laser line, whilst mCherry and tdTomato were excited using the 594-nm and 561-nm laser lines, respectively. All imaging took place with pinhole at 1 Airy unit, typically over five z-stacks covering 5 μm. Image analysis took place using ImageJ, using inbuilt and custom macros (available on request). Data were processed in Excel and graphed in Prism. PKAc1 cKD parasites grown in confluent HFFs for 24 hours were treated with 1 μg/mL ATc (or the equivalent volume of 100% EtOH; solvent control) for 6 hours. Following this, parasites were harvested by passage of cultures through a 26-gauge needle and centrifuged (1,500g for 10 minutes at 4 °C). The parasites were then resuspended in fresh culture medium containing either ATc (final concentration, 1 μg/mL) or the equivalent volume of 100% ethanol and added to flasks containing confluent HFFs for up to 24 hours. The parasites were then harvested from their host cells and loaded with Fura-2 as described previously [60]. Cytosolic Ca2+ measurements were performed at 37 °C using a PerkinElmer LS 50B Fluorescence Spectrometer as described previously [60]. Graphs were generated and statistical tests performed using Prism 7 software. Parental and PKAc1cKD parasites expressing GCaMP6/mCherry were treated ±ATc for 14 hours and resuspended in either EC buffer (141. 8 mM NaCl, 5. 8 mM KCl, 1 mM CaCl2,1 mM MgCl2,5. 6 mM Glucose, 25 mM HEPES, pH 7. 4) or IC buffer (5 mM NaCl, 142 mM KCl, 2 mM EGTA, 1 mM MgCl2,5. 6 mM Glucose, 25 mM HEPES, pH 7. 4). Tachyzoite samples were then read on a FACS LSRII. Samples were spiked with 2× concentrated stocks of equivolume of either BIPPO or A23187 at 20 seconds post-acquisition and read for a total of 100 seconds. Values were imported into Prism 7 software for plotting and statistical testing. Secretion assays were performed and quantitated as described previously [61]. Briefly, cells were resuspended in Ringer’s buffer, shifted to 37 °C and allowed to secrete for 60 minutes. Secretion was arrested by placing cells on ice for 2 minutes. Parasites were separated from the soluble secreted proteins by centrifugation at 8,000 rpm, 4 °C, for 2 minutes. A total of 85 μL of supernatant was removed and centrifuged again to remove any remaining cells, and 75 μL of supernatant was removed and boiled with reducing Sample Buffer. The parasites pellet was washed with PBS and boiled in reducing Sample Buffer. Secreted proteins were analysed using antibodies to indicated factors by western blot: TOM40 (1: 1,000) [62], GRA1 (1: 2,000) [63], microneme protein 2 (MIC2) (1: 5,000) [64], AMA1 CL22 (1: 1,000) [65], SUB1 (1: 1,000) [66], and PLP1 [43]. | Central to pathogenesis and infectivity of Toxoplasma and related parasites is their ability to move through tissue, invade host cells, and establish a replicative niche. Ca2+-dependent signalling pathways are important for activating motility, host cell invasion, and egress, yet how this signalling is turned off after invasion is unclear. Here, we show that a cAMP-dependent protein kinase A (PKA) is essential for rapid suppression of Ca2+ signalling upon completion of host cell invasion. Parasites lacking this kinase rapidly invoke an egress program to re-exit host cells, thus preventing the establishment of a stable infection. This finding therefore highlights the first factor required for Toxoplasma (and any related apicomplexan parasite) to switch from invasive to the replicative forms. | Abstract
Introduction
Results
Discussion
Materials and methods | cell motility
parasite groups
medicine and health sciences
viral transmission and infection
enzymes
microbiology
enzymology
parasitic diseases
parasitic protozoans
in vivo imaging
parasitology
apicomplexa
tachyzoites
protozoans
toxoplasma
research and analysis methods
imaging techniques
proteins
protein kinases
biochemistry
eukaryota
host cells
cell biology
virology
biology and life sciences
organisms | 2018 | Protein kinase A negatively regulates Ca2+ signalling in Toxoplasma gondii | 14,746 | 194 |
Innate and type 1 cell-mediated cytotoxic immunity function as important extracellular control mechanisms that maintain cellular homeostasis. Interleukin-12 (IL12) is an important cytokine that links innate immunity with type 1 cell-mediated cytotoxic immunity. We recently observed in vitro that tumor-derived Wnt-inducible signaling protein-1 (WISP1) exerts paracrine action to suppress IL12 signaling. The objective of this retrospective study was three fold: 1) to determine whether a gene signature associated with type 1 cell-mediated cytotoxic immunity was correlated with overall survival, 2) to determine whether WISP1 expression is increased in invasive breast cancer, and 3) to determine whether a gene signature consistent with inhibition of IL12 signaling correlates with WISP1 expression. Clinical information and mRNA expression for genes associated with anti-tumor immunity were obtained from the invasive breast cancer arm of the Cancer Genome Atlas study. Patient cohorts were identified using hierarchical clustering. The immune signatures associated with the patient cohorts were interpreted using model-based inference of immune polarization. Reverse phase protein array, tissue microarray, and quantitative flow cytometry in breast cancer cell lines were used to validate observed differences in gene expression. We found that type 1 cell-mediated cytotoxic immunity was correlated with increased survival in patients with invasive breast cancer, especially in patients with invasive triple negative breast cancer. Oncogenic transformation in invasive breast cancer was associated with an increase in WISP1. The gene expression signature in invasive breast cancer was consistent with WISP1 as a paracrine inhibitor of type 1 cell-mediated immunity through inhibiting IL12 signaling and promoting type 2 immunity. Moreover, model-based inference helped identify appropriate immune signatures that can be used as design constraints in genetically engineering better pre-clinical models of breast cancer.
The discovery of molecular targeted therapies revolutionized the treatment of breast cancer. Tamoxifen, a small molecule inhibitor of the estrogen receptor, was the first drug to inhibit the growth of breast cancer cells that depend on female sex hormones. More recently, trastuzumab was developed to inhibit the growth of breast cancer cells that overexpress HER2, an oncogenic member of the epidermal growth factor family of receptors [1]. Based upon their demonstrated clinical impact, a pre-operative biopsy sample is used to guide treatment based upon expression of the hormone receptors for estrogen (ER) and progesterone (PR) and the epidermal growth factor receptor HER2 [2]. While these molecular targeted therapies have improved survival, de novo and acquired resistance to these therapies present challenges for achieving a durable clinical response [3], [4]. The difficulties in achieving a durable clinical response using molecular targeted therapies have sparked a renewed interest in viewing cancer from an evolutionary perspective [5]–[7]. Thinking about cancer from an evolutionary perspective involves three key concepts. First, tumors are comprised of a heterogenous population of cells. While non-genetic sources of heterogeneity have been recognized for several decades [8], recent studies of breast cancer using next generation sequencing have revealed the genetic heterogeneity associated with oncogenesis [9]–[14]. Second, the different cell types contained within the tumor microenvironment - stromal cells, malignant clones, and cells of the immune system - and their collective interactions comprise a dynamic system. Dynamic systems typically have control mechanism that aim to maintain the system in a desirable state, such as tissue homeostasis, in the presence of external perturbations [15]. Identifying the control architecture in biological systems remains a central challenge. Third, cells of the population impinge upon a selective fitness landscape that determines their fate. The selective landscape is composed of intracellular and extracellular adaptive control mechanisms that regulate tissue homeostasis. Many intracellular control mechanisms are well studied and form the foundation of the hallmarks of cancer [16]. Innate and adaptive immunity function as extracellular control mechanisms that regulate cellular homeostasis [17] and, in contrast, are not well understood [5]. Cytokines coordinate innate and adaptive immunity and defects in their action have pathogenic implications. For instance, Interleukin-12 (IL12) is a cytokine that is produced by innate immune cells and acts upon Natural Killer cells, CD8+ Cytotoxic T cells, and CD4+ T helper cells to initiate a type 1 cell-mediated adaptive immune response [18]. Genetic mutations in IL12p40 and one component of the IL12 receptor, IL12Rβ1, have been observed in patients with recurrent mycobacterial disease, suggestive of insufficient type 1 cell-mediated immunity [19], [20]. Genetic deletion of other component of the IL12 receptor, IL12Rβ2, increases susceptibility to spontaneous autoimmunity, B-cell malignancies, and lung carcinomas [21]. Originally called Natural Killer (NK) Cell Stimulating Factor, IL12 also enhances the ability of NK and CD8+ Cytotoxic T cells to lyse target cells, a mechanism exploited for tumor immunotherapy. For example, IL12 was used as an adjuvant to promote NK-cell mediated killing of HER2-positive breast cancer cells in patients treated with trastuzumab [22]. As an adjuvant for tumor immunotherapy, toxicity restricts the systemic delivery of IL12 [23]. However, local delivery of IL12 to the tumor microenvironment promotes tumor regression in the B16 melanoma model [24] and in the EL4 thymoma model [25]. Given that genetic defects in IL12 signaling increase cancer incidence and enhanced local delivery of IL12 promotes tumor regression, we recently asked whether malignant cells alter the selective fitness landscape by locally inhibiting the response of immune cells to IL12. Using the B16 model for melanoma, we identified that tumor-derived Wnt-inducible signaling protein 1 (WISP1), a beta-catenin responsive oncogene [26], exerts paracrine action on immune cells by inhibiting their functional response to IL12 [27]. The objective of this retrospective study of invasive breast cancer was three-fold: 1) to determine whether the gene signature associated with type 1 cell-mediated cytotoxic immunity was correlated with overall survival, 2) to determine whether WISP1 expression is increased in invasive breast cancer, and 3) to determine whether a pattern of gene expression consistent with inhibition of IL12 signaling axis correlates with WISP1 expression.
In this retrospective cohort study, our first objective was to determine whether there were distinct cohorts within the invasive breast cancer arm of the TCGA study that can be defined based upon type 1 cell-mediated immune response. Samples included in the analysis were limited to those derived from patients diagnosed with variants of invasive breast cancer (n = 520) and to those obtained from normal breast tissue (n = 61) (see Dataset S1). As listed in Table 1, normalized expression values were obtained for genes that are associated with T cells, macrophages, and Natural Killer cells and the functional roles that these cells play in cell-mediated immunity. As macrophages and T cells can enhance or inhibit cell-mediated immunity depending on polarization bias, genes associated with alternative polarization states were also included. Specifically, macrophages within the tumor microenvironment are thought to either promote (M1) or inhibit (M2) cytotoxic cell-mediated immunity [28], [29]. Similarly, effector T cells alter their functional role in coordinating adaptive immunity depending on the cytokines secreted by and the transcription factors expressed by different subsets, which include type 1, type 2, type 17, or T regulatory subsets [30], [31]. We also included genes associated with the IL12 cytokine family, as IL12 links innate to adaptive immunity and other members of this cytokine family have competing effects on immune bias (e. g. , [32]). Based upon immune-related gene expression, tissue samples hierarchically clustered into two main groups (Figure 1 and Figure S1). While more patients were associated with Group 1 (n = 370 versus n = 150), the two groups had similar patient population characteristics with no difference between age, tumor stage, menopause status, or lymph node status (Table 2). Based on either molecular pathology or PAM50 intrinsic subtypes [33], the molecular characteristics of tumors associated with these two groups were significantly different (p–value<1x10-4 - Table 2). Triple negative (TN) breast cancer samples were significantly enriched in group 2 with an odds ratio of 3. 48 versus 0. 38 for group 1. As expected, samples positive for either estrogen receptor (ER) or progesterone receptor (PR) and negative for HER2 were enriched in group 1. Given that molecular subtyping guides therapy selection, 6-year overall survival was estimated using Kaplan-Meiers curves for group 1 versus group 2 cohorts stratified by adjuvant treatment, if known, and by molecular pathology (Figure 2). Patients with TN breast cancer that clustered with group 2 exhibited an increase in overall 6-year survival compared with patients with TN breast cancer that clustered with group 1 (p–value<0. 03 with hazard ratio = 0. 191 (95% C. I. 0. 037–0. 995) ). While the number of events is low for the TN breast cancer group, the overall trend in survival was also observed at earlier time points (1-year survival p–value<0. 032 with hazard ratio<0. 01; 3-year survival p–value = 0. 126 with hazard ratio = 0. 28; and 5-year survival p–value = 0. 054 with hazard ratio = 0. 22). Patients with HER2+ breast cancer also exhibited a similar trend, but the difference in overall survival did not reach a similar level of significance. In contrast, no difference in 6-year survival was observed between the group 1 and group 2 cohorts treated using adjuvant hormone therapy, if known, or were positive for either ER or PR. Effective anti-tumor immunity depends on the product of the number of immune cells present within the tumor microenvironment and the efficacy of the immune cells present to elicit cell-mediated cytotoxic immunity. To infer mechanistic differences in anti-tumor immunity between the patient cohorts, we estimated the magnitude and the quality of anti-tumor immunity from the gene expression measurements. The relative magnitude of immune cell infiltration was inferred from the average expression values for genes associated with NK cells, T cells and macrophages (Table 1). Compared to samples derived from normal tissues, the group 1 cohort exhibited a gene expression signature associated with reduced NK cell, T cell, and macrophage recruitment (Figure 3A). In contrast to group 1, the immune cell signature in the group 2 cohort suggested an increase in NK cell, T cell, and macrophage recruitment relative to normal breast tissue. As immune cell polarization influences the efficacy of anti-tumor immunity, we used model-based inference to determine the polarization signature. In normal breast tissue, T helper cells were primarily polarized towards Th2 (p–value<0. 001) and macrophages were polarized towards a M2 phenotype (Figure 3B, p–value<0. 001). In invasive breast cancer, group 1 cohort exhibited a mixed Th2 and Th17 immune bias and a strong Th1 bias was associated with group 2 samples (p–value<0. 001). Consistent with the Th1 bias in group 2 samples, the genes associated with a type 1 cell-mediated immune response were also consistently expressed at higher levels in the group 2 cohort compared to group 1 and normal tissue samples. Compared to the null hypothesis for immune cell polarization, the macrophage polarization bias in the group 1 cohort could not be distinguished from a pattern of random gene expression and samples from the group 2 cohort exhibited a strong M1 bias (p–value<0. 001). Collectively, the genes associated with type 1 cell-mediated immunity can be used to identify cohorts within invasive breast cancer that correlate with improved 6-year survival, specifically in patients with TN invasive breast cancer. Principal component analysis (PCA) was used to gain insight into the molecular differences between cohorts identified using hierarchical clustering. As the samples from normal tissues clustered predominantly with group 1, the group 1 cohort was subdivided into three subsets (Figure S2A) with normal tissue samples clustering into group 1a. Of the total variation contained within the gene expression data, the first four principal components (PCs) captured 54% of the variance (Figure S2B). Differences among the clustered cohorts were observed for the first four PCs while no significant differences among groups were observed for the rest of the PCs (Figure S2C). Based upon the loading coefficients for individual genes, the magnitude of PC1 corresponded to the extent of T cell-mediated (increases with CD2, CD247, CD3G, CD3D, CD8A, and CD28 expression) type 1 cytotoxic immunity (increases with FASLG, IFNG, GZMB, TBX21, IL12RB2, EOMES, PRF1, B2M and decreases with GATA3 expression) and PC2 captured a correlation between WISP1 and the T cell lineage-defining transcription factors GATA3 and PPARG. As described in the supplemental Text S1, a similar gene expression signature was observed in the gene expression results reported by Gluck et al. (GSE22358 [34]). PC projections of the patient samples suggest that the extent of T cell-mediated cytotoxic immunity within invasive breast cancer is a continuous property, with TN breast cancer more prevalent at higher values for PC1 and HER2+ breast cancer more prevalent at lower values for PC1 (Figure S2C). Yet, the hierarchical clustering subdivided this continuous property into two discrete cohorts. In contrast to PC1, PC2 separated samples derived from invasive breast cancer from normal breast tissue and, based upon the loading coefficients for PC2, suggests that an increase in WISP1 expression correlates with oncogenic transformation (Figure 3C - p–value<1x10-15). The average intensity of WISP1 antibody staining in an independent tissue microarray that contained samples from normal (n = 3) and breast carcinoma tissue (n = 9) were used to validate that an increase in WISP1 correlates with oncogenic transformation (Figure 4, panels A–C). The tissue microarrays were consistent with the gene expression data such that WISP1 was increased in invasive breast cancer compared to normal breast tissue (p–value<0. 001). T cell polarization is driven by competition among lineage-defining transcription factors that are induced by the exogenous action of polarizing cytokines [30]. By inducing T-bet, Interleukin-12 (IL12) acts, in part, upon CD4+ T helper cells to organize an effective type 1 cell-mediated immune response [18]. As a potential inhibitor of a type 1 cell-mediated immune response, in vitro co-culture assays identified WISP1 as a paracrine regulator of immune cells by inhibiting response to IL12 [27]. Here, we found that the loading coefficients associated with PC2 suggest that the variation in WISP1 was also associated with two T cell lineage-defining transcription factors: GATA3 and PPARG. Specifically, WISP1 expression was correlated with GATA3 expression (Figure 3C - p–value<1x10-9) and exhibited a negative correlation with PPARG expression (see Figure S4A - p–value<1x10-15). The functional connection between WISP1 and immune polarization is strengthened by the observations that GATA3 expression also correlates with GATA3 protein abundance (Figure 4D - p–value<1x10-15) and that GATA3 expression does not correlate with changes in genome copy numbers of GATA3 (Figure S4B - p–value=1). In addition, GATA3 expression exhibited a negative correlation with IL12 receptor β 2 (Figure S4C - p–value<1x10-15). Compared to other putative immunosuppressive mechanisms present in the tumor microenvironment, the WISP1-GATA3 signature was the only mechanism that was higher in the group 1 cohort relative to the group 2 cohort (see Figure S5). Peroxisome proliferator-activated receptor (PPARG) is a ligand-activated transcription factor that plays an important role in regulating immunity and oncogenesis [35]. For instance, Th2 polarization is associated, in part, with increases in GATA3, IL6, and PPARG gene expression [31]. In contrast, Chung et al. report that PPARG can form an inhibitory complex with nuclear factor of activated T cells (NF-AT) that inhibits the transcription of IL4 in T helper cells [36]. In a mouse model of atopic asthma, pharmacologic activation of PPARG reduces the canonical Th2 cytokines IL4 and IL13 and GATA3 protein in lung extracts [37]. Collectively, these results suggest that the gene set - CD4, IL4, IL5, IL10, and GATA3 - is a better signature for a type 2 bias of T helper cells as an increase in PPARG expression may sensitize Th2 effector cells to negative regulation by PPARG ligands and that GATA3 and PPARG may be regulated independently. Using this revised Th2 cell gene signature, the group 1 cohort is associated with an increase in type 2 bias relative to the group 2 cohort and normal breast tissue (Figure S6 - p–value<0. 001). The group 2 cohort exhibited a strong type 1 bias while a Th17 bias was observed in samples from normal breast tissue. An expression signature consistent with inducible T regulatory cells was not observed in any of the cohorts (p–value<0. 001). Using a Cox proportional hazards regression model, we also found that polarization towards a T helper type 1 phenotype was a predictor of survival independent of the molecular pathology (see). Using bootstrap resampling of the genes listed in Table 1, the model-based inference of type 1 polarization is a better predictor of improved survival than 95% (i. e. , p–value<0. 05) of the random immune signatures for the 1 Yr, 3 Yr, and 6 Yr outcomes and than 93% of the random signatures for 5 Yr outcome (see Figure S7). In addition to secreting WISP1, B16 model for melanoma also overexpresses one component of the IL12 receptor, IL12Rβ2, that, in vitro, creates a local cytokine sink for IL12 [27]. We have also found that STAT4 is phosphorylated irreversibly, creating a short term memory to IL12 signaling that is limited by cell proliferation [38]. Local delivery of IL12 to the tumor microenvironment promotes tumor regression in the B16 melanoma model [24] and in the EL4 thymoma model [25]. Collectively, these studies imply that signaling by endogenous IL12 within the tumor microenvironment helps to maintain T cell polarization when cognate tumor antigens induce T cell proliferation [39] and that manipulating this extracellular control mechanism may impart a survival advantage to the collective tumor population. Finally, we wanted to determine whether IL12 receptor β2 was increased relative to IL12 receptor β1 in samples derived from tumors with active type 1 cell-mediated immune response. In invasive breast cancer, the ratio of IL12RB2 to IL12RB1 expression was increased in the group 2 cohort relative to the group 1 cohort (Figure 3D - p–value<1x10-15). As we had previously observed an imbalance in copy numbers of the components of the IL12 receptor in malignant melanocytes, we measured IL12 receptor β1 and IL12 receptor β2 copy numbers in the 184A1, BT474, SKBR3, and MDA-MB-231 cell lines by flow cytometry (Figure 5, panels A–B). While the copy numbers in these cell lines were lower than what we had observed in the B16F0 and B16F10 cell lines, the ratio of IL12 receptor β2 to IL12 receptor β1 was increased in the triple-negative breast cancer model, MDA-MB-231, relative to the other three cell lines (see Figure 5C - p–value<0. 05). We also observed that abundance of major histocompatibility complex (MHC) class I molecules varied among these cell lines in a pattern consistent with differences in type 1 cell-mediated immunity (Figure 5D). Specificity in directing T cell-mediated cytotoxic immunity depends on the presentation of peptides bound to MHC class I molecules. A reduction in MHC class I expression may reduce the efficacy of type 1 cell-mediated immunity in controlling tumor growth. Given that the gene expression signature associated with type 1 cell-mediated immunity varied among the invasive breast cancer samples, we ascertained whether there were any basal differences in copy numbers of MHC class I among cell lines derived from the different cohorts. The BT474 and SKBR3 cell lines are cell models of HER2+ breast cancer, a subset that exhibited a gene expression signature associated with a low type 1 cell-mediated response. The 184A1 cell line is a cell model of normal breast tissue, which had an intermediate type 1 cell-mediated response signature. The MDA-MB-231 cell line is a cell model of triple negative breast cancer, which exhibited a high type 1 cell-mediated response signature. Copy numbers of MHC class I molecules were assayed by flow cytometry (Figure 5D). Collectively, the copy numbers of MHC class I molecules varied among the cell lines: 184A1 expressed 307K copies while BT474 and SKBR3 expressed lower copies (16K and 10K, respectively) and MDA-MB-231 expressed higher copies (916K). The lower copies in the BT474 and SKBR3 cell lines is consistent with previous studies that report that over-expression of HER2/Neu reduces MHC class I expression [40], [41]. In summary, basal differences in MHC class I and IL12 receptor copy numbers among the 184A1, BT474, SKBR3, and MDA-MB-231 cells lines are consistent with a model of invasive breast cancer where the molecular subtypes are distinguished by differences in type 1 cell-mediated immunity. Modeling breast cancer in mice inevitably involves some degree of abstraction - one must determine key elements associated with the human disease and select model systems that incorporate those elements. Historically, transplantable models for cancer, like the B16 melanoma and 4T1 breast cancer models, have been used to study anti-tumor immunity in vivo. Cell lines that were derived from spontaneous tumors can be manipulated in vitro to express defined tumor antigens and re-introduced into a syngeneic host. However, transplantable models have been criticized as they do not resemble established spontaneous tumors (e. g. , [42], [43]). One of the advances associated with pre-clinical drug discovery and development has been the development of genetically engineered mouse models (GEMM) for cancer that incorporate alterations in known oncogenes and tumor suppressors. Breast cancer GEMMs spontaneously develop lesions in mammary tissue that histologically resemble the human equivalent [44]. Yet GEMMs are not without criticism, as Jacks and coworkers suggest that genetically engineered mouse models of cancer may underestimate the mutational and antigenic load of most human cancers [45]. Given the observed immune gene expression signature observed in human breast cancer tissue, we also wanted to determine whether most common breast cancer GEMMs exhibit similar immune gene signatures as the human disease. Similar to the TCGA analysis presented in Figure 1, mRNA expression results from 122 breast cancer tumor and normal breast tissue samples obtained from a variety of genetically engineered mouse models (GEMM) for breast cancer (see Figure 6 - [GEO: GSE3165]) [46]. Within this GEMM data set, four gene expression clusters were identified based upon a subset of genes associated with anti-tumor immunity and immunosuppressive mechanisms: a normal group and three breast cancer groups. In contrast to the TCGA data, WISP1 was up-regulated in only a small subset of 7,12-dimethylbenz[a]anthrazene (DMBA) -induced breast cancer models. Moreover, the immune gene expression signatures in the breast cancer GEMMs suggested that NK cell infiltrate is unchanged, T cell infiltrate is suppressed, and that macrophages are elevated relative to normal breast tissue (see Figure 7A). This signature is different from the immune cell gene signatures observed in the TCGA data set that suggest that NK cells, T cells, and macrophages were either collective decreased in group 1 or collectively increased in group 2 relative to normal tissues. In terms of immune polarization, macrophages exhibit a similar pattern in the GEMMs compared to the human samples, where macrophages in normal tissue are skewed towards an M2 and macrophages in tumor tissue are skewed towards M1 phenotypes (see Figure 7B). In contrast, the inferred T cell polarization states in GEMMs are different from human samples. A regulatory T cell signature is highest in normal mammary tissue and the different tumor models exhibit mixed immune polarization signatures that overlap with the null hypothesis signature. The GEMM results also suggest that reduced T cell recruitment to mammary tumors in the GEMMs reduces the signal associated with T cell polarization, which in turn makes identifying the T cell polarization state from homogenized tissue samples difficult. Similar discordance has been reported between human inflammatory diseases and their corresponding mouse models [47]. Collectively, the immune signatures in the set of GEMMs are unlike that observed in the TCGA samples and motivate developing pre-clinical mouse models that better reflect the immune signature associated with invasive breast cancer in humans. Given the discordance between GEMMs and humans, the likelihood for a type 2 error in testing immunotherapies using spontaneous mouse models for breast cancer seems high. Unfortunately, developing GEMMs that better mimic human anti-tumor immunity is a recommendation that has persisted for over three decades [48]. In contrast to the 1980s, this work illustrates how data obtained from large-scale studies, like the Cancer Genome Atlas, coupled with in silico model-based inference methods can be used to identify appropriate immune signatures. These immune signatures can be used as design constraints in genetically engineering better pre-clinical mouse models of cancer.
In this retrospective study of the invasive breast cancer arm of the Cancer Genome Atlas study, we made three main observations. First, type 1 cell-mediated cytotoxic immunity was correlated with increased survival in patients with invasive TN breast cancer and was predictor of survival independent of molecular pathology. In the absence of a molecular targeted therapy, an increase NK and T cell infiltrate and a shift from type 2 towards type 1 cell-mediated immunity correlated with a survival advantage in patients with invasive TN breast cancer. These correlates were also identified independent of tumor staging and lymph node involvement. This observation is consistent with a significant body of literature that correlate various molecular characteristics of anti-tumor immunity with overall survival in breast cancer, as reviewed by Andre et al. [49]. While many studies focus on particular immune-related molecules, gene expression profiling provides a less biased view of the immunologic properties of anti-tumor immunity. As one of the most extensively studied cancers, gene expression profiling of primary tumors (e. g. , [50]–[54]) and tumor stroma [55] have identified immune classifiers of clinical outcome in breast cancer. In contrast to a focus on primary tumors, samples from age-matched normal breast tissue were included in the analysis to gain insight into the immunologic changes associated with oncogenic transformation, which led to the second observation. Second, oncogenic transformation in invasive breast cancer was associated with an increase in WISP1 gene expression (p–value<1x10-15) and protein abundance (p–value<0. 001). Prior studies of WISP1 in breast cancer have been mixed. Xie and coworkers observed higher WISP1 gene expression in 20 of 44 breast cancer samples while the remaining samples exhibited levels of expression similar to normal breast tissue [56]. In contrast, Davies and coworkers found that WISP1 was reduced in samples obtained from breast tumors compared to normal breast tissue [57]. Here, the focus on invasive breast cancer and larger sample size may explain some of the observed differences. WISP1 is a member of the family of connective tissue growth factors that is induced by nuclear localization of beta-catenin [26] and participates in stem cell differentiation and tumorigenesis [58]. While the details remain to be fully elucidated, on-going work suggests that proteolytic cleavage of E-cadherin enables membrane-bound beta-catenin to localize to the nucleus and induce WISP1 expression [27]. In vivo, loss of E-cadherin plays an important role in the metastatic potential of cancers [59]. While WISP1 has been reported to influence neurodegeneration and osteogenesis [60], a signaling mechanism has yet to be identified despite a report suggesting that WISP1 binds the integrin [61]. Third, the gene expression signature in invasive breast cancer was consistent with WISP1 as a paracrine inhibitor of type 1 cell-mediated immunity through inhibiting IL12 signaling and promoting type 2 immunity. In particular, we found a highly significant correlation between WISP1 and GATA3 (p–value<1x10-9) and a highly significant negative correlation between WISP1 and PPARG (p–value<1x10-15). While the increase in GATA3 is consistent with WISP1 as an inhibitor of IL12 signaling, the negative correlation between WISP1 and PPARG was unexpected and suggests that the reduction in PPARG may relieve transcriptional repression of type 2 cytokine production by T cells. Conventionally, exploratory data analysis is used to identify genes or clusters of genes that correlate with clinical outcome (e. g. , [50]–[55]). To identify gene set classifiers, the gene expression data set is subdivided into a training set - to discover significant gene clusters - and a validation set - to confirm that the gene cluster correlates with outcome. Here a more targeted approach was used. Previously, we used a phenotypic assay that incorporated in vitro, in silico, and unbiased proteomics methods to discover a paracrine mechanism by which a mouse model for melanoma regulates immune response to IL12 [27]. To validate this mechanism, a retrospective analysis was used to identify whether a gene expression signature that is consistent with WISP1 as a paracrine regulator of anti-tumor immunity exists in invasive human breast cancer. This gene signature was embodied in PC2 and corresponded to oncogenic transformation. Inverse relationships between type 1 cell-mediated cytotoxic immunity and GATA3 and between IL12RB2 and GATA3 were also captured in PC1 and PC3, respectively. Collectively, the first three PCs captured 49% of the overall variance in gene expression. The limitations of the analysis are such that the TCGA data reflect homogenized tumor tissue and the genes that are associated with the immune polarization signatures have pleiotropic biological roles. For instance, GATA3 plays a role in both mammary epithelial [62] and immune cell biology [63]. In terms of mammary epithelial cell biology, GATA3 is thought to inhibit breast cancer metastasis. GATA3 is up-regulated in luminal epithelial cells and down-regulated in basal epithelial cells, which have a higher propensity for invasion and metastasis. In this TCGA study, mutations in GATA3 have a higher prevalence [14]. However, less than 2% of samples that exhibited basal-like and HER2-enriched intrinsic subtype characteristics had mutations in GATA3, specifically truncation mutations. TN breast cancer samples were predominantly the basal-like subtype while HER2+ samples corresponded to the HER2-enriched subtype. In terms of immune cell biology, GATA3 promotes type 2 and counter regulates type 1 T cell polarization, as captured in the analysis by the negative loading coefficient for GATA3 in PC1 and the reciprocal relationship between GATA3 and IL12RB2 in PC3. Collectively, the data suggest that changes in GATA3 may reflect a signature derived from immune cell biology rather than mammary epithelial cell biology. We found that GATA3 expression was correlated with WISP1 and that GATA3 was up-regulated in invasive breast cancer compared to normal tissue. The increase over normal in GATA3 was especially prevalent in patient samples associated with the group 1 cohort, a cohort with reduced overall survival for patients with TN breast cancer. In contrast, GATA3 expression was lower relative to normal tissue in a subset of the group 2 cohort, a cohort with improved overall survival for patients with TN breast cancer. Despite the increase in WISP1 associated with oncogenesis, the differences in immune bias between the patient cohorts may reflect intrinsic differences in sensitivity to local reprogramming of tumor-infiltrating lymphocytes. While this cross-sectional tumor biopsy study can not rule out the possibility that the observed signatures are due to systemic alterations in immune response, identifying local factors could help improve the efficacy of many of the immunotherapies currently in clinical trials, such as adoptive T cell transfer or immune checkpoint inhibitors that increase systemic T cell numbers. In summary, effective anti-tumor immunity is proportional to the number and to the cytotoxic activity of immune cells that enter the tumor microenvironment. Recent advances in cancer immunotherapy stem from increasing the number of tumor-infiltrating immune cells by inhibiting immune checkpoints or adoptive T cell therapy. Mirroring the clinical results of these therapies, we found that a gene signature consistent with enhanced type 1 cell-mediated cytotoxic immunity is a predictor of overall survival in invasive breast cancer independent of molecular pathology. In addition, this study also supports a link between epithelial-to-mesenchymal transition - through secretion of WISP1 - and repression of type 1 cell-mediated cytotoxic immunity - through inhibition of IL12 signaling. From an evolutionary perspective, the results also suggest that oncogenic transformation in invasive breast cancer alters the selective fitness landscape through reducing the efficacy of innate and adaptive immunity, which function as important extracellular control mechanisms. Restoring these extracellular control mechanisms will require a better understanding of the dynamics associated with the shift in polarization from type 1 to type 2 within tumor-infiltrating lymphocytes and the sensitivity of this axis, in terms of both quantity and quality, to pharmacological action. From a translational science perspective, these findings motivate a directed effort to demonstrate - using pre-clinical mouse models of invasive breast cancer that more accurately represent the immune signature in human disease - that inhibiting these paracrine immunosuppressive cues will improve the overall response of current cancer immunotherapies. Moreover, model-based inference helped identify the immune signatures that can be used as design constraints in genetically engineering better pre-clinical mouse models of cancer. This knowledge may be of particular importance to patients with TN breast cancer, a patient group that is underserved by the current generation of molecular targeted therapies.
Expression of genes associated with type 1 cell-mediated immunity, as summarized in Table 1, in normal breast tissue and invasive breast cancer were obtained from the breast cancer arm of The Cancer Gene Atlas (TCGA) study [14]. In brief, homogenized samples obtained from primary tumor (n = 520) and matched normal breast tissue (n = 61) were obtained from patients that were newly diagnosed with invasive breast adenocarcinoma following surgical resection and that received no prior treatment for their disease (chemotherapy or radiotherapy). Subsequent treatments, clinical characteristics and biomarkers, and overall outcome for all of the 520 tumor tissue samples and reverse-phase protein array (RPPA) data for only 340 of the tumor tissue samples were downloaded from the TCGA website (https: //tcga-data. nci. nih. gov/tcga/). The median age at diagnosis was 59 years of age and the median follow-up time for overall survival was 22 months. Tumor and normal breast tissue gene expression for this cohort was obtained following array normalization by processing through the Oncomine database (www. oncomine. org). In brief, Level 2 data obtained using Agilent mRNA expression microarrays were downloaded from the TCGA website (https: //tcga-data. nci. nih. gov/tcga/) and the intensity of a given gene probe was normalized to the median of the probe intensities across the entire array sample. Expression of a given gene is expressed in terms of a log2 median-centered ratio, where genes that have a value less than zero are expressed at a level less than the median and genes with a value greater than zero are expressed at levels higher than the median. The abundance of WISP1 in invasive breast cancer and normal breast tissue was quantified by immunohistochemical analysis using a tissue microarray derived from de-identified human breast tissue samples, as provided by the Human Protein Atlas (www. proteinatlas. org, Stockholm, Sweden [64]) and in accordance with approval from the Uppsala University Hospital Ethics Committee. The tissue microarray analysis included samples from 9 breast adenocarcinomas (6 ductal and 3 lobular) and 3 normal breast tissues from women that ranged in age from 23 to 87 years. The tissue microarrays were processed as described by Uhlen et al. [65] and probed using a rabbit polyclonal antibody against WISP1 (ab10737 - Abcam, Cambridge, MA) that was validated by providing partly consistent staining patterns with another antibody and gene/protein characterization data. WISP1 staining was visualized using diaminogenzidine and microscopic tissue features were visualized by counterstaining with Harris hematoxylin. Immunohistochemically stained tissue microarrays were scanned at 20× resolution (1 mm diameter) and provided as an 8-bit RGB JPEG image. Pathological assessments of the images were annotated manually. The average intensity of WISP1 staining per tissue sample was quantified by deconvoluting the intensity of WISP1 staining from nonspecific hematoxylin tissue staining in R using the EBImage package and deconvolution approach described by Ruifrok and Johnston [66]. Polarization of T helper cells and macrophages into different subtypes are defined by differences in gene expression [29], [31]. The genes associated with each subset are summarized in Table 1. The log2 median-centered ratios of subset-defining genes were normalized to the standard deviation of the observed values across the entire cohort, that is a z-score. Immune cell polarization among alternative subsets is assumed to be a mutually exclusive process. Using Bayes theorem, the conditional probability that cells contained within a homogenized tissue sample exhibit a polarization bias, as represented by a model (), given the observed multi-gene expression signature, , can be express as: (1) where is the likelihood of observing data given the polarization model, is the prior for the model, and is the number of polarization subsets. As we have equal ignorance a priori as to how well the competing polarization models describe the data, the priors for each model are set equal to. The likelihood of observing increased expression of a defined multi-gene signature () is the product of the likelihood of observing increased expression of each gene () within the signature. The likelihood can be defined using a simple Euclidian metric based upon the z-score [67], [68]: (2) such that a higher z-score for genes () associated with a subset () and a lower z-score for genes associated with a different subset () corresponds to a higher likelihood. Bootstrap resampling is an effective computational method for estimating the uncertainty associated with a calculated value [69]. Bootstrap resampling with replacement (= 1000) from the set of all observed gene expression values was used to establish a predicted polarization bias for an equivalent size patient cohort (= 200) that is consistent with a null hypothesis, that is the gene expression values are random samples and contain no information regarding immune cell polarization. The distribution in posterior probability of immune bias for a given cohort was obtained using kernel density estimation. Statistical significance associated with the mean posterior probability of immune bias for a given cohort was compared to the null hypothesis. A p–value corresponds to the likelihood that the observed posterior probability (or a more extreme value) of immune bias for a cohort is due to random chance. A p–value<0. 05 was considered significant. The variation and correlation among the gene expression measurements were characterized using principal component analysis (PCA), which is described in more detail in the Supplemental Text S1. The scoring coefficients for each of the top 10 principal components are listed in Dataset S1. Statistical differences between means were assessed using unpaired Student' s t-tests. All Student' s t-tests were two-sided. Statistical significance associated with a correlation between gene expression values within a sample was assessed using a Pearson product-moment correlation coefficient. A one-sided test of the Pearson' s correlation coefficient was used to determine whether a correlation coefficient was positive or negative. A p–value<0. 05 was considered significant. Overall survival time was used as a clinical outcome metric. To estimate cumulative survival probability, Kaplan-Meier survival curves were estimated from the cohort overall survival data. Statistical significance associated with a difference in survival between two groups was estimated using the Peto & Peto modification of the Gehan-Wilcoxon test and the Cox proportional hazards regression model, as implemented in the R survival package. Based upon criticism of other gene signatures associated with cancer survival [70], the significance of the hazard ratio associated with a type 1 immune polarization bias was estimated by comparing the hazard ratio predicted by type 1 immune polarization model against a distribution in hazard ratios predicted from an ensemble of random models that were created by bootstrap resampling (= 1000). Each random model was created by randomly assigning a small subset of genes selected from the genes shown in Table 1 to one of four polarization states. The number of genes associated with each of the four polarization states was the same as listed in Table 1. A p–value corresponds to the likelihood that the observed hazard ratio (or a more extreme value) associated with type 1 immune polarization is due to random chance, where a p–value<0. 05 was considered significant. All analyses were performed using R software version 2. 14. 1 (http: //www. r-project. org). Overall, the study was performed in concordance with the REMARK guidelines [71]. The nontumorigenic human breast epithelial cell line 184A1 was obtained from ATCC (Manassas, VA), the human HER2+ breast cancer cell lines (BT474 and SKBR3) were kindly provided by Dr. Jia Luo (University of Kentucky; Lexington, KY), and a cell model of triple negative breast cancer (MDA-MB-231) was a gift from Dr. J. M. Ruppert (West Virginia University). The 184A1, BT474, and SKBR3 cell cultures were maintained at 37C in 5% CO2 in media supplemented as described previously [72]. Similarly, the MDA-MB-231 cell line was maintained in Dulbecco' s Modification of Eagle' s Medium (DMEM) supplemented with 10% (v/v) heat inactivated fetal bovine serum (FBS) (Hyclone, Inc. , Logan, UT), L-glutamine, and penicillin/streptomycin (BioWhittaker, Walkersville, MD). Allophycocyanin (APC) -conjugated mouse anti-human CD212 (IL12Rβ1 - Clone 2. 4E6) and APC-conjugated mouse anti-human HLA-A, B, C (Clone G46-2. 6) were purchased from BD Biosciences (San Diego, CA, U. S. A.). Phycoerythrin (PE) -conjugated mouse anti-human IL-12 receptor β2 (IL12Rβ2 - Clone 305719) was purchased from RnD Systems (Minneapolis, MN). ChromPure human IgG (whole molecule) were purchased from Jackson Immuno Research (West Grove, PA, U. S. A.). Quantum Simply Cellular uniform microspheres conjugated to anti-mouse IgG were purchased from Bangs Laboratories (Fishers, IN). Fluorescence-activated cytometry was performed as described [27], [38]. Quantum Simply Cellular calibration beads that contain four Quantum Simply Cellular microsphere populations with different mouse IgG antibody binding capacities were stained with fluorophore-conjugated monoclonal antibodies specific for IL12Rβ1, IL12Rβ2, or HLA-ABC. The cells were analyzed using a FACSAria flow cytometer and FACSDiva Version 6. 1. 1 software (BD Biosciences). No stain controls were used as negative flow cytometry controls. Single stain controls were used to establish fluorescent compensation parameters. Cellular events were identified by forward and side scatter characteristics. On average, events were analyzed. Flow cytometry data was exported as FCS3. 0 files and analyzed using R/Bioconductor [73]. | Effective anti-tumor immunity is proportional to the number and to the cytotoxic activity of immune cells that enter the tumor microenvironment. Recent advances in cancer immunotherapy stem from increasing the number of tumor-infiltrating immune cells by inhibiting immune checkpoints or adoptive T cell therapy. Here, we used computational methods to identify potential mechanisms present within the tumor microenvironment that limit the efficacy of anti-tumor immunity. Specifically, we found that oncogenic transformation is associated with the induction of tumor-derived biochemical cues, namely Wnt-inducible signaling protein-1, that locally suppress anti-tumor immunity. Moreover, we used model-based inference to demonstrate that a gene signature consistent with effective type 1 cell-mediated cytotoxic immunity is a predictor of overall survival independent of molecular pathology. Interestingly, patients with triple negative breast cancer were more enriched in the cohort associated with type 1 cell-mediated immunity. As this immune gene signature is not present in current genetically engineered mouse models of breast cancer, the results help identify design constraints for engineering better pre-clinical models of breast cancer. Demonstrating efficacy in pre-clinical animal models is a pre-requisite for bringing improved cancer immunotherapies into the clinic. | Abstract
Introduction
Results
Discussion
Materials and Methods | bioengineering
systems biology
biomedical engineering
medicine
oncology
forms of evolution
microevolution
immune evasion
immunotherapy
basic cancer research
cancer treatment
biology
computational biology
evolutionary biology
engineering | 2014 | Induction of Wnt-Inducible Signaling Protein-1 Correlates with Invasive Breast Cancer Oncogenesis and Reduced Type 1 Cell-Mediated Cytotoxic Immunity: A Retrospective Study | 10,774 | 287 |
The human cytomegalovirus (HCMV) is extremely prevalent in the human population. Infection by HCMV is life threatening in immune compromised individuals and in immune competent individuals it can cause severe birth defects, developmental retardation and is even associated with tumor development. While numerous mechanisms were developed by HCMV to interfere with immune cell activity, much less is known about cellular mechanisms that operate in response to HCMV infection. Here we demonstrate that in response to HCMV infection, the expression of the short form of the RNA editing enzyme ADAR1 (ADAR1-p110) is induced. We identified the specific promoter region responsible for this induction and we show that ADAR1-p110 can edit miR-376a. Accordingly, we demonstrate that the levels of the edited-miR-376a (miR-376a (e) ) increase during HCMV infection. Importantly, we show that miR-376a (e) downregulates the immune modulating molecule HLA-E and that this consequently renders HCMV infected cells susceptible to elimination by NK cells.
HCMV is a dsDNA herpesvirus, which is highly prevalent in the human population. It establishes its life long latency by using a diverse panel of sophisticated immune evasion strategies, and specifically by manipulating the expression of Human Leukocyte Antigen (HLA) class I molecules [1]–[4]. One of the fascinating examples with this regard is the viral protein UL40 that encodes a signal peptide similar in its sequence to the signal peptide of HLA class I molecules. The UL40 signal peptide is processed and loaded specifically in the groove of HLA-E and thus induces its surface expression, where it can inhibit NK cells function [5], [6]. Natural Killer (NK) cells are innate immune lymphocytes, which recognize and eliminate hazardous cells such as virally infected, transformed and damaged cells. The activity of NK cells is regulated by a balance of signals generated by activating and inhibitory receptors [7]–[10]. While the activating ligands of NK cells are diverse, most of the NK inhibitory ligands belong to the HLA class I family. Thus, reduction in the normal expression levels of HLA class I proteins will result in the activation of NK cells [11], [12]. Among the HLA class I proteins, the non-classical HLA-E is distinct in that it presents in its groove only a limited variety of peptides, which are primarily derived from the signal peptides of other HLA class I molecules [13], [14]. HLA-E is mainly recognized by the inhibitory CD94/NKG2A heterodimer and by the activating CD94/NKG2C receptors, both are expressed by T cells and NK cells [15]. Interestingly, expansion of NK clones expressing NKG2C is often seen HCMV-seropositive individuals [16], yet the function of these NK clones is unclear. Several activating NK cell receptors exist, that bind to divers ligands. The NKG2D receptor is one of the potent NK cell activating receptors, recognizing 8 different ligands: UPBP 1–6, MICA and MICB [17], [18]. While the ULBP family has orthologous in mice and in primates, mice do not encode MIC genes and most primates encode only a single MIC gene. We have shown in the past that the expression of MICA and MICB is controlled by cellular and by viral miRNAs [19]–[21]. Interestingly, one of the miRNA that was shown by us to target MICB, miR-376a [19], is known to be edited by ADAR enzymes [22]. RNA editing is a post-transcriptional modification that generates diversity in RNA molecules and in proteins. A-to-I RNA editing is catalyzed by the adenosine deaminase acting on RNA (ADAR) enzymes [23], [24], which bind dsRNA structures (that are not completely defined), and can thus edit protein-coding mRNAs and non-coding RNA molecules such as microRNAs (miRNAs). The editing of miRNAs can result in a change in the specificity of the miRNA, especially if the editing event occurs in the seed sequence [24]. Two active ubiquitously expressed ADAR genes are known, ADAR1 and ADAR2 [24]. The ADAR1 enzyme has two considerably different isoforms; ADAR1-p150, which is induced by type I interferon (IFN) [25] and is located mainly in the cytoplasm and ADAR1-p110, that is thought to be constitutively expressed and is located mainly in the nucleus [26]–[28]. ADAR proteins are physiologically crucial as knock-out of Adar1 is embryonic lethal [29]–[31] and knock-out of Adar2 results in behavioral abnormalities including epileptic seizures [32]. ADAR proteins also function during viral infections and were shown to be either proviral or antiviral [33]. Here we show that specifically during HCMV infection ADAR1-p110, and not ADAR1-p150, is induced. We further show that increased editing of miR-376a is observed upon HCMV infection and that the edited miR-376a (miR-376a (e) ), downregulates the expression of HLA-E to render infected cells susceptible to elimination by NK cells.
Because it is known that ADAR1-p150 is induced during several viral infections [34]–[37] and since it is unknown whether ADAR proteins are induced during HCMV infection we initially analyzed the expression of the two active ADAR proteins, ADAR1 and ADAR2, prior and following HCMV infection of Human Foreskin Fibroblasts (HFFs). ADAR2 was undetected in HFF cells prior (Fig. S1a), and following infection (data not shown). Surprisingly, although it is known that the expression of ADAR1-p150 is induced following interferon (IFN) treatment [25] (Fig. S1b, quantified in Fig. S1c), only the expression of ADAR1-p110, which was considered until now constitutive, significantly increased following HCMV infection (Fig. 1a, quantified in Fig. S1d). Next, we investigated whether ADAR1-p110 would be induced by additional HCMV strains. We infected cells with AD169, TB40/E, low passage Merlin strain, and with a clinical isolate (CI) and observed increased expression of ADAR1-p110 following infection by all strains (Fig. 1b, quantified in Fig. S1e). The induction of ADAR1-p110 was not cell specific as it was observed also in ARPE-19 cells (epithelial origin) infected with the TB40/E strain (Fig. 1c, quantified Fig. S1f). Next, to examine whether the induction of ADAR1-p110 is virus specific, we infected HFF cells with additional herpesviruses; HSV1 and HSV2, and did not observe any changes in ADAR1-p110 expression (Fig. 1d, quantified in Fig. S1g). We also infected A549 cells with the DNA virus Adenovirus (AdV, Fig. 1e, quantified in Fig. S1h), and with the RNA viruses Influenza and the human metapneumovirus (hMPV, Fig. 1f, quantified in Fig. S1i). We used A549 in these cases, as HFF cells are not permissive to infections with these viruses. Following infection we did not observe changes in ADAR1-p110 expression (Fig. 1e and f quantified in Fig. S1h and I, respectively). Yet, the expression of ADAR1-p150 was increased by these viruses (Fig. 1e and f quantified in Fig. S1h and I, respectively). Thus we conclude that following HCMV infection ADAR1-p110 is specifically induced. To elucidate the mechanism of ADAR1-p110 induction we cloned all known ADAR1 promoters upstream to a Firefly reporter. Expression of the ADAR1 gene is controlled by four alternative promoters, of which three drive the expression of ADAR1-p110 (Fig. 2a, genomic regions 1B, 1C and 2, white boxes, translation initiation at Met296) and a fourth promoter, IFN-inducible, that controls the expression of ADAR1-p150 (Fig. 2a, genomic region 1A, black box, translation initiation at Met1). The activity of each reporter was assessed under IFN-α or IFN-β treatments or during HCMV infection (Fig. 2b). As previously reported, the genomic region 1A induced the reporter' s activity following IFN treatment [25] (Fig. 2b). Importantly, while genomic regions 1C and 2 did not affect the activity of the Firefly reporter in the various treatments (Fig. 2b), genomic region 1B strongly induced Firefly activity, only following HCMV infection (Fig. 2b). We also assessed the activity of the various reporters in ARPE-19 cells infected with TB40/E strain and obtained similar results (Fig. 2c) To investigate at which stage of infection is ADAR1-p110 induced, we transfected HFF cells with the 1B reporter, and then infected the cells in the presence or absence of late-gene expression inhibitor, Ganciclovir (GCV, Fig. 2d). As can be seen, GCV treatment did not affect ADAR1-p110 induction (Fig. 2d). Indeed when we followed the kinetics of ADAR1-p110 induction we observed that the induction of the 1B reporter' s expression was very rapid, and was detected as early as six hours post infection (Fig. 2e, black bars). Moreover, it was not induced when a UV-inactivated virus was used (Fig. 2e, grey bars), indicating that the 1B promoter is activated only due to infection by a transcriptionally active virus, either by the virus itself or as a cellular response to HCMV infection. MiR-376a was shown to be one of the primary miRNAs that undergo editing by ADAR proteins [22]. Here we observed that ADAR1-p110 is specifically induced during HCMV infection (Figs. 1 and 2). Thus, we were next interested to evaluate the levels of the edited-miR-376a (named here miR-376a (e) ) during HCMV infection. As miR-376a and miR-376a (e) differ in only one nucleotide, it is impossible to distinguish between the two miRNAs by qRT-PCR. Hence we performed next generation small RNA sequencing on uninfected and HCMV infected HFF cells and observed an increase in miR-376a (e) levels following infection (Fig. 3a). To demonstrate that ADAR1-p110 has a role in the editing of miR-376a during HCMV infection we knocked down ADAR1. Because the transcript of ADAR1-p110 is included within the transcript of ADAR1-p150 it is impossible to knock down only ADAR1-p110 without affecting the expression of ADAR1-p150. We therefore prepared two shRNA constructs, one that targets ADAR1-p150 specifically (shADAR1-p150) and another that targets both isoforms of ADAR1 (shADAR1) and transduced HFF cells with these constructs. The KD of ADAR1-p110 by shADAR1 was very efficient as demonstrated by WB analysis of uninfected and HCMV infected HFF cells (Fig. 3b, quantified in Fig. 3c), while the KD of ADAR1-p150 affect only the expression of ADAR1-p150 (Fig. S2). Next, we evaluated the levels of miR-376a (e) in the shADAR1 cells following HCMV infection by small RNA next generation sequencing and observed reduced levels of miR-376a (e) as compared to control cells (Fig. 3d). To corroborate the above findings, we cloned ADAR1-p110 into a lentiviral vector and transduced HFF cells. The over expression of ADAR1-p110 was validated by WB (Fig. 3e, quantified in Fig. 3f). We analyzed the levels of miR-376a (e) by small RNA next generation sequencing and observed increased levels of miR-376a (e) specifically in ADAR1-p110-expressing HFF cells (Fig. 3g). So far we showed that ADAR1-p110 is induced specifically during HCMV infection and that ADAR1-p110 can edit miR-376a. Interestingly, we demonstrated in the past that miR-376a down regulates the expression of the stress-induced ligand MICB [19]. Because the editing of miR-376a occurs in its seed region, it is expected (as previously reported [22]) that the target spectrum of miR-376a (e) will be different than that of miR-376a. Indeed, miR-376a (e) does not have a full seed base-pairing with the 3′UTR of MICB (Fig. 4a). Thus, we next tested whether miR-376a (e) is able to control the expression of MICB. RKO cells were transduced with lentiviral vectors encoding miR-376a, miR-376a (e), or a control miRNA. As previously reported [19], a moderate, yet statistically significant reduction in MICB expression was observed in cells transduced with miR-376a (Fig. 4b, left, quantified in Fig. 4c), while the expression of other NKG2D ligands expressed on RKO cells was not affected (Fig. S3a). In contrast, little or no change in MICB levels was observed in cells transduced with miR-376a (e) (Fig. 4b right, quantified in Fig. 4c). Furthermore, dual luciferase reporter assays demonstrated that miR-376a (e) is unable to repress reporter' s activity when fused to the 3′ UTR of MICB (Fig. 4d). These observations were supported by NK killing assays against the transduced RKO cells in which the expression of miR-376a lead to reduced NK cell killing while expression of miR-376a (e) had no effect (Fig. 4e). Thus, we concluded as we previously reported [21], [38] that even a moderate change in MICB expression that is mediated by miR-376a is sufficient to affect NK cytotoxicity and that miR-376a (e) does not affect MICB expression. In light of future findings describe below, we also assessed the levels of HLA-E on RKO cells and found that they do not express HLA-E (Fig. S3b). We next transduced RKO cells with lentiviral vectors expressing either ADAR1-p110 or a control vector and assessed the levels of the NKG2D ligands (Fig. 4f, quantified Fig. 4g). Although RKO cell already express ADAR1 (Fig. S4a) overexpression of ADAR-p110 (black histograms), moderately elevated the levels of MICB only (Fig. 4f, quantified in Fig. 4g), probably due to reduced miR-376a levels. As an additional control we also overexpressed ADAR1-p110 in cells that do not express or express little amounts of MICB (293T and BJAB, but express ADAR1-p110, Fig. S4a) and observed no changes in the expression of MICB or in other NKG2D ligand (Fig. S4b). Thus, we concluded that editing of miR-376a abolishes its ability to regulate the expression of MICB. Because miR-376a (e) does not control the expression of MICB (Fig. 4) and since we showed that ADAR1-p110 is strongly induced specifically upon HCMV infection (Figs. 1 and 2) leading to increase levels of miR-376a (e) (Fig. 3). We next wondered whether miR-376a (e) might regulate the expression of immune genes that are known to be affected during HCMV infection. To that end we utilized the online algorithm RNAhybrid (http: //bibiserv. techfak. uni-bielefeld. de/rnahybrid/submission. html) to screen for possible miR-376a (e) binding sites in 3′UTR of various related genes. One of the top candidates was HLA-E, which was predicted to have two binding sites for miR-376a (e) in its 3′ UTR (Fig. 5a), yet no such sites were predicted for the unedited miR-376a. Whether HLA-E is regulated by miRNAs is unknown. Therefore, we initially tested this by knocking down the RNase III enzyme Drosha (one of the central enzymes involved miRNA biogenesis [39]). KD of Drosha (which was validated by WB, see Fig. S5a, quantified in Fig. S5b) resulted in elevated levels of HLA-E (Fig. 5b, quantified in Fig. 5c). To demonstrate that miR-376a (e) specifically, controls the expression of HLA-E we overexpressed miR-376a (e), miR-376a or a control miRNA in BJAB cells (we selected BJAB cells because they express HLA-E). Indeed reduced HLA-E levels were observed only in cells transduced with miR-376a (e), whereas the unedited miR-376a or control-miRNA had no effect (Fig. 5d, quantified in Fig. 5e). To validate that miR-376a (e) regulates HLA-E by directly binding to the predicted binding sites, we fused the 3′ UTR of HLA-E downstream to a Firefly reporter. Dual luciferase assays were performed in cells transduced with lentiviruses expressing miR-376a (e), miR-376a or control miRNA and then transfected with the Firefly reporter. While expression of miR-376a did not affect the reporter' s activity (Fig. 5f), repression was observed in cells expressing miR-376a (e) (Fig. 5g). To demonstrated that miR-376a (e) regulates HLA-E expression by direct binding to the predicted sites (Fig. 5a), we generated reporters bearing single (mut187 or mut1342) and double (mut187 and mut1342, named Dmut) mutations in the predicted binding sites (Fig. S6). All mutant reporters abolished the miR-376a (e) -mediated repression (Fig. 5g). Thus, we concluded that miR-376a (e) directly binds the 3′ UTR of HLA-E at the predicted binding sites and that both binding sites are necessary for the regulation of HLA-E by miR-376a (e). Finally, qRT-PCR analysis of the relative abundance of HLA-E mRNA in cells transduced with miR-376a (e) demonstrated no effect as compared to control cells (Fig. 5h), suggesting that miR-376a (e) represses HLA-E expression through translational inhibition. Because we demonstrated that ADAR1-p110 and editing of miR-376a are induced specifically following HCMV infection and since we showed that miR-376a (e) regulates HLA-E, we next tested whether miR-376a (e) controls HLA-E during HCMV infection. We initially validated that the miR-376a (e) binding sites in the 3′ UTR of HLA-E are targeted during HCMV infection. HFF and ARPE-19 cells were transfected either with the WT HLA-E 3′ UTR Firefly reporter or with the Dmut reporter and then the cells were infected with the AD169 (HFF cells) or the TB40 strains (ARPE-19 cells). The reporter' s activity was repressed by both HCMV strains only when it was fused to the WT 3′ UTR of HLA-E and not when fused to the mutant 3′UTR (Fig. 6a). We next proceeded to investigate whether ADAR1 has a role in the regulation of HLA-E during HCMV infection. To that end HFF cells were transduced either with shADAR1, with shADAR1-p150 or with a control, infected with HCMV and then the levels of HLA-E were assessed by FACS (Fig. 6b, quantified in Fig. S7a). HLA-E levels increased in all cells following infection (probably due to the binding of the UL40 leader peptide), however the shADAR1 transduced HFF cells showed an even higher expression of HLA-E (Fig. 6b, quantified in Fig. S7a). We next wanted to demonstrate that miR-376a (e) specifically controls the endogenous expression of HLA-E during HCMV infection. For this we generated an anti-miRNA sponge construct, consisting of six adjacent binding sites for miR-376a (e), fused downstream to an eGFP cassette. Sponge constructs sequester the specific miRNA from its original targets, enabling their expression [40]. HFF cells were transduced with a control sponge (See Text. S1: Supporting Methods) or with an anti-miR-376a (e) sponge, and then infected with HCMV. The levels of HLA-E and of classical HLA class I proteins were analyzed by FACS (Fig. 6c, quantified in Fig. S7b). The expression of HLA-E was induced following infection however in HCMV infected HFF cells transduced with the anti-miR-376a (e) sponge the expression of HLA-E was significantly higher than that in the control cells (Fig. 6c, red histogram, quantified in Fig. S7b). Infections of all cells was similar as an equivalent reduction in the levels of classical HLA class I molecules was observed (Fig. 6c, right). Importantly, increased levels of HLA-E following transduction of HFF cells with the anti-miR-376a (e) sponge, were observed following infection with AD169, Merlin, CI and the TB40/E strain (Fig. 6d, quantified in 6e). Thus we concluded that following HCMV infection ADAR-p110 is induced, miR-376a is edited and that the edited miRNA, miR-376a (e), regulates the expression of HLA-E. After we demonstrated that antagonizing ADAR1-p110 and miR-376a (e) affects HLA-E during HCMV infection we next wanted to test the functional implications of these findings. HCMV infected HFF cells were initially transduced with the shADAR1 lentiviruses indeed showed reduced NK killing (data not shown). Yet, when the interaction between HLA-E and CD94 (the chain that recognizes HLA-E) were blocked, only a partial restoration of killing was observed (data not shown), suggesting that the ADAR1 proteins affect other molecules and pathways, in addition to HLA-E, that control NK cell activity. Thus to demonstrate the functional effect miR-376a (e) has on NK killing of HCMV infected HFF cells we utilize the sponges mentioned above and for that we initially verified the sponge' s specificity. HFF cells were transduced with the sponges against miR-376a and against miR-376a (e) and MICB levels were evaluated. In agreement with our above results in which we showed that miR-376a (e) does not control the expression of MICB (Fig. 4b), only the anti-miR-376a sponge had an effect on the levels of MICB in uninfected HFF cells (Fig. 7a). In contrast, during HCMV infection only the anti-miR-376a (e) sponge led to increase levels of HLA-E (Fig. 7b). We have also tested whether the sponges affect MICB expression following HCMV infection and observed that none of the sponges affect MICB expression during infection (Fig. S8). This is because the anti-miR-376a (e) sponge does not antagonize MICB (Fig. 7a) and because, as we have previously shown the anti-miR-376a sponge by itself is not sufficient to alter the levels of MICB during HCMV infection due to a synergistic interaction with the viral miR-UL112 [19]. HLA-E is the known ligand of two immune receptors; the inhibitory heterodimer CD94/NKG2A and the activating heterodimer CD94/NKG2C [15], [41]–[43]. It was demonstrated that the proportion of NKG2C+ NK cells increases significantly after HCMV infection [16], yet the functional significance of this expansion is still unclear. Thus, to test the functional significance of the miR-376a (e) -mediated control of HLA-E during infection we isolated NK cells from four HCMVneg and four HCMVpos donors (Fig. 7c). Consistent with previous reports [16], the proportion of the NKG2C+ NK population was indeed higher in the HCMVpos as compared to HCMVneg donors, while the proportion of NKG2A+ NK cells was lower in the HCMVpos donors (Fig. 7c). In contrast, the proportion of CD94+ NK cells (CD94 is the common chain of the NKG2A and NKG2C heterodimers that recognizes HLA-E) did not significantly change between donors (Fig. 7c). Next, HFF cells transduced either with the anti-miR-376a (e) -sponge or with a control sponge, were infected with HCMV and subjected to killing by NK cells derived from HCMVneg and HCMVpos donors. Irrespective of the donors' NKG2C+ NK cells percentage, the expression of the anti-miR-376a (e) -sponge which led to increased levels of HLA-E (Fig. 6c), resulted in inhibition of NK function and reduced killing of infected cells (Fig. 7d). Blocking of CD94 led to equivalent levels of killing of all cells by all donors (Fig. 7d). We also performed the reciprocal experiments in which we transduced HFF cells with miR-376a (e) (which leads to the down regulation of HLA-E). In this scenario, the killing of the infected HFF was increased (due to reduced inhibition) and as above, the increased killing was observed by all donors, irrespective of their phenotype (Fig. 7e). Blocking of CD94 restored the killing to that observed with cells infected by control miRNA (Fig. 7e). Together, these results demonstrate the dominance of the inhibitory signal generated by NKG2A in controlling NK cell activity and the importance of the regulation of HLA-E by miR-376a (e) during HCMV infection. We next wanted to investigate the influence of ADAR1-p110 induction on HLA-E expression in vivo. However, HCMV is very different from the mouse CMV (MCMV) and as the human virus infects only humans there is practically no murine model that enables us to investigate HCMV infection in vivo. Moreover, no sites for the mouse orthologue of miR-376a (e) (mmu-miR-376b (e) ) were predicted in the 3′UTR of the mouse orthologue of HLA-E, Qa-1b. HCMV is highly prevalent in the human population and is the leading cause of congenital infection, associated with severe birth defects and intrauterine growth retardation. The decidua is the maternal part of the placenta through which infection of HCMV is probably transmitted to the fetus. Recently, a new organ culture model was established in which human decidua is maintained, ex-vivo, and infected with HCMV [44]. This unique model enables us to examine some of our observations under physiological conditions. Decidual organs were infected with the AD169 strain of HCMV that expresses IE72-GFP enabling the detection of infected cells (GFPpos). As expected, the HCMV infected cells (GFPpos) demonstrated typical enlarged cell morphology (Fig. 8). The organ cultures were next immunofluorescencenly stained for ADAR1 or HLA-E expression either early (36 hrs) or late (7days) during HCMV infection. In agreement with our above results, HCMV infected cells (GFPpos cells) demonstrated elevated nuclear ADAR1-p110 staining at both early and late time points (Fig. 8a–d). In contrast, HLA-E levels were initially increased in HCMV infected cells (GFPpos, Fig. 8e and 8f), probably because of the viral UL40 protein, but were then reduced as infection prolonged (Fig. 8g and 8h), probably due to the editing of miR-376a (e) which results in the down regulation of HLA-E.
ADAR1 encodes two isoforms: ADAR1-p110, which was until now thought to be constitutively expressed, and ADAR1-p150, which is IFN-induced. These two isoforms are located at different compartments in the cell; ADAR1-p110 is located primarily in the nucleus, and ADAR1-p150 is located mainly in the cytoplasm [24]. It is therefore expected that each isoform will have a specific range of RNA targets, and thus different roles in different physiological conditions. The role of ADAR1 proteins (especially ADAR1-p150) during infection was mostly investigated in the context of RNA viruses such as influenza virus [34], hepatitis C virus [35] Rift Valley fever virus [36] and mumps virus [37]. ADAR enzymes can also edit transcripts of DNA viruses (including viral miRNAs [33]), but in DNA virus infections it is unknown which ADAR protein mediates these editing events and what are the functional consequences of such editing. Herein we demonstrate that the expression of ADAR1-p110 is induced during HCMV infection. We further show that this induction in ADAR1-p110 expression is HCMV specific and that it was not observed when other viruses were used. Additionally we show that miR-376a is edited during HCMV infection and that knockdown or over expression of ADAR1-p110 affects miR-376a editing, suggesting that ADAR1-p110 controls the editing of miR-376a during HCMV infection. Whether ADAR1-p110 is induced by a viral component or by cellular is still unknown. We have shown that the kinetics of ADAR1-p110 induction is very rapid, and that the anti-viral drug Ganciclovir can not block this induction. In an attempt to investigated whether one of the immediate early genes of HCMV are responsible for the ADAR1 induction we expressed the viral transcription activators IE72 and IE86 in HFF cells, but did not observe any change in the levels of ADAR1-p110 (data not shown). Thus, further work is required to reveal whether it is a cellular or a viral component that is required for the ADAR1-p110 induction. We also show that HLA-E, which was not known previously to be regulated by miRNAs, is downregulated by miR-376a (e) and that this consequently leads to the elimination of the HCMV infected cells by NK cells. The question of why HCMV “allows” the induction of ADAR1 is interesting; particularly in view of the data we provided which show that the ADAR1-p110 induction is HCMV specific. It is possible that HCMV itself is dependent on cellular components of RNA editing for successful gene expression and infection (as in the case of viral miRNAs [45]) and thus can not evade the anti-viral response mediated by ADAR1-p110. Indeed it was shown that ADAR enzymes can edit transcripts of DNA viruses [33]. It is also possible that the virus developed mechanisms, which are yet unknown, to temper with the ADAR1-p110 response. This might be in similarity to what has been recently shown regarding IFI16 sensing of HCMV and the counter mechanism of the virus mediated by pUL83 [46]. Perhaps this balanced response enables the co-existence of the virus with its host. As we have demonstrated that an UV-inactivated virus does not induce ADAR1-p110, it is tempting to speculate that a viral component controls ADAR1-p110 induction. Yet, it is also possible that upon live virus infection cellular-downstream elements trigger the induction of ADAR1-p110. Thus whether the induction of ADAR1-p110 is an anti-viral response or a viral-mediated tactic that the host turns to its advantage is yet to be determined. Previously we showed that upon HCMV infection the virus downregulates the activating ligand MICB by the viral miRNA miR-UL112 [47]. We also demonstrated that miR-UL112 acts synergistically with the cellular miR-376a (that also regulates MICB), and that this leads to reduce killing of the infected cells [19]. Here we show that editing of miR-376a renders the infected cells susceptible to NK cell cytotoxicity. We were able to distinguish between these two seemingly contradicting observations because we used two different viruses. In all current experiments we had to use the AD169 laboratory strain of HCMV as similarly to clinical strains it encodes a UL40 protein with a leader peptide capable of being presented in the groove of HLA-E (GenBank accession number FJ527563). In our previous work we used the TB40/E strain that has a mutation in UL40 (GenBank accession number EF999921) and is unable to be presented by HLA-E and thus is unable to induce HLA-E expression [48]. Therefore, when TB40/E is used editing of miR-376a has no effect on HLA-E levels and we could nicely observe that the infected cells were less susceptible due to the viral and cellular miRNAs operating to reduce NK cell killing [19]. It is difficult to determine which strategy is superior to which, the viral down regulation of MICB or that of the host through increased editing and down regulation of HLA-E. During infection, both the virus and the host employ many mechanisms that affect the progress of infection. As HCMV is known to infect in vivo a wide rage of cells and furthermore as the virus has two modes of infection (latent and lytic) the question of who has the upper hand in this battle - the virus or the host, becomes very complicated. Thus, whether an infected cell will be killed or not depends not only on the specific cell in question but also on the mode and stage of infection as well as on the balance between the anti-viral and the viral strategies as shown here and in our previous publications [19], [49]. It was demonstrated here and previously that HCMVpos individuals have increased proportion of NKG2C+ NK cells [16]. HCMV infection is a prerequisite for the expansion of these NKG2C+ NK cells as infection with other viruses, such as Hantaviruses, induces the expansion of the NKG2C+ population only in HCMVpos individuals [49]. Although the expansion of the NKG2C+ NK cells is well established the role and functional significance of the NKG2C+ NK cells is still unclear and actually it is thought that it is only few NKG2C+ NK clones that expand [49]. Importantly, it was demonstrated that inhibition by NKG2A is dominant over the NKG2C activation [47]. Our current results are in line with these observations as we show that although the proportion of NKG2C+ cells is higher in HCMVpos individuals, the outcome of neutralizing miR-376a (e) during infection (which leads to increased HLA-E levels) is inhibition of NK cell function irrespective of the donors' serotype. The new mechanism of HLA-E regulation via miRNA editing discovered here is unique to humans. The mouse orthologous protein of HLA-E, Qa-1b, does not contain predicted binding sites for the mouse orthologue of miR-376a (e). Thus, to demonstrate the induction of ADAR1 and reduction in HLA-E expression following HCMV infection in physiological settings, we used a decidua organ culture model [44] and observed that during HCMV infection ADAR1-p110 expression is induced and that HLA-E expression is reduced. This unique system could be used in the future to study exciting questions regarding the relationships between the virus and its human host.
The following antibodies were used: anti-hMICA (159227), anti-hMICB (236511), anti-hNKG2A (131411), anti-hNKG2C (134591) and anti-hCD94 (131412, blocking) all were purchased from R&D systems. The W6/32 hybridoma was purchased from ATCC. For HLA-E detection, the MEM-E07/08 antibodies were used (generously provided by V. H.), Cloning and generation of lentiviruses and lentiviral vectors (ADAR1 knockdown, over-expression and sponge) was as previously described [50]. The specific sequences are listed in Text S1: Supporting Methods. The Drosha-knockdown vector and control vector were purchased from Sigma Aldrich and encode for puromycin resistance (HeLa cells were grown in the presence of 3. 5 ug/ml puromycin). In the present study, the following viruses were used: HCMV (strains: AD169, Merlin TB40/E and a clinical isolate), HSV1, HSV2, Adenoviruse, hMPV and Influenza (the A/PR8 strain). HCMV UV inactivation prior to viral adsorption performed with the UV Stratalinker 2400 (StrataGene) at 0. 99 Joule. Viral inactivation was confirmed in plaque assays. Total RNA was extracted with TRI reagent (Sigma), and was treated with Turbo-DNase (Ambion). For generation of cDNA libraries the M-MLV reverse-transcriptase (Invitrogen) was used for reverse transcription (according to manufactures instructions), in the presence of a poly dT primer. The 3′ UTRs of MICB and of HLA-E were cloned downstream to a Firefly reporter in the pGL3 vector (Promega) via the XbaI restriction site. Twenty-four hours prior to transfection of the reporter plasmids, cells were plated to 50% confluence in 24well plates, in triplicates. 200 ng/well of the pGL3 vector and 50 ng/well of the pRL-CMV vector were transfected with the TransIT-LT1 transfection reagent (Mirus Bio), and relative Firefly activity was assessed 48 hrs post transfection. Firefly/Renilla activity ratio was normalized to that in the control cells, and then relative activity of the reporter was calculated. The various ADAR1 promoter genomic regions were cloned upstream to a Firefly reporter via the XhoI and NheI restriction (promoters 1B, 1C and 2), or NheI and HindIII (promoter 1A) in the pGL4. 14 vector (Promega). Forty-eight hours prior to transfection of the reporter plasmids, HFF cells were plated to 60% confluence in 24well plates, in triplicates. 250 ng/well of the pGL4. 14 vector and 50 ng/well of the pRL-CMV vector were transfected with the TransIT-LT1 transfection reagent (Mirus Bio). Four hours after transfection, the media was removed, cells were washed, and media containing the reported treatment was added (mock, 1000 units/ml of IFN-α (Peprotech), 1000 units/ml IFN-β (Peprotech) or infection with AD169 strain of HCMV at MOI 1). Unless stated otherwise, Firefly activity was assessed 48 hrs post transfection. Firefly/Renilla activity ratio was calculated relatively to that in the mock cells. Primers for the amplification and mutations in the 3′ UTR of HLA-E and for the cloning of the various genomic regions are listed in Text. S1: Supporting Methods. Lysates of cells were prepared by lysing cells in ice-cold 0. 6%SDS in 10 mM Tris (pH 7. 5) buffer containing a cocktail of protease inhibitors (Roche). Lysate were separated using SDS–polyacrylamide gel electrophoresis and transferred to a NitroCellulose membrane. The following antibodies were used for protein detection: anti-hHLA-E (MEM-E02, generously provided by V. H.), anti-hADAR1 (SigmaAldrich, prestige antibodies, HPA003890), anti-hADAR2 (SigmaAldrich, clone ADAR2-8), anti-Drosha (Abcam, polyclonal), anti-αTubulin (Santa Cruz Biotechnology, B-7). Quantification of blots was performed with the ImageJ software. NK cells were isolated from healthy donors via MACS separation kit (Miltenyibiotech) and grown in the presence of IL-2 (peprotech). Target cells were grown over night in the presence of 35S added to a Methionine-free media (Sigma). Prior to incubation with the effectors, cells were washed, counted, and 5000 cells/well were plated. For each target, the spontaneous 35S release was calculated by cells which were not incubated with effector cells, and maximum 35S release was calculated by applying 0. 1M of NaOH to the target cells. The level of 35S release was measured after 5 hours of incubation with effectors (at 37°c) by a β-counter TopCount (Packard). When blocking antibody was used, NK cells were preincubated with 0. 5 ug/well of the blocking antibody on ice for 1 hour, and then the 35S labeled target cells were added for 5-hour incubation. Decidual tissues from women undergoing first-trimester elective pregnancy terminations were obtained by deep scraping to obtain maternal tissue from the basal plate and placental bed encompassing the decidua with interstitial trophoblastic invasion. The study was approved by the Hadassah Medical Center Institutional Review Board and was performed according to the Declaration of Helsinki, good clinical practice guidelines, and the human experimentation guidelines of the Israeli Ministry of Health. All donors gave written informed consent. Preparation of organ culture was as described [45]. For infection of decidual organ cultures, immediately after the sectioning the tissues were placed in 48-well plates (∼5 slices/well to maintain optimal viability) and were inoculated with the virus (40 PFU/well). The following antibodies were used: anti-hHLA-E (MEM-E02, diluted 1∶500, generously provided by V. H.), anti-hADAR1 (SigmaAldrich, prestige antibodies, HPA003890, diluted 1∶650), detected by Goat anti-mouse DyLight647 or Goat anti-Rabbit DyLight647 (Jackson). Images were obtained using a Leica SP50 confocal microscope, using a 40× (1. 25 NA) Leica oil objective. ImageJ software was used for image processing and quantification. Single cell intensities were normalized to the average fluorescence intensity of all cells after reduction of background intensity. The Illumina HiSeq 2000 platform was used. All sequenced samples underwent initial pre-processing prior to differential expression analysis. Briefly, sequencing adapters were clipped and bases with quality lower than 30 were removed from both ends of each sequenced read. Reads shorter than 16 nucleotides after clipping and trimming were discarded. Remaining reads were mapped against the mature human miRNA reference sequences (downloaded from miRBase [51], allowing up to 2 mismatches between read and reference using BWA [52]. Reviewing the reads mapping to miR-376a (MIMAT0000729), a pileup of the supported alleles and their sequencing quality was produced using SAMtools [52]. The expected sequencing error rate was calculated per-position and a binomial cumulative distribution was implemented to differ significant modifications from sequencing errors. | The human cytomegalovirus (HCMV) infects a high percentage of the human population. HCMV infection is life threatening to immune-compromised individuals and when transmitted to the fetus can cause severe birth defects. Thus, it is crucial to understand the anti-viral strategies that are induced in response to HCMV infection. By using molecular tools, next generation deep sequencing of small RNAs and a novel human decidual organ culture we demonstrate that in response to HCMV infection the cell launches an anti-viral response that via miRNA editing enables the elimination of HCMV infected cells. We find that following HCMV infection a specific ADAR1 isoform, ADAR1-p110, is strongly induced and that in addition editing of miR-376a is also increases. We continue to show that the edited-miR-376a then downregulates the inhibitory molecule HLA-E, and by doing so promotes the elimination of HCMV infected cells by NK cells. Finally we demonstrate that in decidual tissues ADAR1-p110 is induced and that HLA-E levels are reduced in response to HCMV infection. | Abstract
Introduction
Results
Discussion
Materials and methods | medicine
infectious diseases
immune cells
nk cells
immunology
cytomegalovirus infection
biology
viral diseases | 2014 | MicroRNA Editing Facilitates Immune Elimination of HCMV Infected Cells | 10,639 | 273 |
The Joint Evolutionary Trees (JET) method detects protein interfaces, the core residues involved in the folding process, and residues susceptible to site-directed mutagenesis and relevant to molecular recognition. The approach, based on the Evolutionary Trace (ET) method, introduces a novel way to treat evolutionary information. Families of homologous sequences are analyzed through a Gibbs-like sampling of distance trees to reduce effects of erroneous multiple alignment and impacts of weakly homologous sequences on distance tree construction. The sampling method makes sequence analysis more sensitive to functional and structural importance of individual residues by avoiding effects of the overrepresentation of highly homologous sequences and improves computational efficiency. A carefully designed clustering method is parametrized on the target structure to detect and extend patches on protein surfaces into predicted interaction sites. Clustering takes into account residues' physical-chemical properties as well as conservation. Large-scale application of JET requires the system to be adjustable for different datasets and to guarantee predictions even if the signal is low. Flexibility was achieved by a careful treatment of the number of retrieved sequences, the amino acid distance between sequences, and the selective thresholds for cluster identification. An iterative version of JET (iJET) that guarantees finding the most likely interface residues is proposed as the appropriate tool for large-scale predictions. Tests are carried out on the Huang database of 62 heterodimer, homodimer, and transient complexes and on 265 interfaces belonging to signal transduction proteins, enzymes, inhibitors, antibodies, antigens, and others. A specific set of proteins chosen for their special functional and structural properties illustrate JET behavior on a large variety of interactions covering proteins, ligands, DNA, and RNA. JET is compared at a large scale to ET and to Consurf, Rate4Site, siteFiNDER|3D, and SCORECONS on specific structures. A significant improvement in performance and computational efficiency is shown.
Interface residues are essential for understanding interaction mechanisms and are often potential drug targets. Reliable identification of residues that belong to a protein-protein interface typically requires information on protein structures [1] and knowledge of both partners. Unfortunately, this information is often unavailable and for this reason, reliable site prediction using a single protein, independently from its partners, becomes particularly valuable. Interactions of a protein with ligands, other proteins, DNA or RNA are all characterized by sites which either are conserved, present specific physical-chemical properties or fit a given geometrical shape [2], [3]. At times, the interface presents a mixture of these three signals. Interfaces differ from the rest of the protein surface typically because buried interface residues are more conserved than partially buried ones and because the sequences associated with interfaces have undergone few insertions or deletions. However, on average, the most conserved patches of residues overlap only the 37. 5% (±28%) of the actual protein interface and an analysis of 64 different types of protein interfaces (formed from close homologs/orthologs or from diverse homologs/paralogs) demonstrated that conserved patches cannot clearly discriminate protein interfaces [4]. The composition of interacting residues appears to distinguish between different types of interfaces [5], [6]. In particular, hydrophobic residues [7] and specific charge distributions [5], [8] have been shown to be characteristic of protein-protein interfaces. Protein interaction sites with ligands, DNA and RNA are usually highly conserved and the signal of conservation is likely to be sufficient for good predictions. The same does not hold true for protein-protein interfaces, where we show that combining information coming from conservation and the specific physical-chemical properties of the interacting residues, enhances the signal. We propose a predictive method, named Joint Evolutionary Trees (JET), that extracts the level of conservation of each protein residue from evolutionary information, combines this information with specific physical-chemical properties of the residues, and predicts conserved patches on the protein surface of known three-dimensional structures. Defined in this way, JET is able to detect protein interfaces with very different types. It does not require information on potential interaction partners and it belongs to the family of methods which have been inspired by the Evolutionary Trace approach (ET) [9], [10]. Similarly to ET, JET analyzes a protein sequence and structure, and finds information (from a careful analysis of the evolutionary distances between sequences homologous to) on binding interfaces by detecting conserved patches on the surface of the structure of. JET has been designed with large-scale applications in mind which requires the approach to be adjustable for different datasets and to guarantee predictions even with weak signals. Because of this, various evolutionary hypotheses on protein interfaces have been tested and new methodological approaches have been developed within JET. Two main hypothesis on interaction sites have been tested. The first asserts that specific physical-chemical properties of patches always co-exist with some degree of conservation of the patch. The second claims that interaction sites on a protein surface are composed of an internal core which is conserved, with concentric layers of residues around the core which are progressively less conserved. We also addressed four main methodological points. The first concerns the problem of accurately quantifying the strength of residue conservation in a set of sequences whose similarity to has been automatically evaluated by PSI-BLAST. This means reducing the interfering effects of sequences wrongly selected by PSI-BLAST (that is, sequences that are not homologous to) on the topology of the associated distance tree, and ensuring, as far as possible, diverse sequence identity within the samples. To this end, we introduce a new discrete combinatorial paradigm of computation to investigate potentially large sets of biological sequences by randomly sampling small subsets a sufficient number of times to ensure statistical overlap of the sampled sets. This method turns out to be powerful and also computationally efficient. The second point concerns the core of the ET methodology which relies on the definition of a trace, a notion that quantifies the conservation of a residue position within a distance tree of sequences similar to and that was originally introduced in [9]. This definition turns out to be insufficient to properly characterize residue conservation and a “hybrid” definition was proposed in [11] which combines the original notion of a trace, based on tree topology, with information entropy of the residue position within the pool of aligned sequences. In JET, we clarify the limits of the original combinatorial definition by redefining a trace based on tree topology and demonstrate that information entropy is not required. The third point concerns the evaluation of patches of conserved residues as potential internal cores of interaction sites. We tested the hypothesis that such cores correspond to the largest patches found for the protein and observed that this is generally the case. A novel method estimating the size of relevant clusters of conserved residues and of clusters of residues with specific physical-chemical properties has been tailored around the specific protein being treated. The method is based on a random generation of clusters over the protein surface of the protein in question. An evaluation of the size of a cluster based on a random generation is used also in [11]. The important difference between the two approaches is that, in the latter case, the estimation is made for arbitrary proteins. Finally, since JET is based on the random choice of small sets of sequences for constructing multiple trees, it could yield slightly different answers in different runs. This fluctuation has been analyzed and exploited to further improve our algorithm. An iterative version of JET (iJET) provides a list of consensus residues belonging to interaction patches. When JET is used for large-scale analyses, this turns out to be a safe and successful approach. When the user uses JET on a single protein, it is possible to run it once, or to explore the set of potentially interacting residues by varying a consensus threshold during iterations. In difficult cases, this can allow the user to refine the detection of interacting residues.
JET performs a PSI-BLAST search [12] at http: //www. ncbi. nlm. nih. gov/Blast. cgi, or locally, to select as many as 5000 sequences. It does it on chains with at least 20 residues. Retrieved sequences are filtered to eliminate redundant sequences, that is sequences with >98% sequence identity to, and to eliminate very divergent sequences, that is sequences with <20% sequence identity. A second filter is defined on the length of the alignment which should cover at least the 80% of the length of the reference sequence, and on the number of inserted gaps which should be <10% of the size of the alignment. A third filter cuts-off sequences with an e-value ≥10−5. If the pool of remaining sequences does not contain at least 100 sequences, then the cut-off on the length of retrieved alignments is automatically decreased by 10% of the length of progressively until reaching 51% of the length of the reference sequence (this condition ensures that all selected sequences will overlap with each other). If the number of sequences retrieved is insufficient, we reset the length of the alignment to 80% of the length of the reference sequence and restart the analysis with an e-value of 10−4. We repeatedly increase the e-value and decrease the length by filtering sequences progressively with e-values 10−3,10−2,10−1,1, 10,100, until a sufficient number of sequences is retrieved. At the end of the retrieval step we obtain a set of selected sequences. We want to align small sets of sequences in approximately times. With the purpose of using most of the information contained in and to guarantee overlapping of sequences among trees, we set whenever and we fix otherwise. Each set of sequences contains the reference sequence. Since the distribution of sequences based on sequence identity might not be uniform, we order sequences in in four classes characterized by 20–39% (including 20 and 39), 40–59%, 60–79%, and 80–98% sequence identity. This ensures a comparable set of representatives for different groups of identity within each set of aligned sequences. We then randomly select distinct sequences from each class. (If is not an integer, we pick the remaining sequences, that is sequences, successively, starting from the class of sequences characterized by the smallest sequence identity.) We require that each class contains enough sequences to ensure diversity within the generated alignments. Ideally, this corresponds to requiring that the inequality holds, where is the number of distinct sequences in the class with. In practice, we may find classes with insufficiently varied sequences to supply the sets to be aligned. In this case, if the class is empty, we ignore it. If it is not empty, we decrease the number of sequences to pick up within this class to a maximum such that. We pick the missing sequences from the other classes, satisfying. We order the classes with respect to the combinations and choose the sequences starting from the class with greatest value. In the event that there is a class where the inequality cannot be satisfied due to lack of sequences, we decrease the coefficient 2 within the inequalities (for all) by a maximum of five steps towards the coefficient 1. For each step we apply the procedure above to the new class of inequalities. This way, we obtain a good compromise between an ideally uniform distribution of sequence identities within an alignment and the diversity of sequences amongst different alignments. Sequences in a pool are aligned using CLUSTALW with the Blosum62 matrix [13]. The Score Distance method [14] has been used to define the distances between sequences obtained by the alignment; no contribution is made for gaps in the sequence nor by the ends. To align distantly related proteins, Gonnet [15] and HSDM [16] matrices are preferable and an automatic selection between Blosum62, Gonnet and HSDM has been implemented in JET. The criteria is as follows. Given an alignment of two sequences the score distance method computes the effective score of the alignmentwhere is the score produced by the alignment using a substitution matrix, , is the e-score value of the matrix (for Blosum62, for Gonnet, and for HSDM) and N is the number of pairs of aligned residues. Based on this, one computes distances between two sequences as To properly compute distances, one needs to guarantee. In the case of distantly related proteins, it is possible that and the value can become negative. When this occurs for some pairs using Blosum62, we take sequences and (whenever different from the reference sequence) out of the set and recompute distances until the condition is satisfied for all pairs. We require that the number of sequences in the tree covers 75% of the original number of sequences and is ≥10 (this corresponds to the minimal size of an acceptable tree). If at least one of these conditions is not satisfied then we repeat the analysis using the Gonnet method. If this also fails to pass the test the HSDM method will be used. For each multiple alignment, a distance tree is constructed based on the Neighbor Joining algorithm (NJ) [17]. The midpoint rooting method is used to find the point that is equidistant from the two farthest points of the tree, and to root the tree there. If are two nodes belonging to a branch of, let be the distance between and provided by the tree construction. The root of has rank 1. A node, which is not a leaf, has rank, if all nodes of such that have rank and at least one of them has rank. If two nodes (which are not leaves) are such that then their rank is the same. The maximum rank definable on a tree is, that is the number of sequences in See Figure 1, top. Consensus sequences of rank n and backtrace sequences of rank n are used to define tree traces. Let be the sequence associated with the leaf in. A consensus sequence associated to a leaf of is a sequence (of the same length as) where position is occupied by the residue in aligned to the i-th residue of If no residue in is aligned to the i-th position of then a gap will appear in the consensus sequence. A consensus sequence of a node of rank n is a sequence (of the same length as) where the position is occupied by those residues common to the consensus sequences associated with the children of. See Figure 1, top. A back-trace sequence of a node of rank, is a sequence (of the same length as) which records all residues in the consensus sequence associated to that do not already belong to the back-trace of the father of. The back-trace sequence of the root is the consensus sequence of the root. See Figure 1, bottom. Given, let be a node in with rank; we look at all positions along the branches of such that and we collect in a set subtrees of associated with positions of level as follows: given a position of level along some branch (defined below), we include the subtree of rooted at this point in only if the subtree contains more than two nodes; if the position coincides with a branching node of, then we include two copies of the subtree in. Each subtree in has a backtrace associated to its root. A tree trace of level is a residue which is not a tree trace of level and that occurs in backtraces of at least 2 subtrees in. A residue in the backtrace sequence of the root of is conserved in all sequences, in particular in, and it is called a tree trace of level 1. Notice that this definition is much weaker than the corresponding definition of trace for ET. In fact, in ET, a residue is a trace of level only when the residue is conserved in all subtrees of. See Figure 2. The set of tree traces resulting from the analysis of all generated metric trees will be used to define the relative trace significance for the residues in the PDB structure. Let be the generated trees, and index the residue positions in. We say that a residue at position in is a trace with degree of significancewhere is the tree trace level of residue in tree, is the maximum level of and is the number of trees where the residue appears as a trace. Values vary in the interval [0,1], and represent an average over all trees of the residue importance: traces appearing often at small (big) levels will get values close to 1 (0). We can consider in the formula to be smaller than the maximum level attainable, that is. This corresponds to the 95% (a default parameter of the method) of residues which have a trace value for a tree. Note that the condition does not imply that some residues have no trace (indeed traces are read out of many trees). The average trace value for a residue is computed with respect to the relative trace significance of it and the one of its neighboring residues: where is the set of residue positions which are neighbors of (that is, a neighbor is a residue with a distance <5Å from of at least one of its atoms), and where we fixed by default the weight values at and, favoring the residue compared to its neighbors. is the actual value that is used in JET to rank residues and to establish the importance of a residue position. Surface residues are residues with at least 5% of accessible surface [18]. Surface atoms have at least 1Å2 of accessible surface. Accessibility is calculated with NACCESS 2. 1. 1 [19] with a probe size of = 1. 4Å. In practice we shall use surface atoms belonging to surface residues only. A surface cluster is a set of surface residues to which a residue belongs if at least one of the surface atoms of is at distance <5Å from a surface atom in some other residue of the cluster. Several surface clusters can be detected for a single protein. Note that a surface cluster contains residues that are in contact because of surface atoms and excludes contacts based on internal atoms. As a consequence of this definition, clusters which are not contiguous patches at the protein surface are separated and, in some cases, several smaller surface clusters are obtained. See Figure 3. This definition reflects the idea that protein-protein interactions depend on atomic-level detail. An inverse relation between the fraction of the surface covered by the interface and the total protein surface has been observed in [20] based on a dataset of 1256 protein chains. We approximated the data in [20] with the function (plotted in Figure 4), where is the number of surface residues. We used this analytical expression to parameterize the clustering algorithm described below. Two thresholds are defined from the distribution of trace values computed with JET. The cluster-trace threshold is the trace determined with a confidence level of on the distribution of trace values and the residue-trace threshold is the trace determined with a confidence level of for the same distribution. These thresholds are used to construct and evaluate appropriate clusters. The clustering algorithm is structured in three steps. The first two steps are used to construct “cluster seeds” that will be extended into clusters at the third step of the construction. First, the algorithm orders all trace residues from the largest to the smallest. Next, it chooses residues with the highest trace value, greater than the residue-trace threshold, and either creates a new isolated cluster or adds the residue to an old cluster by checking that the average trace of the new cluster (either the isolated one or the one obtained by extension) is greater than the cluster-trace threshold. Notice that residue traces may be smaller than the cluster trace threshold. The set of clusters obtained in this way is filtered by the next step of the algorithm. In the second step, the algorithm computes a threshold for the size of the “cluster seeds”. To do so it takes the distribution of trace values obtained by running JET on a given protein and randomly reassigns the same trace distribution to surface residues of the protein. It clusters with the clustering algorithm described above and repeats this procedure 6000 times. It calculates the distribution of the size of the clusters and the distribution of the number of clusters obtained, to determine the percentile of a size, that is, the fraction of the population which has a size, and the percentile of the number of clusters, that is the fraction of the population with a number of clusters. Then it selects the clusters in (obtained in the first step) those within a percentile of size <0. 1. For all other clusters, it considers more relaxed conditions for selection. Namely, it selects clusters which are smaller in size, but have a high average trace compared to the others in. This notion is coded into the following two numerical conditions: where computes the percentile in a distribution and are set at 0. 15 and 1, respectively, for a first round of selection and to 0. 25 and 0. 95 for a second round of selection. If no cluster is selected, then the algorithm goes back to the random distribution, repeats the analysis by increasing the percentile level by 10% and recomputes a new, more lax, threshold until at least one cluster is found. The clusters obtained at the end of the second step of the clustering algorithm are called cluster seeds. The third step of the algorithm extends the cluster seeds with neighboring residues by maintaining a sufficiently high average trace of the cluster. To do this, the cluster-trace threshold is set at a confidence level of. Neighboring surface residues are those that respect the definition of a cluster once added. The algorithm collects all neighboring surface residues and adds them one by one by decreasing trace value, each time checking that the cluster-trace and the residue-trace thresholds are respected. When all neighboring residues are treated, the algorithm extends the resulting cluster further by searching for a new set of neighboring surface residues and by applying the extension procedure described before until no further extension is obtained. The algorithm then outputs the final set of clusters. The number of clusters may be smaller than the number of cluster seeds because extension may lead to the fusion of some initial clusters. Note that the residue-trace threshold guarantees that we are going to cluster a pool of residues among the best trace residues. The cluster-trace is used to guarantee that the average trace of the seed cluster remains high. Statistical analysis of physical-chemical properties of protein-protein interfaces reveals a biased amino-acid content within interfaces and allows the definition of propensity values for interface residues [21]. These values are listed in Text S1. We use propensity values to rank residues in a protein. For each residue we definewhere, is the degree of significance of, is the propensity value of. Notice that the formula is similar to and parameters, and are defined as for. We employ the ranking on for computing cluster seeds based on physical-chemical signals by running the first and second step of the clustering procedure (with a cluster-trace threshold determined by using the distribution of trace values dependent on). Then we compute cluster seeds based on conservation using the ranking of trace values (with a cluster-trace threshold computed from the trace distribution). Cluster seeds in the set are extended with the third step of the clustering procedure. For this we use a mixed trace value for a residue at position, instead of the usual trace value, which is defined asthat is, the average between trace and propensity. The cluster-trace threshold is computed from the distribution of mixed trace values. Note that cluster seeds detected by different signals can again fuse into a single cluster as discussed later for a allophycocyanin structure. Size of predicted clusters computed with mixed trace values and number of surface residues are reported in Figure 4 for all proteins in the Huang dataset. Points fluctuate around the curve (which represents reference values) and this is due to the multiple parameters used for clusterisation. The experimental interaction sites for the proteins listed in Text S2, Text S3 and Text S5 are determined using the crystal structure of the protein and NACCESS [19] for the detection of residues exposed to the solvent. JET finds signals corresponding to different interactions of a protein, namely with other proteins, ligands, DNA or RNA, as well as the chain-chain interactions in multimeric proteins. Hence, it becomes important to consider all it is known of such interactions to correctly evaluate predictions (see Figure 5). Given a protein, we considered all interactions between its chains. In addition, we collected information on other potential interactions by searching in the PDB archive for protein complexes containing a chain that displays at least 95% sequence identity to a chain in the PDB file of the experimental structure. All sites for the homologous chains (defined by an interaction with other chains in the “homologous” PDB file) are considered. For all PDB files (the reference and the homologous ones), we also looked at all chain-ligand interactions described in them, and selected those involving ligands that are known to have a functional role. For this, we used a list of enzyme compounds associated to reactions stored in KEGG database (a flatfile was downloaded at ftp: //ftp. genome. jp/pub/kegg/ligand/enzyme/enzyme) and discharged all compounds which were absent in the list. All identified interactions were grouped together to define the set of “true” interacting residues of the experimental structure to be evaluated. We define a residue to belong to an interaction site if at least 10% of the accessible surface of the residue (within the protein) becomes non accessible due to the interaction (within the complex). To properly evaluate JET performance on a given protein we rely on the following quantities: the number of residues correctly predicted as interacting (true positives, TP), the number of residues correctly predicted as non-interacting (true negatives, TN), the number of non-interacting residues incorrectly predicted as interacting (false positives, FP) and the number of interacting residues incorrectly predicted as non-interacting (false negatives, FN). We use four standard measures of performance: sensitivity, specificity accuracy and positive predictive value. We also consider scores to evaluate the statistical pertinence of the above measures. Expected values are calculated as, , , , where is the coverage obtained with JET, is the number of surface residues predicted by JET, is the total number of surface residues and is the number of residues in the real interaction site. Note that the calculation of expected values assumes that residues have been selected at random as being positives on the structure of the protein under study. This means that expected values depend on the protein studied. We can now compute sensitivity, specificity accuracy and positive predictive values for the random case: , , , respectively. Pertinence scores are computed as follows: sensitivity score, specificity score, accuracy score and PPV score. To compare JET performance and the ET analysis described in [22], we used the Matthews' correlation coefficient (MCC) [23] defined as where. When JET gives no answer (for example, due to an insufficient number of sequences retrieved by PSI-BLAST), then all interface residues are treated as negatives with, being the difference between the surface size and interface size (computed as the number of residues) and being the interface size. Notice that are all the negatives. All evaluation scores reported in the Tables, Text S2, Text S3, and Text S5 are multiplied by 100. The Huang dataset of 62 protein complexes constituted of 43 homodimeric chains, 24 heterodimeric chains and 19 transient chains [4] has been used to test JET performance and to compare it to ET (see below). The PDB code, chain and size of all proteins in the Huang dataset are listed in Text S2. Some of the chains appear in complexes of different types: heterodimers and homodimers include four combinations of the same chains, and homodimers and transients include two combinations. Several additional protein structures discussed in the text are listed in Text S3. All results reported in Tables, Text S1, Text S2, and Text S3 have been obtained with sequences retrieved from the PSI-BLAST server. To check JET behavior on interfaces belonging to different functional categories, we used the Kanamori dataset of 265 interfaces which contains 72 signal transduction proteins, 43 enzymes, 19 inhibitors, 36 antibodies, 31 antigens, 64 other proteins [22]. This dataset was originally constituted to evaluate the possibility to employ information on residue conservation coming from ET to direct docking. Structures of proteins and complexes used for the analysis were downloaded from Protein Data Bank http: //www. rcsb. org/pdb/home/home. do. ET predictions (that is, residue average trace values and clusters) have been obtained using locally ET Viewer (http: //mammoth. bcm. tmc. edu/traceview/). ET default values are: 500 BLAST retrieved sequences, a sequence identity between 26–98% for retrieved sequences, a cut-off of 0. 7 on the length of retrieved sequences, a maximum BLAST e-value at 0. 05, a coverage of 25% for clustering residues belonging to the whole protein (not only those lying on the surface). Note that for small proteins, the 25% protein coverage corresponds essentially to surface coverage, but that, in general, one should expect ET to cover much less protein surface. ET predictions were taken from [22]. iJET was run with default values and complexes interfaces were evaluated with NACCESS. Six chains (1cdk: I, 1cdm: B, 1i4o: C, 1jdp: H, 1nrn: R and 1vrk: B) in Kanamori dataset were too small (≤20aa) to be evaluated with iJET and in this case the evaluation of the complex considered TP = 0 and FN = 0 for these chains. JET has been implemented in Java and Java 3D. A list of all default values for JET parameters and instructions on how to use it is given in Text S4. The program can be found at http: //www. ihes. fr/̃carbone/data. htm. JET output files can be visualized with available programs like VMD, used to generate all figures of protein structures in this article [24].
Large-scale predictions of interaction sites from evolutionary signals are highly sensitive to the degree of variability within the available sequences. The Huang dataset contains a pool of proteins which, overall, turns out to be quite well-sampled by a PSI-BLAST search. This resulted in an average of 358,210,61 and 29 sequences for the 20–39%, 40–59%, 60–79%, 80–98% identity classes for the whole set of proteins. There are however a few exceptions which are worth discussing since a large-scale approach needs to handle such cases appropriately. Notably, an adjustment of the number of trees and number of sequences in a tree is important to ensure the most appropriate sequence sampling within the trees. Depending on the size of a protein we should expect that a different proportion of residues will belong to interfaces [20]. As discussed in Material and Methods, the clustering algorithm uses an estimation of the size of the expected interaction site as a function of the size of the protein. This estimation varies significantly for proteins of different sizes. Roughly, corresponds to an interface that covers <10% of the entire surface, to <15%, to <25% and to a fraction varying (rapidly) from 25% to 90%. It might seem that for small proteins, JET covers a large proportion of the surface, but this has advantages as illustrated by the 85aa long Mdm2 protein chain 1ycr bound to the transactivation domain of p53. JET predictions cover 46% of the surface and by doing so, detect 71% of the interaction site, that is 10 residues interacting with P53 out of 16 (see comparisons with ET below). One of the characteristics of JET is to use several distance trees of randomly sampled sequences instead of just one distance tree grouping all sequences recovered with PSI-BLAST. In Figure 6, we show the improvement in JET predictions solely due to dividing sequences amongst several trees. This is done by varying the number of trees, and by evaluating JET performance. (For each, we ensure that JET treats roughly the same quantity of sequence information by requiring each tree to contain sequences. Note that due to a random choice of sequences for each tree, there is a high probability that the trees will share some common sequences). Improvements come from a consensus in residue trace values as a consequence of the degree of significance of a trace. This is determined by the number of trees used in the prediction. The plot shows better predictions for larger number of trees and also that the methodology leads to decreasing the noise due to incorrect alignments, the presence of non-homologous sequences in the pool, biased samples and so on. Figure 7 illustrates the execution time of JET (excluding the PSI-BLAST step) when it is applied to the same pool of 400 sequences, but varying the number of trees. Sequences in the pool are all homologous to the sequence of chain 9atc: A. If trees are considered, each tree contains sequences which are randomly selected in the four identity classes as explained in Material and Methods. The plot shows that execution time is proportional to the number of sequences in the trees, with the major contribution coming from the CLUSTALW alignment. Given a protein structure, JET estimates the size of the largest surface cluster for the protein (obtained by taking the largest cluster computed over 6000 iterations of the random clustering procedure). Based on the number of estimated residues, it predicts an interaction site of the appropriate size. The need for a structure-specific estimation results from the absence of a correlation between protein size and size of the largest surface cluster as illustrated in Figure 8 (black dots) for the Huang dataset. On the contrary, there is a linear correlation between protein size and cluster size when all protein residues are considered (see Figure 8, (grey dots), where random clustering is carried out on all protein residues and not only surface residues). Based on this property, [25] proposed a linear correlation and used it to predict the largest acceptable protein cluster for a given protein. Our analysis shows that cluster size predictions based on structure-specific estimations are better. JET is a prediction tool that uses evolutionary information to detect conserved interaction sites and was inspired by the Evolutionary Trace approach. Comparisons with ET are therefore necessary. The performance of JET and ET on the Huang dataset are presented in Text S2 for each protein and a synthesis is provided in Table 2 (compare lines “ET” and “JET-cons”) for homodimer, heterodimer and transient interfaces. The two systems perform in a comparable way when clustering is not applied. After clustering, JET covers 38% () of the interface against 35% for ET. The interface residues predicted by JET correspond to real interface residues with a probability of 0. 6 () against 0. 5 for ET. JET prediction scores are two times better than random predictions (). JET found 86% of residues which are not in the interface () against the 84% for ET. The combination of these evaluating factors implies an average accuracy of 71% for JET against 68% for ET. Differences in ET and JET performance with and without clustering suggests that the clustering procedure employed in ET is less successful than that proposed here. In [10], it is argued that ET works best for families of homologous proteins with sequence identities higher than 40%. JET correctly detects functional sites of protein families well below this threshold. In Text S2, we provide the number of sequences retrieved with PSI-BLAST and the sequence identity classes for all proteins in the Huang dataset. For a large majority of these proteins, most retrieved sequences fall into the 20–39% class. For small proteins, the usage of an adapted curve (discussed above, see Figure 4) for evaluating protein coverage, also improves JET performance with respect to ET. An example is the Mdm2 protein chain 1ycr: A (discussed above, see also Figure 9D) where a 46% JET coverage contrasts with a 24. 6% ET coverage, and results in the detection of 71% of the interaction sites (10 residues out of 16 interacting with P53) against only 41% for ET (5 residues out of 16). To understand this contrast, it is important to look at the scores, , and listed in Text S3, which describe behavior of the two approaches compared to a random choice of residues. Note that in the case of 1ycr: A, most of the 25% residues covered by ET are surface residues, since the chain is small. Over the Huang dataset, a considerable improvement of JET performance is shown in Table 2 (compare lines “ET” and “JET - cons+pc”) when clustering is carried out on mixed traces, coupling both conservation signals and physical-chemical properties (, , and). In this way JET predictions improve considerably, as seen in the allophycocyanin structure 1all: B (Figure 10 and Text S3) where by using physical-chemical properties together with conservation, JET detects 66. 7% of the interaction site, while conservation alone only detects 51. 1%. ET detects 40% of the site. The leucine dehydrogenase structure 1leh in Figure 11, again illustrates that using physical-chemical properties improves the detection of interaction sites: the ligand site is constituted by very conserved residues, while the protein interface displays strong physical-chemical signals. The latter, combined with residue conservation, help JET to extract a suitable cluster describing the interaction site. ET fails to detect the site (see Text S3). Here, the PDB file used did not contain information on the ligand interaction and thus residues predicted to belong to the ligand interface were erroneously classed as false positives by our automatic procedure. This example illustrates the difficulty of a large-scale evaluation of a prediction system. To check whether clusters predicted in different runs of JET represent a consensus or not we iterated JET 10 times and analyzed its performance. Namely, given a protein, we considered a consensus prediction defined as the ensemble of residues that appear in a cluster at least times, for, out of the 10 iterations of JET on the protein structure. We then evaluated JET on each protein of the Huang dataset for increasing values of (see Figure 12). As expected, for increasing, predictions show a better PPV, but a worse sensitivity. This corresponds to an increased selectivity in choosing residues to belong to clusters. If conservation is coupled with physical-chemical properties, then specificity, accuracy and PPV curves show the best prediction at. The evaluation of JET iterated 7 times on the Huang dataset is presented in Table 2 (line “iJET”). The take-home message from this study is that different runs of JET are likely to provide slightly different outcomes and that a robust prediction of residues at the interface can be drawn from iterations. In this case, JET obtains very good average scores: , , , . Compare it with ET: , , , . JET is consistently better for homodimers, heterodimers and transients interfaces. It is important to stress that the iterative procedure suggests a list of residues that do not necessarily form clusters (as defined above), but patches of residues (and possibly isolated residues) that have been consistently (that is, in most JET runs) been classed as being part of an interaction interface. The iterated version of JET (based on 10 iterations) is called iJET. In Figure 13, we illustrate iJET behavior for different values of on several protein structures. Structures B, C, D show that for (column in the middle) we could detect residues belonging to the real interface that are missed for (right hand column). This means that in a single protein analysis, it could be worthwhile for the user to try different values of and evaluate the best ad hoc. For instance, for the structures of Figure 13, residues appearing <7 times (colored in pink) are not always interface residues (see A and B). For a large-scale analysis such tests are impossible and the value of needs be fixed. As we have shown, setting at 7 is appropriate for the Huang dataset. In Text S2 and Text S3, a comparison between iJET and single-run JET (using combined conservation signals and physical-chemical residue properties) shows that single-run JET can produce better results than the average obtained with iJET. The reason for this lies in the variable information content of pools of sequences retrieved by PSI-BLAST. This suggests that many of the sequences may be noisy in relation to the interaction site, although this noise can be eliminated in certain runs. Note that if JET is run on a single tree constructed out of sequences retrieved by PSI-BLAST, the result remains identical for all iterations. Computational strategies to ameliorate iJET will be discussed elsewhere. Small proteins are clearly more difficult to analyze than large ones. This is shown in Table 3 that revisits the performance of iJET presented in Table 2 with respect to protein length. Small proteins (with <200aa) display a less stable behavior compared to larger ones (≥200aa): evaluation scores for the two classes of large proteins in Table 2 are closer than for the two classes of small proteins. Specificity and accuracy remain essentially unchanged for large proteins and much lower values are attaint for small proteins. As expected, best sensibility and PPV are reached for small proteins due to a large coverage (see Figure 4). iJET has been compared to Consurf [26] and Rate4Site [27] on the Src SH2 domain of the 1fmk structure discussed in [26], [27]. iJET run 10 times on sequences which were automatically downloaded from the PSI-BLAST site, and where each residue trace value is the maximum trace value over the 10 runs. Consurf and Rate4Site run on 233 homologous sequences (Figures 2,3A, and 3B in [27]). The site between SH2 and the C-tail of the tyrosine kinase domain predicted by iJET is comparable to Consurf and Rate4Site predictions (compare Figures 2 and 3 in [27] and Figure 14, left). The three systems do not detect any residue in the SH2-kinase domain interface nor in the SH2-linker loop site. iJET detects as important (due to both conservation and physical-chemical properties) residue TRP148 sitting in the SH2–SH3 domain interface (Figure 14, right). Consurf detects no conserved residue, while Rate4Site identifies the site. By using 34 close SH2 homologues from the Src family [27], clear signals of conservation belonging to the multiple interaction sites are detected by the three systems. This is expected since the Src family is highly conserved. In this case, SH2–SH3 domain and SH2-kinase domain are well detected (see Figure 3 in [27] and Figure 15). The SH2-linker loop interface is detected as highly variable by Consurf while Rate4Site assigns to it an average conservation. iJET correctly detects the site even though it assigns to it a signal of average strength (see Figure 15, left). This is possible because of the cluster seed extension procedure in the clustering algorithm that does not require a residue to be conserved to belong to a cluster. It is important to see that no other residue located close to the conserved region is erroneously detected by iJET (see Figure 15, right). In conclusion, JET appears to perform better than Consurf and slightly less well than Rate4Site for the SH2–SH3 site. It demonstrate to be a successful platform for detecting very difficult signals like the linker loop interaction, where both Consurf and Rate4Site failed. Compared to Consurf, we can observe that it is able to detect important residues (such as TRP148) starting from a very mixed pool of sequences. It is interesting to notice that iJET and Rate4Site agree on the very variable residues, contrary to Consurf prediction of variability (see residues on top of the structure in Figure 14, left, and on bottom of the structure in Figure 14, right, and compare them to Figures 2 and 3 in [27]). iJET is compared to siteFiNDER|3D [28], Consurf and ET Viewer 2. 0 on the N-terminal domain of MukB (1qhl: A). iJET run on its own set of homologous proteins selected from its PSI-BLAST output, and important residues are defined to appear at least 8 times over the 10 JET runs. iJET pool of sequences gave rise to the expected prediction with 29 residues out of 227 residues in the chain, exhibiting a higher specificity than siteFiNDER|3D evaluated on its own dataset of sequences (45 over 227). The important residues determined by iJET are all clustered around the putative G-loop and include Gly34, Asn36, Gly37 and Lys40 from the Walker-A motif (Figure 16). This result shows the high specificity of iJET. Consurf run on its own set of sequences detects the Walker-A site but with a specificity of 37 out of 227 residues, therefore lower than iJET. ET Viewer 2. 0 run on its own dataset failed to make a useful prediction. As Consurf and ET Viewer 2. 0 (when these latter are applied to some well chosen dataset of sequences), JET detects as conserved other residues which lie in the same face of the molecule (like Glu202 and Tyr206), and this suggests a possible role in dimerization of MukB. We analyzed the structure of Arginine kinase (1bg0) discussed in [29]. We run iJET and we selected as important those residues that appear in JET clusters for 10 runs (Figure 17). Notice that this is a very restricting condition for selection. iJET detected as important (and conserved) and as belonging to the interaction site the functionally specific residues GLU225, ARG229, ARG280 and ARG309 [30] (it misses ARG126). These residues, as well as all others forming the interaction pocket, are not detected as conserved in [29], using SCORECONS [31] and the program alignment MUSCLE as input. Also, [29] detects only 2 over 5 residues as functionally important. This example confirms the results obtained at large scale on the Huang dataset, where we see that binding pockets are usually well detected. It shows JET accuracy in detecting conservation signals. JET is capable of detecting very different types of interface, as illustrated here with several case studies. Some belong to the Huang dataset (Text S2) and others are listed in Text S3. See Figures 13 and 9. Comparison with ET provides an evaluation of the power of JET in these cases. A large-scale analysis of interfaces with different functional classification is realized on the Kanamori dataset (Text S5) and follows. Even though JET detects several interaction sites and any evaluation is difficult, we compared it with the performance of ET on the Kanamori dataset of proteins organized in functional classes, where specific pairwise interactions were targeted. The overall evaluation scores attaint by iJET cannot be very good due a potentially erroneous increase of false positives coming from JET detection of multiple interaction sites, but an honest comparison of iJET to ET can be drawn on functional classes following [22]. Namely, we considered 265 protein interfaces belonging to different functional classes: signal transduction proteins, enzymes, inhibitors, antibodies, antigens and others [22], and considered as positives, the residues in the two interacting chains that belong to the interface. We found that iJET performs well in signal transduction proteins, enzymes and inhibitors, while a poor behavior is recorded on antigen and antibody interface predictions (see Table 4, Figure 18, and Text S5). We observe an improvement with iJET compared to ET [22]. The striking difference between our analysis and [22] is that for us inhibitors work essentially as well as enzymes. The MCC computed by [22] on the inhibitors class is −0. 01 (with a standard deviation of 0. 14) while we obtain a MCC of 0. 26 (with a standard deviation of 0. 11) which is comparable to the MCC of 0. 28 (and standard deviation of 0. 13) obtained for enzymes. Similar prediction quality for enzymes and inhibitors is not explainable by similar evolutionary pressure of enzyme-inhibitor partners since the two protein classes display asymmetric residue conservation [22], [32]. (See remarks above on receptor/inhibitor pockets.) iJET good performance on the inhibitors class might be due to the fact that iJET takes into account also physical-chemical properties for residue evaluation, and that it detects interaction sites accordingly to the biological hypothesis that clusters are formed by a conserved internal core surrounded by successively less conserved layers of residues. A careful analysis of the distribution of conserved residues on the inhibitor interaction sites should be able to clear out this point, but this will be done somewhere else. In conclusion, our finding support the crossed usage of iJET predictions with docking algorithms, leading to a reduction of the docking search space for signal transduction proteins, enzymes and inhibitors.
Conserved patches of residues on a protein surface can help to suggest the location of an interaction site. We have tested the evolutionary hypothesis that interaction sites are constituted by a conserved internal core, surrounded by successively less conserved layers of residues. Based on this hypothesis we were able to develop a new criterion for extending conserved patches (that is “cluster seeds”), improving predictions of realistic interface clusters. The impact of this extension step in the algorithm is nicely illustrated in Figure 15 where a residue belonging to the linker loop interface of SH2 is detected by the extension and remains non predicted by other systems. By making multiple iterations, iJET predicts 40% of the interfaces for proteins within the Huang dataset (with, and), and more than 50% of the interfaces for proteins in Text S3 (with, and). For more than a quarter of the proteins in Huang dataset, more than 50% of their sites are correctly predicted, and for 6 out of the 12 proteins in Text S3,60% of the true site is identified by iJET. We tested the evolutionary hypothesis that specific physical-chemical properties of residues forming interaction sites should co-exist with signals of residue conservation. We were able to show that a combination of conservation signals (even if low) and physico-chemical interface propensity values indeed leads to successful predictions. Future developments of JET will include an intelligent detection of patches satisfying specific physical-chemical properties based on propensity values differentiating multiple types of interaction [6]. JET and iJET can be used for large-scale analysis or as platforms to make in silico experiments on protein interfaces. These latter are possible due to the flexible parameterization provided by the system. Each step of JET can be monitored and improved by an accurate ad hoc understanding of the protein under study (this might end up into an explicit consideration of protein length, availability of homologous sequences, distribution of homologous sequences in sequence identity classes, expected conservation, etc.). The first hand information coming from a run of JET are the clusters that it provides. Notice that for large-scale comparison of JET and ET, we considered a hit to be the set of clusters issued by a single run of JET. For comparison with iJET, we considered a hit to be the set of clustered residues issued by iJET, with (pertinency of is discussed above). In single protein analysis, we might want to look for functionally specific residues, and it might be more appropriate to adopt very selective conditions, for instance by asking for a residue to appear in 10/10 clusters. If the aim is to discriminate between residue importance, it might be useful to use the maximal mixed trace for residues issued over 10 runs of JET, or as before, to select as important those residues appearing in 10/10 clusters. These measures are easily accessible to the user in the output files. Examples of the application of these criteria to single proteins are discussed in Results. Multiple interaction sites often occur on a protein surface and this makes evaluating JET difficult since only some of these sites may be experimentally characterized. JET is nevertheless capable of detecting all residues patches which are susceptible to be involved in interactions with other ligands or macromolecules. An example illustrating this point is leucine dehydrogenase 1leh (see Figure 11) Which has both a protein-protein interface and a ligand-binding pocket. The absence of information on the conserved pocket in the corresponding PDB file leads to apparent false positives when JET is used automatically (see Text S3), but such information can be valuable and can be used by the biologist in formulating new hypotheses. Lastly, it is remarked that JET can be applied to protein sequences for which the structure is unknown, if the structure of a homologous protein is available. This approach can again be valuable to the biologist, notably in guiding site-specific mutagenesis experiments [33]. | Information obtained on the structure of macromolecular complexes is important for identifying functionally important partners but also for determining how such interactions will be perturbed by natural or engineered site mutations. Hence, to fully understand or control biological processes we need to predict in the most accurate manner protein interfaces for a protein structure, possibly without knowing its partners. Joint Evolutionary Trees (JET) is a method designed to detect very different types of interactions of a protein with another protein, ligands, DNA, and RNA. It uses a carefully designed sampling method, making sequence analysis more sensitive to the functional and structural importance of individual residues, and a clustering method parametrized on the target structure for the detection of patches on protein surfaces and their extension into predicted interaction sites. JET is a large-scale method, highly accurate and potentially applicable to search for protein partners. | Abstract
Introduction
Materials and Methods
Results
Discussion | computational biology/macromolecular structure analysis
molecular biology/molecular evolution
computational biology/evolutionary modeling
molecular biology/bioinformatics
computational biology/macromolecular sequence analysis
evolutionary biology/bioinformatics | 2009 | Joint Evolutionary Trees: A Large-Scale Method To Predict Protein
Interfaces Based on Sequence Sampling | 12,231 | 194 |
Human papillomaviruses are causally associated with 5% of human cancers. The recent discovery of a papillomavirus (MmuPV1) that infects laboratory mice provides unique opportunities to study the life cycle and pathogenesis of papillomaviruses in the context of a genetically manipulatable host organism. To date, MmuPV1-induced disease has been found largely to be restricted to severely immunodeficient strains of mice. In this study, we report that ultraviolet radiation (UVR), specifically UVB spectra, causes wild-type strains of mice to become highly susceptible to MmuPV1-induced disease. MmuPV1-infected mice treated with UVB develop warts that progress to squamous cell carcinoma. Our studies further indicate that UVB induces systemic immunosuppression in mice that correlates with susceptibility to MmuPV1-associated disease. These findings provide new insight into how MmuPV1 can be used to study the life cycle of papillomaviruses and their role in carcinogenesis, the role of host immunity in controlling papillomavirus-associated pathogenesis, and a basis for understanding in part the role of UVR in promoting HPV infection in humans.
Papillomaviruses are species-specific, epitheliotropic, double-stranded DNA viruses. There are over 200 strains or genotypes of human papillomaviruses (HPVs) [1]. Mucosotropic HPVs are the most common sexually transmitted pathogens, and a subset of these viruses cause 5% of human cancers, including cervical cancer, other anogenital cancers, and a growing fraction of head and neck cancers (reviewed in [2]). Other HPVs cause cutaneous warts, which are among the most common ailments treated by dermatologists [3–6]. They arise most frequently among children [7,8], and impose a significant burden in immunocompromised patients, particularly amongst organ transplant recipients [9–11]. They are ubiquitous in nature and can persist in the skin asymptomatically for years most clearly in context of immunosuppressed patients [9,12]. A subset of cutaneous HPVs also has been causally associated with skin cancer (reviewed in [9,13,14]). The study of papillomavirus-induced disease has long been hindered by the absence of any identified strains of virus that infect laboratory mice. This limitation was overcome with the recent identification of the murine papillomavirus, MmuPV1, isolated from cutaneous warts arising on the T-cell deficient NMRI-FoxN1nu/nu strain of laboratory mice [15]. MmuPV1 belongs to the pi-papillomaviridae genus and is phylogenetically related to cutaneous HPVs and other animal PVs that cause cutaneous disease in exotic rodent species [16]. MmuPV1 causes warts in cutaneous epithelium as well as in mucosal epithelium lining the female reproductive tract and oral cavity [15,17–20], and in some cases these lesions show signs of neoplastic progression [20,21]. Multiple studies have shown that the ability of MmuPV1 to cause overt disease is largely restricted to immunodeficient strains of mice [18,20]. Epidemiological studies have suggested that there is a correlation between exposure to ultraviolet radiation (UVR) and the prevalence of cutaneous HPVs in healthy and immunosuppressed patients, respectively [22,23]. Cutaneous HPVs are more commonly found at anatomical sites exposed to sunlight, and a history of blistering sunburn is associated with prevalent and persistent cutaneous HPV infections [12,22–24]. UVR has also been shown to play a role in papillomavirus-associated disease caused by animal papillomaviruses that infect the African multimammate rat (MnPV) or the cottontail rabbit (CRPV) [25–27]. In this report we demonstrate that UVR, specifically UVB (280 to 315 nm) assists in development of MmuPV1 dependent papillomas and associated malignant progression to squamous cell carcinomas, in immunocompetent strains of mice. We further show that there is a correlation between UVR-induced susceptibility to MmuPV-1 associated disease and UVR-induced immunosuppression. These findings provide a potential explanation for the role of UVR-mediated immunosuppression in papillomavirus-associated disease in humans.
Several lines of evidence support a link between UVR exposure and papillomavirus infection [22,23,26]. We sought to determine if UVR facilitates MmuPV1-induced papillomatosis in immunocompetent strains of mice. We used the inbred FVB/NJ strain of mice for our initial studies because it has been classically used to study chemically induced skin tumorigenesis [28]. Ears and tails of 8–10 weeks old FVB/NJ mice were infected with 108 viral genome equivalents (VGE, a measure of the amount of encapsidated genomes in a stock of virus) of MmuPV1 virions following topical scarification of the epidermis. Twenty-four hours post-infection (h. p. i.), mice were or were not exposed to varying doses of UVB whole body irradiation (280-320nm range UVR). Mice were then observed weekly for papillomatosis (Fig 1A). By 3 months post-infection greater than 50% of infected ear sites developed papillomas in FVB/NJ mice treated with 300mJ/cm2 UVB (Fig 1B, Table 1, S1 Fig-Panel A). Sites on the tails of the same mice infected with an equal dose of virus failed to develop papillomas. We did not see any papillomas develop at infected sites on non-irradiated mice or mice treated with a lower dose (150mJ/cm2) of UVB (Table 1). Mock-infected mice exposed to the same doses of UVB did not develop warts (Table 1). Lateral transmission of MmuPV1 has been observed in immunodeficient strains of mice experimentally infected with MmuPV1 [15,20]; however, we did not see any warts arise at uninfected sites on the UVB-treated, MmuPV1-infected FVB/NJ mice. The presence and size of warts arising on the MmuPV1-infected ears sites of UVB-treated FVB/NJ mice were monitored for a total of 6 months. In this time frame 17% of the sites developed papillomas that completely regressed (Fig 1D and 1E: red lines in Fig 1E). 25% of sites developed papillomas that partially regressed (Fig 1D and 1E-blue lines). Another 17% of sites developed papillomas that continued to grow over the 6 month period (Fig 1D and 1E—green lines). 41% of the sites did not develop papillomas. To address whether UVB is exerting its effect directly on the virus, we tested whether UVB treatment at 300mJ/cm2 24 hours before infection (h. b. i.) had the same impact on papilloma incidence. Over the initial 4-month period post-infection, papillomas arose with similar frequency as in groups of mice exposed to UVB 24 hours after infection (Fig 1B, P = 0. 814, log-rank, two-sided). Again, we saw complete regression of a similar fraction of papillomas when mice were observed up to 6 months post-infection (Fig 1C). We also found that a subset of animals infected with MmuPV1 and treated with a UVB fourteen days post-infection also developed papillomas at a similar frequency (S1 Table). To learn if the UVA spectra (315-400nm range UVR) also caused immunocompetent mice to become susceptible to MmuPV1-induced papillomatosis, we infected FVB/NJ mice with 108 VGE of MmuPV1 at sites on the ears and exposed them to 300 J/cm2 of UVA 24 hours post-infection. We failed to observe papillomas arise at any sites on these mice (S2 Table). Combining UVA (300 J/cm2) and UVB (300 mJ/cm2) treatment gave a similar incidence of papillomas as seen with UVB alone (S2 Table). We also tested the susceptibility of other commonly used, wild-type, inbred strains of mice to MmuPV1-induced disease with varying doses of UVB (S3 Table): 30% of infected ear sites on C57/BL6 mice developed papillomas at the three-month time point when treated with UVB at a higher dose of 600mJ/cm2. Only one infected site on a BALB/c mouse developed a papilloma when exposed to 1200mJ/cm2 UVB (S3 Table). These results indicate that there are strain differences in susceptibility of mice to MmuPV1-induced disease following UVB irradiation. Because UVB-treated FVB/NJ mice were most susceptible to MmuPV1 disease (p = 0. 041 FVB/NJ vs C57/Bl6 or 0. 0044 FVB/NJ vs BalbC, at 300 mJ/cm2), we pursued all further studies of the role of UVB in inducing MmuPV1 infection using the FVB/NJ strain of mice. To investigate the relationship between viral dose and papilloma incidence in FVB/NJ mice treated with UVB, we infected mice on their ears with 108,107 or 106 VGE of MmuPV1 and exposed the mice to 300mJ/cm2 of UVB 24 hours post-infection (Fig 2). In parallel, T-cell deficient FoxN1nu/nu mice were also infected at ear sites with the same stock of virus at designated doses in the absence of UVB-treatment (Fig 2). By 3 months post-infection, 90% of ear sites in FoxN1nu/nu mice infected with 108 and 107 VGE MmuPV1 developed papillomas (Fig 2A and 2B –Right). At this time-point, papillomas developed in UVB-irradiated FVB/NJ mice infected with 108 and 107 VGE MmuPV1, but at lower penetrance (Fig 2A and 2B—Left) and no papillomas formed on UVB-treated FVB/NJ mice infected with 106 VGE of MmuPV1 up to 6 months post-infection (Fig 2B- Right). In contrast, 37. 5% of FoxN1nu/nu mice ear sites infected with 106 VGE of MmuPV1 develop papillomas (Fig 2B-Left). Non-irradiated FVB/NJ mice infected with 108,107 or 106 VGE of MmuPV1 did not develop any papillomas. These results indicate that a higher threshold in the amount of virus is required to see papilloma induction in immunocompetent animals after UVB irradiation. Furthermore, there was complete regression of a subset of papillomas in FVB/NJ mice infected with 107 and 108 VGE of MmuPV1 when monitored up to 6 months post-infection (Fig 2B). There was no regression of papillomas in immunodeficient FoxN1nu/nu mice. Lesions arising on the ear sites of MmuPV1-infected FVB mice treated with UVB were harvested at 6 months post-infection and analyzed histopathologically. They showed sessile papilloma-like morphology with hyperkeratosis (Fig 3A, left; S1 Fig-Panel B). Cells within the stratum granulosum showed presence of koilocytes consistent with productive papillomavirus infection (Fig 3A, middle). The MmuPV1 major viral capsid protein was expressed in these papillomas as indicated by L1-specific immunofluorescence (Fig 3A, right). Viral capsid protein was mostly detected in the koilocytes and other cells within the upper layers of the epithelia (suprabasal and terminally differentiated) with some L1-positive cells occasionally observed in the basal layer consistent with prior findings in immunodeficient mice [19]. The sessile papillomas were accompanied by multiple areas of atypical squamous cell hyperplasia, several of which were suggestive of early neoplastic transformation, consistent with other reports [17,19,20]. In our case, however, we also found focal areas of malignant progression consistent with squamous cell carcinoma with invasion extending into follicular structures deep within the dermis as evident from cytokeratin-14 staining (Fig 3B). These areas of malignant progression were observed in several histopathologically scored lesions (e. g. see S1 Fig- Panel B). Foci of chronic inflammation were also noted in the lesions, some of which extended into the atypical dermis (Fig 3B). Southern blot quantification showed that viral extracts harvested from papillomas from immunocompetent mice had a 1000-fold reduction in the amount of VGE/mg of wart harvested when compared to the papillomas arising in immunocompromised FoxN1nu mice (S2 Fig). Regardless, the viral extracts from immunocompetent mice were infectious and caused papillomas in FoxN1nu mice (Fig 3B). MmuPV1 DNA-specific in situ hybridization coupled with L1 immunohistochemistry analysis of MmuPV1-induced ear papillomas in UVB irradiated FVB/NJ mice (Fig 3C) showed presence of amplified viral DNA and L1 capsid. Robust L1 expression was seen throughout the papilloma most frequently in the suprabasal layers of the epithelia. The FISH positive cells were comparatively less frequent and predominantly seen in the spinous epithelium occasionally showing co-localization with L1 positive nuclei. We observed that while regions of papillomatosis showed presence of amplified MmuPV1 DNA areas of malignancy showed little to no presence of MmuPV1 amplified DNA (S3 Fig). Presence of the viral major capsid protein (L1) and amplified viral DNA, which are markers for the productive stage of the viral life cycle [reviewed in [29]], coupled with findings from the transmission experiments (Fig 3B), confirm that MmuPV1 establishes a productive infection in immunocompetent mice following UVB treatment. There are several lines of evidence in the field of photoimmunology indicating that UVB impairs a variety of immune responses in humans and laboratory animals both locally, within UV-irradiated skin, and systemically, at distant sites [30–32]. To test whether UVB-assisted pathogenesis in MmuPV1-infected FVB/NJ mice is due to a systemic or a local effect of UVB-irradiation, we infected FVB/NJ mice at ear sites as described previously. Twenty-four hours after infection, mice were anesthetized and the infected ear sites were shielded with tin foil leaving the rest of the mouse exposed to 300mJ/cm2 UVB irradiation (Fig 4A). Mice were then observed weekly to score for papillomatosis up to six months post-infection. At the end of 16 weeks approximately 55% of infected mice developed papillomas on ear sites (Fig 4B). There was no significant difference in the temporal onset of papillomas between shielded and unshielded mice (P = 0. 938, log-rank, two sided). The animals were kept under observation for 6 months and scored weekly for papilloma incidence. We found that by six months post-infection 27% of the papillomas completely regressed in the mouse cohort whose infection sites were shielded, similar to that seen in the unshielded cohort, where 28. 5% of the papillomas had completely regressed. These results indicate that systemic effects of UVB on host biology must contribute to the ability of UVB to induce MmuPV1-dependent disease. Through the study of different strains of immunodeficient mice, it has been found that T-cell deficiency is necessary for the development of MmuPV1-dependent papillomas [18,20]. Several studies have demonstrated that UVB causes cell-mediated immunosuppression in mice [33,34], reviewed in [30,32,35,36]. Cell-mediated immunosuppression by UVB has traditionally been measured by monitoring delayed type hypersensitivity (DTH) responses [30,32,35,37,38]. A single exposure to UVB irradiation is sufficient to inhibit DTH responses in some strains of mice [39,40]. To determine if the single dose of UVB irradiation (300mJ/cm2) that makes FVB/NJ mice susceptible to MmuPV1-induced papillomatosis (Table 1) is sufficient to cause immunosuppression in this strain of mice, we measured DTH responses in these and control mice not treated with UVB [41,42]. Ten days post-UVB treatment, FVB/NJ mice were sensitized to an antigen by topically applying 0. 5 mg 1-Chloro-2,4-Di-Nitrobenzene (DNCB) on the shaved backs of the mice. Five days later we determined the level of immune response to this antigen by applying 0. 2 mg DNCB to a distant site (the ears) and monitoring DTH responses (Fig 5). DTH was assessed by measuring ear swelling every 24 hours for four days. A single exposure to UVB (300mJ/cm2) was capable of causing immunosuppression in FVB/NJ mice as evidenced by a marked decrease in swelling in response to DNCB challenge in the UVB treated mice compared to non-UVB treated mice (Fig 5). This difference was statistically significant (p<0. 005, T-test). To assess whether this immunosuppression was systemic or local, we repeated the DTH assay on mice in which we shielded their ears from UVB as described previously. These mice also displayed reduced swelling in response to challenge with DNCB on their ears, indicating that UVB-induced immunosuppression in these mice is systemic (Fig 5). This observation suggests that UVB causes systemic immunosuppression that may assist MmuPV1-dependent disease (Figs 1–4). To determine if UVB-mediated immune suppression correlates with papilloma incidence we performed long-term DTH assays in FVB/NJ mice following infection with MmuPV1 and UVB-treatment. Mice were infected with MmuPV1 in their right ear and exposed to a single dose of UVB (300 mJ/cm2) 24 hrs post-infection as described previously (Fig 1A). Ten days post UVB-treatment animals were sensitized to antigen by painting 0. 5 mg of DNCB on their shaved backs. Mice were monitored for papilloma formation. Three months post infection, animals were challenged with 0. 2 mg DNCB on the uninfected, left ear. We chose the three-month time period because this is the time by which FVB/NJ mice treated with UVB develop the maximal number of MmuPV1-induced warts (Figs 1C and 2B). Post-challenge we measured ear swelling up to 96 hrs (Fig 6). We found that in the control group (animals not treated with UVB) a long-lived DTH response was established as evidenced by swelling of the ears. UVB-treated, uninfected control mice displayed a range in levels of DTH response, indicative of variable levels of long-term immune suppression. This was also seen in UVB-treated, MmuPV1-infected mice. Of specific note, the mice in this latter group that retained immune-suppression at the end of 3 months were the same ones that had developed MmuPV1-induced warts. The difference in ear swelling between the UVB-irradiated animals infected with MmuPV1 that developed warts and those that did not develop warts was statistically significant (P = 0. 016, Wilcoxon rank-sum, two-sided). These data demonstrate a strong correlation between long-term, UVB-induced immunosuppression and MmuPV1-dependent pathogenesis.
The lack of infection models in a tractable laboratory animal has limited our ability to study the pathogenesis of papillomaviruses in their natural hosts. The murine papillomavirus, MmuPV1, isolated from cutaneous warts arising on immunodeficient NCR-FoxN1nu/nu laboratory mice [15,16] is a valuable animal papillomavirus because it provides us, for the first time, the opportunity to study papillomavirus infections in the context of a genetically manipulatable host. Prior to this study, MmuPV1-associated papillomatosis has been described primarily in the context of immunodeficient strains of mice [17,18,20,21], though there have been reports that MmuPV1 can cause warts on hairless strains of mice, which are thought to be immunocompetent [15,43]. Consistent with a role of the host immune system in limiting MmuPV1-induced disease, Handisurya, et al. (2014) found that Cyclosporin A treatment is required for induction and maintenance of MmuPV1-induced papillomas in immunocompetent mice [18]. These authors did find that, at very high doses of virus (1012 VGEs), the SENCAR strain of mice, selectively bred for high susceptibility to skin tumor induction by chemical carcinogens [44], developed papillomas, but these papillomas regressed within two weeks of appearing. In this study we investigated whether the UVB spectra impacts susceptibility of immunocompetent mice to MmuPV1. There was limited prior evidence suggesting the role of UV radiation in other animal models for papillomavirus infection. Mastomys natalensis Papillomavirus (MnPV) is a rodent papillomavirus that is shown to cause papillomatosis in the African multimammate rat [25]. MnPV DNA was found in UV induced tumors in HRA/Skh mice that are hairless but immunocompetent [26]. Further, studies showed that cell-free extracts containing MnPV enhanced UV-induced tumorigenesis [45]. We found that when FVB/NJ mice were exposed to high doses of UVB (Table 1) greater than 50% of infected sites developed papillomas by 3 months, some of which persisted for 6 months (Figs 1–4). UVB also induced MmuPV1-dependent papillomatosis in other strains of immunocompetent mice (S3 Table). Histopathological analysis of the papillomas in the MmuPV1/UVB infection model also indicated progression to squamous cell carcinoma (Fig 3, S1 Fig-Panel B). Interestingly, we observed that while regions of papillomatosis showed many cells that had amplified viral DNA, we only observed a few cells harboring amplified viral DNA in papilloma-associated malignant regions (S3 Fig). More sensitive in situ hybridization techniques will be required to determine whether MmuPV1 viral genomes are lost as malignant progression arises, as is thought to occur in beta-HPV associated non-melanoma skin cancers [46]. Several lines of evidence supports the hypothesis that UVB is having an indirect effect on increasing the susceptibility to MmuPV-1 induced papillomatosis by inducing systemic immunosuppression. First, we did not find any significant difference in susceptibility to wart formation in mice that were treated with UVB 24 hours before or 24 hours after infection with MmuPV1 (Fig 1). Second, UVB' s ability to increase susceptibility to MmuPV1-induced papillomatosis was equally efficient whether or not the site of infection was exposed to UVB (Fig 4). And third, long-lived immunosuppression correlated with wart formation (Fig 6). A second but not mutually exclusive hypothesis is that UVB also directly influences MmuPV1-induced pathogenesis. Supportive of this hypothesis, expression of the viral genes in K14HPV8 transgenic mice were increased following exposure to UVR [47]. The physiological relevance of this observation remains unclear, however, because the HPV8 genes are under the control of a heterologous, keratin 14 promoter in the context of this transgenic mouse model. The MmuPV1 infection model should provide a valuable experimental platform for further testing this second hypothesis. The ability of papillomaviruses to persist in their host in the absence of causing overt disease, i. e. latency, has long been suspected. Compelling evidence for latency comes from the observation that cyclosporine-induced immunosuppression led to elevation of the viral DNA copy number at sites of wart regression in cottontail rabbits infected with rabbit oral papillomavirus, consistent with reactivation of virus from latency [48]. In the MmuPV1-infection model, ELISA analysis of serum from C57BL/6 mice infected with MmuPV1 showed seroreactivity to MmuPV1 virus particles at 70 days post-infection even though they did not develop warts [49]. Likewise, seroconversion was seen in all mice (n = 20) in a group of SKH-1 mice infected with MmuPV1, of which 3 mice actually developed papillomas [43]. Together, these observations indicate that the virus is presented to the immune system even in the absence of causing overt disease. Whether this results from the original exposure of the animals to the virus or the consequence of a latent infection remains unclear. Our own observation that a single exposure to UVB 14 days post-infection led to papillomatosis at sites infected with MmuPV1 (S1 Table) indicates either that infectious virus is stably retained at the site of infection for that period of time and then initiates infection post UVB exposure or that latent infections arose that were then activated by UVB-induced immunosuppression. It remains to be determined whether latency arises in the MmuPV1 infection model, and, if so, the nature of this latency. In this study we observed an unexpected tissue specificity in terms of susceptibility to MmuPV1-induced papillomatosis; while infection of the ear led to efficient formation of papillomas in UVB-treated FVB/NJ mice, papillomas did not arise at tail sites that were infected in the same mice. Others have reported that there is some site specificity for MmuPV1-induced disease in immunodeficient mice [19,20]. For example, while tail and muzzle of immunodeficient mice are susceptible to MmuPV1-induced papillomatosis, the torso skin is not. There are limited studies that have been directed towards looking at difference in susceptibility of different cutaneous sites of mice to cancer. Therefore the molecular differences between the sites are not very clear. In our studies with HPV 16 transgenic (K14E6, K14E7) mice we have consistently observed epithelial hyperplasia most extensively in the ear skin [50–52]. Likewise, ear skin of HPV 38 E6/E7 transgenic mice display patches of hyperproliferation [53]. Recently, there has been one report that suggests that difference in miR-155 expression of ear versus chest/torso skin of K14HPV16 mice could explain the difference in susceptibility of different tissues to develop HPV-mediated carcinogenesis [54]. It is also possible that grooming behaviors of mice, such as scratching of ear, promotes wounding [55] thereby allowing virus to access lower layers of epithelia better. In our studies we observed a clear viral-dose dependence in causing disease. Interestingly, we found that a viral dose of 107 VGE or greater was required to see papillomas in UVB-irradiated FVB/NJ mice whereas 106 VGE of MmuPV1 were sufficient to induce papillomas on BALB/c-Foxn1nu/nu mice, albeit at low incidence over the same 6-month observation period (Fig 2). This suggests that there is a higher threshold of virus required to see MmuPV1-dependent disease in UVB treated immunocompetent animals. We posit that this observation reflects, at least in part, the fact that long-term immunosuppression induced by UVB in FVB/NJ mice only occurs in a subset of the UVB-treated mice (Fig 6). In our testing of different genetic backgrounds, only FVB/NJ was susceptible to MmuPV1-associated pathogenesis at 300mJ/cm2, whereas C57/BL6 and BALB/c mice were susceptible to MmuPV1-associated pathogenesis at higher doses of UVB (S3 Table). This observation is consistent with a previous study that suggests that different genetic backgrounds respond differently to UVB in terms of levels of immunosuppression based upon DTH assays [56]. While that study did not test FVB/NJ mice, they found that C57/BL6 mice were more sensitive to UVB-induced immunosuppression than BALB/c mice. This correlates with the level of MmuPV1-dependent papillomatosis observed in our studies (S3 Table). In our study we observed that UVB, but not UVA alone makes immunocompetent mice susceptible to MmuPV1-dependent papillomatosis (S2 Table). UVB has been shown to play a key role in initiating and mediating immunosuppression, whereas the mechanisms and roles of UVA in immunosuppression are not well understood [57,58]. UVB can cause both short-term as well as long-term defects in cell-mediated DTH responses [35,38,40], at least in part by inhibiting development of memory T-cells and by causing an overall reduction in T-cell subpopulations in the skin [59]. We observed that UVB caused variable levels of long-term immunosuppression in FVB/NJ animals and that MmuPV1-induced papillomas developed preferentially in those animals retaining long-term immunosuppression (Fig 6). This can explain the observation that papillomas arise in approximately 50% of animals. Based upon these findings we also posit that the observed regression of papillomas (Figs 1–2) correlates with a loss of immunosuppression. The observation that there is a strong correlation between long-term immunosuppression induced by UVB and MmuPV1-dependent pathogenesis is perhaps the most notable finding of this study. This correlation supports the hypothesis that UVB-induced immunosuppression can help drive papillomavirus-induced disease. There is correlative epidemiological data from human studies in which anatomical sites on individuals that are exposed to sunlight, or at which sunburn has occurred are increased in their susceptibility to HPV-induced warts. One difference however, is that in our studies with FVB/NJ mice, the effect of UVB was found to be systemic; UVB irradiation did not have to be applied to the infection site. This raises the interesting question: is there a difference in the role of UVB in HPV-driven pathogenesis compared to MmupV1-driven pathogenesis? Further studies are needed to assess whether a local effect of UVB on MmuPV1 can be identified in mice. Recently, it has been shown that MmuPV1 also infects the mucosal epithelium of the female reproductive tract and oral cavity [21]. In this regards MmuPV1 infection model is truly unique and the systemic immunosuppression by UVB can further be tested in the context of mucosal disease. In conclusion, we have reported the novel finding that UVR makes immunocompetent mice susceptible to development of MmuPV1-induced cutaneous papillomas, and this correlates with UVB-induced systemic immunosuppression. This observation opens the door to pursuing studies using genetically engineered mice to study molecular pathways that mediate the role of UVB in making mice susceptible to papillomavirus-induced pathogenesis, as well as identifying cellular targets of MmuPV-1 encoded factors that mediate their role in pathogenesis.
The following mice were obtained and bred for the purpose of this study (vendor in parenthesis): immunocompetent FVB/NJ (Taconic), BALB/c (Charles River), C57BL/6 (Jacksons Lab); immunodeficient athymic BALB/c FoxN1nu/nu (Harlan). All infected mice were housed in aseptic conditions in micro-isolator cages. Animals were handled only by designated personnel and personal protection gear was changed between cages to prevent any cross contamination from virus. Mice were housed at McArdle Laboratory Animal Care Unit in strict accordance with guidelines approved by the Association for Assessment of Laboratory Animal Care, at the University of Wisconsin Medical School. All protocols for animal work were approved by the University of Wisconsin Medical School Institutional Animal Care and Use Committee (Protocol number: M02478). MmuPV1 virus stock was generated by isolating MmuPV1 virions from papillomas in nude mice as described previously [18,19,60]. The MmuPV1 infection model was established in nude mice using quasivirions generated (as described previously [19,61]) using a clone of MmuPV1 obtained in the Lambert lab. Briefly, this clone of MmuPV1 was made by performing rolling circle amplification on the virus extracts (generously provided by Dr. Aravind Ingle, ACTREC, India) from the original colony of nude mice infected with MmuPV1 [15] followed by cloning into the pUC19 vector [15,16]. To encapsidate MmuPV1 genome we used pMusSHELL, a Mammalian expression plasmid with codon modified L1 and L2 genes of MmuPV1 (generously provided by Dr. Chris Buck, NIH) [60]. To confirm that the virus stock was infectious, nude mice were infected in parallel during each experiment as positive controls. This protocol has been modified from previously described methods of isolating non-enveloped viruses from human skin or tissue samples [62,63]. Animals with warts were euthanized and 10 mg of excised wart was homogenized in 700 μl PBS containing Triton-X-100 (1%). Benzonase was added to the homogenized wart sample and incubated at 37°C for 30 minutes. Collagenase H (2 mg) was added, and the sample was vortexed and then incubated at 4°C overnight. Sodium chloride concentration was adjusted to 0. 8M, the sample was centrifuged for 5 minutes at 5000g, and the supernatant was clarified by ultracentrifugation through an Optiprep (iodixanol) step gradient followed by fractionation as described in detail on the following website: http: //home. ccr. cancer. gov/LCO/pseudovirusproduction. htm. Viral genome equivalence was estimated by comparing the amount of encapsidated viral DNA in the viral stock, liberated by treatment with proteinase K, to known standards of cloned MmuPV-1 genome by Southern analysis using MmuPV1-specific probes, followed by quantification using ImageJ software (S1 Text). In vivo infection with purified MmuPV1 virions was performed on scarified skin of the animals' ears. Animals were anesthetized and tails or inner ears were scarified using a 27-gauge syringe needle to scrape the epithelia (not sufficient to cause bleeding) followed by pipette delivery of virus solution using a siliconized pipette tip. This method is modified from a previously reported infection model [60]. As controls mice were mock infected with vehicle i. e. optiprep as described above. Animals were exposed to a single dose of UVB using a custom designed Research Irradiation Unit (Daavlin, Bryan, OH) [64–66]. This irradiation unit consists of an exposure unit mounted on fixed legs. Within the exposure unit there are four UVA and UVB lamps controlled using Daavlin Flex Control Integrating Dosimeters. In this system, dose units can be entered in milli-Joules per Centimeter Square for UVB (mJ/cm2) and Joules per Centimeter Square for UVA (J/cm2); variations in energy output are automatically compensated to deliver the desired dose. This enables us to expose the animals to an accurate dosimetery of UVB radiation. For accuracy, the machine is periodically calibrated using International Light IL 1400, digital light meter (Daavlin Company). For ear shielding experiments, ear sites were shielded from UV exposure by covering the head of anesthetized mice with tin foil during UV exposure. Papillomas were measured bi-weekly. Fraction of papillomas that completely regressed was computed by expressing number of papillomas at end-point of study (i. e. 6-months post infection) compared to maximum number of papillomas formed (i. e. papillomas at 3-months post infection). Papilloma size was determined by calculating the cubic root of the product of length × width × height to obtain a geometric mean diameter (GMD) as described previously [67]. Data were represented as the means (± standard errors of the mean [SEM] of the GMDs for each test group. DTH assays were performed by topically applying 1-Chloro-2,4-Di-Nitrobenzene (DNCB) (Sigma) using a modification of a method described previously [41]. For short-term immunosuppression studies, mice irradiated with 300mJ/cm2 UVB. Ten days post UVB-irradiation mice were shaved on their backs and sensitized by topically applying 0. 5% DNCB in 50μl vehicle (4: 1 acetone: olive oil). Five days post-sensitization they were challenged with DNCB in the left ear and ear thickness was measured. To assess correlation between wart incidence and long-term immunosuppression mice were either infected with MmuPV1 or mock infected in the right ear followed by treatment with 300mJ/cm2 UVB. Mice were sensitized with DNCB (0. 3% in 50μl vehicle) 10 days post UVB exposure on their backs. Three months post-sensitization, mice were challenged with DNCB in the left ear and ear thickness was measured. Ear thickness was measured by means of Vernier calipers up to 96 hrs. The average of three readings taken at different points across the ear swelling was considered as a single measurement. Ear thickness was reported as the average of the difference between challenged and control ears 24–96 hours post challenge. The control (right) ear was the ear that was challenged with the vehicle only. The standard error in measurement was computed by measuring standard deviation. Skin was harvested, fixed in 4% paraformaldehyde, and embedded in paraffin. Serial sections (5 μm thick) were stained with hematoxylin and eosin (H&E) and evaluated for histopathological features. Immunofluorescent staining was performed on sections after deparaffinizing with xylenes and rehydrating with graded ethanol, respectively. For cytokeratin14 staining, sections were blocked for 1hour at room temperature with goat serum followed by incubation with K14 antibody (Covance) at 1: 1000 dilution for 1hour at room temperature. K14 signals were detected with Alexa-fluor 594 against rabbit. For L1-FISH dual immunofluorescent staining, first L1 immunofluorescent staining was performed. Antigen retrieval was performed using Proteinase K (20 μg/ml) for 15 minutes at 37°C. Samples were blocked for 1hour at room temperature with 5% goat serum and incubated overnight at 4°C with rabbit polyclonal immune serum directed against MusPV1 L1 [18,19] at 1: 5000 dilution (Gift from Dr. Chris Buck, NCI). To proceed with MmuPV1 fluorescent in situ hybridization (FISH), the L1 stained tissue was dehydrated using a series of ice-cold ethanols (70%, 80%, 95%) for 2 minutes each. Slides were dried by placing them in an empty container at 50°C for 5 minutes and then placed in denaturation solution (28 mL formamide, 4 mL 20X SSC pH 5. 3,8 mL water) at 72°C for 2 minutes. The ethanol series was repeated, sections were dried, and denatured digoxigenin (DIG-11-dUTP, Roche) -labeled probe was hybridized to cells overnight at 37°C in a humidified chamber. To make the probe, nick translation was used to label MmuPV1-plasmid DNA with digoxigenin. After washing with 2X SSC and 50% formamide at 50°C (for 30 minutes twice) and 2X SSC at 50°C (for 30 minutes twice), MmuPV1 FISH DNA signals were detected with a digoxigenin-specific antibody conjugated to fluorescein isothiocyanate (Sigma, F3523) at 2% by volume in STM solution (4X SSC, 5% non-fat dried milk, 0. 05% Tween-20,0. 002% sodium azide) for 30 minutes at 37°C. L1 protein signals were detected using Alexa-fluor 488 antibody against rabbit. Nuclei were counterstained with DAPI. All images were captured using a Zeiss AxioImager M2 microscope and AxioVision software version 4. 8. 2 (Jena, Germany). All statistical analyses were performed using MSTAT statistical software version 6. 1. 4 (http: //www. mcardle. wisc. edu/mstat). | Epidemiological studies have implicated that ultraviolet radiation (UVR) from sunlight drive papillomavirus-induced disease in healthy as well as immunocompromised humans. In this report we demonstrate that treatment of immunocompetent mice with UVR renders them susceptible to papillomas and associated squamous cell carcinoma when infected with the recently discovered murine papillomavirus (MmuPV1). Our data further suggest UVR increases susceptibility to virally induced disease by inducing immunosuppression. | Abstract
Introduction
Results
Discussion
Materials and Methods | dermatology
urology
medicine and health sciences
pathology and laboratory medicine
ultraviolet radiation
ears
immunology
immune suppression
cancers and neoplasms
light
electromagnetic radiation
oncology
animal models
otology
model organisms
ear infections
signs and symptoms
sexually transmitted diseases
research and analysis methods
infectious diseases
human papillomavirus infection
warts
mouse models
head
otorhinolaryngology
physics
papillomas
diagnostic medicine
anatomy
genitourinary infections
biology and life sciences
physical sciences
viral diseases
ultraviolet b | 2016 | Role of Ultraviolet Radiation in Papillomavirus-Induced Disease | 10,415 | 124 |
In the last few million years, the hominin brain more than tripled in size. Comparisons across evolutionary lineages suggest that this expansion may be part of a broader trend toward larger, more complex brains in many taxa. Efforts to understand the evolutionary forces driving brain expansion have focused on climatic, ecological, and social factors. Here, building on existing research on learning, we analytically and computationally model the predictions of two closely related hypotheses: The Cultural Brain Hypothesis and the Cumulative Cultural Brain Hypothesis. The Cultural Brain Hypothesis posits that brains have been selected for their ability to store and manage information, acquired through asocial or social learning. The model of the Cultural Brain Hypothesis reveals relationships between brain size, group size, innovation, social learning, mating structures, and the length of the juvenile period that are supported by the existing empirical literature. From this model, we derive a set of predictions—the Cumulative Cultural Brain Hypothesis—for the conditions that favor an autocatalytic take-off characteristic of human evolution. This narrow evolutionary pathway, created by cumulative cultural evolution, may help explain the rapid expansion of human brains and other aspects of our species’ life history and psychology.
In the last few million years, the cranial capacity of the human lineage dramatically increased, more than tripling in size [1–3]. This rapid expansion may be part of a gradual and longer-term trend toward larger, more complex brains in many taxa [3–7]. These patterns of increasing brain size are puzzling since brain tissue is energetically expensive [8–13]. Efforts to understand the evolutionary forces driving brain expansion have focused on climatic, ecological, and social factors [1–3,14,15]. Here we provide an integrated model that attempts to explain both the broader patterns across taxa and the human outlier. To do this, we develop an analytic model and agent-based simulation based on the Cultural Brain Hypothesis (CBH): the idea that brains have been selected for their ability to store and manage information via some combination of individual (asocial) or social learning [16–21]. That is, we develop the idea that bigger brains have evolved for more learning and better learning. The information acquired through these various learning processes is locally adaptive, on average, and could be related to a wide range of behavioural domains, which could vary from species to species. The forms of learning we model could plausibly apply to problems such as finding resources, avoiding predators, locating water, processing food, making tools, and learning skills, as well as to more social strategies related to deception, coercion, manipulation, coordination or cooperation. Our theoretical results suggest that the same underlying selective process that led to widespread social learning [22] may also explain the correlations observed across species in variables related to brain size, group size, social learning, innovation, and life history. Moreover, the parameters in the formal representation of our theory offer hypotheses for why brains have expanded more in some lineages than others [44,23]. Building on the Cultural Brain Hypothesis, our theoretical model also makes a set of predictions that we call the Cumulative Cultural Brain Hypothesis (CCBH). These predictions are derived from the parameters within the CBH model that favor an autocatalytic take-off in brain size, adaptive knowledge, group size, learning, and life history characteristic of human evolution. The CCBH has precedents in other models describing the processes that led to human uniqueness [see 18–20,24,25–29]. Since the CCBH is not a separate model, but instead additional predictions derived from the CBH model, this approach both seats humans within the broad primate spectrum created by the selection pressures we specify, and also accounts for our peculiarities and unusual evolutionary trajectory. That is, the same mechanisms that lead to widespread social learning can also open up a novel evolutionary bridge to a highly cultural species under some specific and narrow conditions—those specified by the CCBH. When these conditions are met, social learning may cause a body of adaptive information to accumulate over generations. This accumulating body of information can lead to selection for brains better at social learning as well as storing and managing this adaptive knowledge. Larger brains, better at social learning, then further foster the accumulation of adaptive information. This creates an autocatalytic feedback loop that enlists sociality (population side and interconnectedness), social learning, and life history to drive up both brain size and adaptive knowledge in a culture-gene co-evolutionary duet—the uniquely human pathway. The juvenile period expands to provide more time for social learning. As biological limits on brain size are reached [e. g. due to difficulties in birthing larger brains, even in modern populations, see 30], increases in the complexity and amount of adaptive knowledge can take place through other avenues, such as division of information (and ultimately, division of labor), mechanisms for increasing transmission fidelity, such as compulsory formal schooling, and further expansion of the “adolescent” period between fertility and reproduction, spent in additional education (i. e. delayed birth of first child) [16]. This process modifies human characteristics in a manner consistent with more effectively acquiring, storing, and managing cultural information. The CBH and CCBH are related, and can be explored with the same model, but we keep them conceptually distinct for two reasons. First, the cumulative culture-gene co-evolutionary process produces cultural products, like sophisticated multi-part tools and food processing techniques, that no single individual could reinvent in their lifetime [despite having a big brain capable of potent individual learning; 19]. The evolution of a second inheritance system—culture—is a qualitative shift in the evolutionary process that demands analyses and data above and beyond that required for the CBH. Second, it’s possible that either one of these hypotheses could hold without the other fitting the evidence—that is, it might be the CCBH explains the evolutionary trajectory of humans, but the CBH doesn’t explain the observed patterns in social learning, brain size, group and life history in primates (or other taxa); or, vice-versa. Our approach is distinct, but related to the Social Brain Hypothesis [SBH; 31], which argues that brains have primarily evolved for dealing with the complexities of social life in larger groups (e. g. , keeping track of individuals, Machiavellian reasoning, and so on). Initial evidence supporting the SBH was an empirical relationship shown between social group size in primates and some measure of brain size [different measures of brain size are typically highly correlated; 32]. Though this relationship does not hold outside the primate order, broader versions of the SBH that encompass other aspects of social cognition have been informally proposed with corresponding evidence from comparative studies. For example, a relationship has been shown between brain size and regular association in mammalian orders [6,33], mating structure in birds and mammals [33], and social structure and behavioral repertoire in whales and dolphins [34]. Efforts to formally explore these ideas isolate four distinct evolutionary mechanisms. First, McNally and collaborators have explored the Machiavellian arms race between cooperation and deception [35,36]. Second, Dávid-Barrett and Dunbar [37] simulate a relationship between coordination costs and group size showing that more complex coordination (and therefore higher cognitive complexity) is required as group size increases. Third, Gavrilets [38] models collaborative ability as a proxy for socio-cognitive competencies, exploring the effect of between-group selection and ecological pressures and showing that between-group competition can select for collaborative ability, which is then further reinforced by ecological pressures. The predictions of this last model are reinforced by a recent data-driven metabolic model tracking energy trade-offs under different types of competition [39]. Finally, exploring a distinct additional mechanism, Gavrilets and Vose [40] simulate an evolutionary competition among males for females in which males can evolve larger brains with learning abilities that permit them to acquire more effective strategies. In his seminal paper, Humphrey [41] highlighted the importance of social learning, along with several other social factors. The theory presented here is therefore consistent with this and other early research that emphasized the learning aspects of the social brain [41,42; for a more recent discussion, see Whiten & van Schaik, 2007,43,44]. However, while many verbal descriptions of the SBH are general enough to encompass most aspects of the CBH, formal instantiations of the SBH each focus on quite distinct evolutionary mechanisms: (1) deception and cooperation, (2) coordination between group members, (3) cooperation in between-competition and against ecological challenges, and (4) learning social strategies. To make progress, we argue that it’s crucial to distinguish the various evolutionary mechanisms that have often been clumped under the “social brain” rubric, and then test for the action of these various mechanisms (which need not be mutually exclusive). The CBH and CCHB are a deliberate shift in focus from “social” to “learning”; a shift with precedence in other theories, most informally expressed [for example, see 15,20,21,23,45,46]. There are, however, some clear departures from most previous approaches. First, crucial to this shift from social to learning is that group size evolves endogenously, rather than as a product of externalities (such as avoidance of predators). Second, learning is assumed to be more general than the skills and cognition required for social living. Individuals could learn skills and knowledge for social coordination, cooperation, and competition, such as social strategies to improve mating, as in [40]. But equally, these skills and knowledge may be related to other fitness relevant domains, such as ecological information about finding food or making tools. Indeed, the generality of adaptive knowledge is critical to the CCBH and the human take-off. In our approach, the potential for a runaway process to explain the human outlier arises neither from a Machiavellian arms race [35,36] nor from sexual selection [40], but instead from the rise of cumulative cultural evolution as a second system of inheritance. Ecological factors are considered in the CBH in terms of survival returns on adaptive knowledge (e. g. easier acquisition of more calories or easier avoidance of predators, where easier means requiring less knowledge). To further develop the CBH and CCBH, our models explore the interaction and coevolution of (1) learned adaptive knowledge and (2) genetic influences on brain size (storage/organizational capacity), asocial learning, social learning, and an extended juvenile period with the potential for payoff-biased oblique social learning (learning from members of the previous generation apart from biological parents). We explicitly model population growth and carrying capacity alongside genes and culture in order to theorize potential relationships between group size and other parameters, like brain size and adaptive knowledge, and also to examine the effects of sociality on the co-evolutionary process through two different parameters. We assume carrying capacity (though not necessarily population) is increased by the possession of adaptive knowledge (e. g. , more calories, higher quality foods, better predator avoidance). Our model incorporates ecological factors and phylogenetic constraints by considering different relationships between birth/death rates and both brain size and adaptive knowledge. This allows us to formalize (and in particular, simulate) these evolutionary processes for taxa facing diverse phylogenetic and ecological constraints.
Three key assumptions underlie our theory: The logic that follows from these key assumptions is first formalized using an analytic approach—an adaptive dynamics evolutionary model [47], available in the Supplementary Materials. This model captures the logic and several of the key predictions of the CBH. We then simulate the logic to capture the co-evolutionary dynamics needed to generate the CCBH. The analytical adaptive dynamics model we present in the Supplementary Materials allows us to understand the evolution of brain size, adaptive knowledge, and reliance on social learning as a function of transmission fidelity, asocial learning efficacy, and survival returns on adaptive knowledge without the complexities of co-evolutionary dynamics and explicit evolution of oblique learning and learning biases. We can derive a set of predictions from the insights gained from this model. The key predictions from the analytical model are that: However, there are several assumptions and implications underlying these basic insights, such as: Brain size and reliance on social over asocial learning will depend on factors that affect availability of adaptive knowledge, which are themselves affected by learning strategies and adaptive knowledge. In other words, there are a range of co-evolutionary dynamics that we have assumed or abstracted away in order to solve this model analytically, but which are crucial to capture and understand the full range of evolutionary dynamics. To understand the conditions under which social learning might emerge (and perhaps more interestingly, extreme reliance on social learning as in humans), we need to explore these co-evolutionary dynamics. We explore these full set of variables and explore these dynamics through an evolutionary simulation. An evolutionary simulation also allows us to properly account for population size, population structure, more sophisticated learning strategies, and life history. This model will bolster and expand on our analytic model and reveal the conditions where adaptive knowledge and brain size will increase. To explore the culture-gene co-evolutionary dynamics, we constructed an agent-based evolutionary simulation that extends our analytic model. In our simulation, individuals are born, learn asocially or socially from their parent with some probability, potentially update by asocial learning or by socially learning from more successful members of their group during an extended juvenile period, migrate between demes, and die or survive based on their brain size and adaptive knowledge. Individuals who survive this process give birth to the next generation. We are mainly interested in the effects of natural selection and learning, so we use a haploid model and ignore non-selective forces such as sex, gene recombination, epistasis, and dominance. The lifecycle of the model, as well as all variables and parameters, are shown in Fig 1 below. This simulation was written in C++ by MM (code in Supplemental Materials). To reduce bugs, two computer science undergraduate research assistants independently reviewed the code and wrote a suite of unit tests using Google’s C++ Testing Framework. The simulation begins with 50 demes, each with a population of 10 individuals. Throughout the simulation, the number of demes was fixed at 50. In early iterations of the model, we explored increasing the number of demes to 100 for some of the parameter space and found no significant impact on the results. Our starting population of 10 individuals is roughly equivalent to a real population of 40 individuals, assuming two sexes and one offspring per parent (4 × 10). As a reference, mean group size in modern primates ranges from 1 to 70 [32]. Each individual i in deme j has a brain of size bij with a fitness cost that increases with increasing brain size. Adaptive knowledge is represented by aij, where 0 ≤ aij ≤ bij. Increasing adaptive knowledge can mitigate the selection cost of a larger brain, but such knowledge is limited by brain size. Our simulations begin with individuals who have no adaptive knowledge, but the ability to fill their bij = 1. 0 sized brains with adaptive knowledge through asocial and/or social learning with some probability. To explore the idea that juvenile periods can be extended to lengthen the time permitted for learning, we included two stages of learning. In both learning stages, the probability of using social learning rather than asocial learning is determined by an evolving social learning probability variable (sij). We began our simulations with the social learning probability variable set to zero (i. e. at the beginning of the simulation, all individuals are asocial learners). To explore the invasion of asocial learners into a world of social learners, we also ran the simulation with the social learning probability variable set to one (i. e. at the beginning of the simulation, all individuals are social learners). Although social learning is widespread in the animal kingdom [22], a realistic starting point is closer to pure asocial learning. Nevertheless, the simulations starting with social learners were often useful in understanding these dynamics, so, in some cases where it is insightful, we report these results as well. Asocial learning allows for the acquisition of adaptive knowledge, independent of the adaptive knowledge possessed by other individuals. In contrast, social learning allows for vertical acquisition of adaptive knowledge possessed by the genetic parent in the first learning stage or oblique acquisition from more knowledgeable members of the deme (from the parental generation) in the second learning stage. The tendency to learn from models other than the genetic parent is determined by a genetically evolving oblique learning probability variable (vij). Thus, the simulation does not assume oblique learning or a second stage of learning [a misplaced critique of related models in our opinion; 50; but a critique not relevant to the present model, 51]. The probability of engaging in a second round of oblique social learning is a proxy for the length of the juvenile period. In the second stage of learning, if an individual tries to use social learning, but does not use oblique learning, no learning takes place beyond the first stage. This creates an initial advantage for asocial learning and cost for evolution to extend learning into an extended juvenile period. We also allow the ability to select a model with more adaptive knowledge (for oblique learning) to evolve through a payoff-bias ability variable (lij). These simulations result in a series of predicted relationships between brain size, group size, adaptive knowledge, asocial/social learning, mating structure, and the juvenile period. Some of these relationships have already been measured in the empirical literature and thus provide immediate tests of our theory. Specifically, several authors have shown positive relationships (notably in primates) between (1) brain size and social group size [44,31,52], (2) brain size and social learning [46,53], (3) brain size and length of juvenile period [54–57], and (4) group size and the length of the juvenile period [56]. Various hypotheses have been proposed for these relationships. Here we argue that they are all a consequence of a singular evolutionary process, the dynamics of which the CBH models reveal. In addition, we find that different rates of evolutionary change and the size of these relationships across taxa [6] may be accounted for by the extent to which adaptive knowledge reduces the death rate (λ in our model). This λ term captures any factor that moderates the relationship between adaptive knowledge and survival. One interpretation, but by no means the only one, is the resource richness of the ecology. For example, richer ecologies offer more ‘bang for the buck’, that is, more calories unlocked for less knowledge, allowing individuals to better offset the size of their brains. Higher λ suggest a richer ecology—or more specifically, an ecology where smarts have a greater return on survival. Indeed, research among primates has revealed that factors affecting access to a richer ecology—home range size or the diversity of food sources—are associated with brain size [58,59]. Thus, our model may help explain why both social and ecological variables seem to be variously linked to brain size. The dynamics of our model also reveal the ecological conditions, social organization and evolved psychology most likely to lead to the realm of cumulative cultural evolution, the pathway to modern humans. These predictions capture the CCBH. Our model indicates the following pathway. Under some conditions, brains will expand to improve asocial learning and thereby create more adaptive knowledge. This pool of adaptive knowledge leads to selection favoring an immense reliance on social learning, with selective oblique transmission, allowing individuals to exploit this pool of growing knowledge. Rogers’ [60] paradox, whereby social learners benefit from exploiting asocial learners’ knowledge, but do not themselves generate adaptive knowledge, is solved by selective oblique social learning transmitting accidental innovations to the next generation. Under some conditions, an interaction between brain size, adaptive knowledge, and sociality (deme size and interconnectedness) emerges, creating an autocatalytic feedback loop that drives all three—the beginning of cumulative cultural evolution.
Overall, our evolutionary simulations produce patterns that are consistent with the existing empirical data, though, of course, our simulation produces many patterns that have not yet been examined. The causal relationships underlying these patterns—the CBH and our simulated instantiation of it—are outlined in Fig 5 below. Before digging into the details, we summarize these relationships as follows: The Cultural Brain Hypothesis predicts that brain size, group size, adaptive knowledge, and the length of juvenile period should be positively intercorrelated among taxa with greater dependence on social learning, but are generally weaker or non-existent among taxa with little social learning. There has been less empirical data published for species with little social learning, perhaps due to a bias toward only publishing statistically significant relationships. The lack of this data makes it difficult to test the asocial regime predictions. The strength of these relationships, overall brain size, and the evolution of different regimes vary, depending on the other parameters in our model. These include ecological factors such as the strength of the relationship between adaptive knowledge and survival (λ), which we will call the “richness of the ecology” as a shorthand, as well as other factors that are themselves products of evolution (which we’ve held fixed as phylogenetic constraints): the relationship between adaptive knowledge and relative reproductive payoffs (φ), which are related to reproductive skew, mating structure, and the level of individual vs between-group selection (we will refer to this as reproductive skew as a shorthand); transmission fidelity (τ); and asocial learning efficacy (ζ). Other models have theorized the evolution of these structures, tendencies, and abilities, but here we are interested in the effect of these factors on the co-evolutionary processes shown in Fig 5. Beyond the hypothesis that social learning, brain size, adaptive knowledge, and group size may have coevolved to create the patterns found in the empirical literature, we are also interested in the conditions under which these variables might interact synergistically to create highly social species with large brains and substantial accumulations of adaptive knowledge (humans). To assess when an accumulation of adaptive knowledge becomes cumulative cultural evolution, we apply a standard definition of cumulative cultural products as being those products that a single individual could not invent by themselves in their lifetime. To calculate this for our species, we ask what the probability is that an individual with the average brain size of the species would invent the mean level of adaptive knowledge in that species via asocial learning. In addition to our main simulations that began with asocial learners, we also ran a set of simulations that began with social learners. Although social learning is widespread in the animal kingdom [22] and the most realistic starting conditions are somewhere in-between these two extremes (no social learning and complete social learning), these realistic conditions are likely closer to no social learning than complete social learning. Nonetheless, running our simulations beginning with social learning provides an upper bound on our predicted patterns and also offers additional insights. There are two key insights. The first is that social learning is maladaptive in a world with little knowledge (Fig 16). With little knowledge for social learners to exploit, asocial learners quickly invade. However, since some social learning is present, once sufficient knowledge has been generated, social learning is again at an advantage, with additional innovations generated in the process of social learning [16]. The second key insight is closely related: consistent with previous models [24], the presence of social learning expands the range of parameters in which cumulative culture is adaptive. Fig 9b shows a greater number of species with social learning (compared to Fig 9a). Fig 15a reveals that more monogamish societies, or at least societies with a reduced reproductive skew or reduced individual selection are more likely to enter the realm of cumulative cultural evolution. Fig 15b reveals that cumulative cultural evolution is more likely to evolve when transmission fidelity is higher. Both Fig 16a and 16b reveal that the range of parameters that lead to the realm of cumulative cultural evolution expands if more social learning is present in the ancestral state.
Our model provides a potential evolutionary mechanism that can explain a variety of empirical patterns involving relationships between brain size, group size, innovation, social learning, mating structures, and developmental trajectory, as well as brain evolution differences among species. It can also illuminate the different rates of evolution and overall brain size that have been found in different taxa and help explain why brain size correlates with group size in some taxa, but not others. In contrast to competing explanations, the key message of the Cultural Brain Hypothesis (CBH) is that brains are primarily for the acquisition, storage and management of adaptive knowledge and that this adaptive knowledge can be acquired via asocial or social learning. Social learners flourish in an environment filled with knowledge (such as those found in larger groups and those that descend from smarter ancestors), whereas asocial learners flourish in environments where knowledge is socially scarce, or expensive but obtainable through individual efforts. The correlations that have been found in the empirical literature between brain size, group size, social learning, the juvenile period, and adaptive knowledge arise as an indirect result of these processes. The Cumulative Cultural Brain Hypothesis posits that these very same processes can, under very specific circumstances, lead to the realm of cumulative cultural evolution. These circumstances include when transmission fidelity is sufficiently high, reproductive skew is in a Goldilocks’ zone close to monogamy (or equally, there is some, but not too much individual-level selection), effective asocial learning has already evolved, and the ecology offers sufficient rewards for adaptive knowledge. In making these predictions, the Cultural Brain Hypothesis and Cumulative Cultural Brain Hypothesis tie together several lines of empirical and theoretical research. Under the broad rubric of the Social Brain or Social Intelligence Hypothesis, different researchers have highlighted different underlying evolutionary mechanisms [35–38,40]. These models have had differing levels of success in accounting for empirical phenomena, but they highlight the need to be specific in identifying the driving processes that underlie brain evolution in general, and the human brain specifically. From the perspective of the CBH, these models have been limited in their success, because they only tell part of the story. Our results suggest that the CBH can account for all the empirical relationships emphasized by the Social Brain Hypotheses, plus other empirical patterns not tackled by the SBH. Moreover, our approach specifies a clear ‘take-off’ mechanism for human evolution that can account for our oversized crania, heavy reliance on social learning with sophisticated forms of oblique transmission (and possibly the emergence of adolescence as a human life history stage), and the empirically-established relationship between group size and toolkit size/complexity [76]—as well as, of course, our species’ extreme reliance on cumulative culture for survival [19]. Our results echo some of the predictions of models of learning and levels of competition. In particular, an early paper by Gavrilets and Vose [40] pitched as a model of Machievellian intelligence might equally be viewed as a model of culture, showing similar co-evolution of brain size, adaptive knowledge and learning ability. A more recent paper by Gavrilets [38] modeled socio-cognitive competencies in competition between groups, between individuals, and against the environment. This model showed how socio-cognitive competencies were enhanced with weaker individual-level selection, which is echoed in the CBH predictions. Finally, a recent paper by González-Forero and Gardner [39] model the energy tradeoff between brains, bodies, and reproduction under different challenges and costs. This energy model takes a different approach to how the variable and parameters are specified, particularly in tracking in ratios of brain size and energy extraction efficiency, making it difficult to directly compare to the CBH and CCBH. While the mapping is not perfect, these are potentially complementary models, particularly in the overall result that humans emerge where competition is 60% ecological, 30% cooperative, and 10% between groups with little individual-level competition, reflecting the importance of a high λ and low φ in our model. The authors conclude by noting how their model may intersect with a model of culture like the CBH in how social learning and life history interact with ecological factors and the relationship between adaptive knowledge and survival. Our simulation’s predictions are consistent with other theoretical work on cultural evolution and culture-gene coevolution. For example, several researchers have argued for the causal effect of sociality on both the complexity and quantity of adaptive knowledge [77,78]. Similarly, several researchers have argued for the importance of high fidelity transmission for the rise of cumulative cultural evolution [29,48,79]. Cultural variation is common among many animals (e. g. , rats, pigeons, chimpanzees, and octopuses), but cumulative cultural evolution is rare [24,80]. Boyd and Richerson [24] have argued that although learning mechanisms, such as local enhancement (often classified as a type of social learning), can maintain cultural variation, observational learning is required for cumulative cultural evolution. Moreover, the fitness valley between culture and cumulative culture grows larger as social learning becomes rarer. Our model supports both arguments by showing that only high fidelity social learning gives rise to cumulative cultural evolution and that the parameter range to enter this realm expands if social learning is more common (see Fig 15). In our model, cumulative cultural evolution exerts a selection pressure for larger brains that, in turn, allows more culture to accumulate. Prior research has identified many mechanisms, such as teaching, imitation, and theory of mind, underlying high fidelity transmission and cumulative cultural evolution [18,28,81]. Our model reveals that in general, social learning leads to more adaptive knowledge and larger brain sizes, but shows that asocial learning can also lead to increased brain size. Further, our model indicates that asocial learning may provide a foundation for the evolution of larger-brained social learners. These findings are consistent with Reader et al. [20], who argue for a primate general intelligence that may be a precursor to cultural intelligence and also correlates with absolute brain volume. And, though more speculative, key mutations, such as the recently discovered NOTCH2NL genes [82,83], may have allowed for the transition from smart asocial learners to larger brained social learners as specified in the narrow pathway of the CCBH. The CBH is consistent with much existing work on comparative cognition across diverse taxa. For example, in a study of 36 species across many taxa, MacLean et al. [84] show that brain size correlates with the ability to monitor food locations when the food was moved by experimenters and to avoid a transparent barrier to acquire snacks, using previously acquired knowledge. The authors also show that brain size predicts dietary breadth, which was also an independent predictor of performance on these tasks. Brain size did not predict group size across all these species (some of whom relied heavily on asocial learning). This alternative pathway of asocial learning is consistent with emerging evidence from other taxa. For example, in mammalian carnivores brain size predicts greater problem solving ability, but not necessarily social cognition [85,86]. These results are precisely what one would expect based on the Cultural Brain Hypothesis; brains have primarily evolved to acquire, store and manage adaptive knowledge that can be acquired socially or asocially (or via both). The Cultural Brain Hypothesis predicts a strong relationship between brain size and group size among social learning species, but a weaker or non-existent relationship among species that rely heavily on asocial learning. Our simulation results are also consistent with empirical data for relationships between brain size, sociality, culture, and life history among extant primates [e. g. 87] and even cetaceans [34], but suggest a different pathway for humans. In our species, the need to socially acquire, store, and organize an ever expanding body of cultural know-how resulted in a runaway coevolution of brains, learning, sociality and life history. Of course, this hypothesis should be kept separate from the CBH: at the point of the human take-off, brain size may have already been pushed up by the coordination demands of large groups, Machiavellian competition, or asocial learning opportunities [19]. For example, Machiavellian competition may have elevated mentalizing abilities in our primate ancestors that were later high-jacked, or re-purposed, by selective pressure associated with the CCBH to improve social learning by raising transmission fidelity, thereby creating cumulative cultural evolution. Thus, the CBH and CCBH should be evaluated independently. Note that in understanding these results, it is worth remembering that our model assumes a relationship between brain size and adaptive knowledge capacity, but not adaptive knowledge; similarly between adaptive knowledge and carrying capacity, but not population size; and between brain size and decreased survival and adaptive knowledge and increased survival. These tradeoffs and co-evolutionary dynamics help us understand why we see stronger or weaker relationships between social and asocial species. Note that our model seeks to (1) show why brain size, adaptive knowledge, social learning, group size, and lifespan are intercorrelated across the animal kingdom (CBH) and (2) how the very same processes that lead to these interconnections, can, under some specific circumstances, lead to the realm of cumulative cultural evolution—the uniquely human pathway. Within the realm of cumulative culture, the dynamics change in ways that are not captured by this model. For example, in order to sustain ever-growing levels of cultural complexity, cultures can generate ways to increase sociality and transmission fidelity. With sufficiently complex culture, mechanisms may evolve to more efficiently share the fruits of rare innovations, allowing for increases in cultural variance that may be individually costly. Moreover, cumulative culture, once acquired, can increase an individual efficacy in subsequent asocial learning [for a discussion of these ideas, see 16]. In developing the simulation, we formalized the minimal set of assumptions and parameters that capture the logic of the CBH and CCBH. There are a number of extensions, variations, and additional parameters that would improve our understanding of the evolution of brain size. There were several assumptions that simplified our model, making it more computationally tractable. Future models may address some of these shortcomings and explore additional parameters. One such improvement is to explicitly track different cultural traits with different cognitive costs and fitness payoffs. By doing this, we could better explore the benefits to migration and cultural recombination. We would also like to more fully explore the impact of the relationship between adaptive knowledge and carrying capacity. Currently, the richness of the ecology only affects individual survival based on paying the calorie cost of costly brains, but the richness of the ecology also affects the carrying capacity of the population with consequent effects for the dynamics between brain size, adaptive knowledge and population size. Another previously mentioned future improvement is the endogenization of transmission fidelity (τ) and reproductive skew (φ). These parameters are themselves subject to genetic and cultural evolutionary processes and thus ought to be modeled as endogenous variables. In our model, we can discuss the effect of different evolutionary outcomes or values of transmission fidelity and reproductive skew, but not their evolution. Two or three regimes emerged in our models based on different ecological and phylogenetic constraints. In a future model, we plan to explore the adaptive dynamics of these different regimes, exploring the invasion fitness of the different equilibrium states discovered in our model. These models will help us better understand the evolutionary dynamics that may have occurred when different previously geographically separated hominin species encountered each other (e. g. , the European encounter between modern humans and their larger-brained Neanderthal cousins). The key improvements that we are eager to explore could be summarized as: (1) endogenizing the evolution of transmission fidelity and reproductive skew, (2) explicitly tracking different cultural traits with different cognitive costs and fitness payoffs, and (3) more thoroughly exploring the brain shrinkage that occurs during the transition from reliance on asocial learning to reliance on social learning. This brain shrinkage (see Fig 16) occurs as social learners invade by “stealing” the knowledge of the asocial learners, without having to figure it out for themselves. Once the population is mostly made up of social learners, brain size begins to increase again. These results hint that the process underlying the Cultural Brain Hypothesis and Cumulative Cultural Brain Hypothesis may also help explain evidence suggesting that human brains have been shrinking in the last 10,000 to 20,000 years [89]. Although this shrinkage in brain size corresponds to shrinking in body size, it may be evidence that our species is not at equilibrium. | Humans have extraordinarily large brains, which tripled in size in the last few million years. Other animals also experienced a significant, though smaller, increase in brain size. These increases are puzzling, because brain tissue is energetically expensive—a smaller brain is easier to maintain in terms of calories. Here we present a theory, captured in an analytic and computational model, that explains these increases in brain size: The Cultural Brain Hypothesis. The theory relies on the idea that brains expand to store and manage more information. Brains expand in response to the availability of information and calories. Information availability is affected by learning strategies (e. g. learning from others or learning by yourself), group size, mating structure, and the length of the juvenile period, which co-evolve with brain size. The model captures this co-evolution under different conditions and describes the specific and narrow conditions that can lead to a take-off in brain size—a possible pathway that led to the extraordinary expansion in our own species. We call these conditions the Cumulative Cultural Brain Hypothesis. These theories are supported by our tests using existing empirical data. | Abstract
Introduction
Models
Results
Discussion | learning
organismal evolution
ecology and environmental sciences
hominin evolution
sociology
human evolution
social sciences
vertebrates
neuroscience
learning and memory
animals
mammals
primates
cultural evolution
cognitive psychology
animal behavior
theoretical ecology
zoology
animal sociality
behavior
hominid evolution
culture
psychology
eukaryota
ecology
natural selection
biology and life sciences
evolutionary biology
cognitive science
amniotes
evolutionary processes
organisms | 2018 | The Cultural Brain Hypothesis: How culture drives brain expansion, sociality, and life history | 8,133 | 245 |
The endoplasmic reticulum (ER) is the site of synthesis of secreted and membrane proteins. To exit the ER, proteins are packaged into COPII vesicles through direct interaction with the COPII coat or aided by specific cargo receptors. Despite the fundamental role of such cargo receptors in protein traffic, only a few have been identified; their cargo spectrum is unknown and the signals they recognize remain poorly understood. We present here an approach we term “PAIRS” (pairing analysis of cargo receptors), which combines systematic genetic manipulations of yeast with automated microscopy screening, to map the spectrum of cargo for a known receptor or to uncover a novel receptor for a particular cargo. Using PAIRS we followed the fate of ∼150 cargos on the background of mutations in nine putative cargo receptors and identified novel cargo for most of these receptors. Deletion of the Erv14 cargo receptor affected the widest range of cargo. Erv14 substrates have a wide array of functions and structures; however, they are all membrane-spanning proteins of the late secretory pathway or plasma membrane. Proteins residing in these organelles have longer transmembrane domains (TMDs). Detailed examination of one cargo supported the hypothesis that Erv14 dependency reflects the length rather than the sequence of the TMD. The PAIRS approach allowed us to uncover new cargo for known cargo receptors and to obtain an unbiased look at specificity in cargo selection. Obtaining the spectrum of cargo for a cargo receptor allows a novel perspective on its mode of action. The rules that appear to guide Erv14 substrate recognition suggest that sorting of membrane proteins at multiple points in the secretory pathway could depend on the physical properties of TMDs. Such a mechanism would allow diverse proteins to utilize a few receptors without the constraints of evolving location-specific sorting motifs.
The endoplasmic reticulum (ER) is the entry site into the secretory pathway, responsible for the folding, maturation, and trafficking of all secreted, membrane-bound, and secretory pathway resident proteins. Once folded, the proteins exit the ER as cargo within COPII-coated vesicles that bud from ER exit sites [1], [2]. Active concentration of proteins into the vesicles [3]–[6] occurs either by direct interaction with the Sec23 and Sec24 subunits of the COPII coat or else are mediated through a diverse group of proteins that function as “adaptors” and have been termed cargo receptors [4], [7]. Cargo receptors allow sorting of cargo that cannot directly bind Sec23/24, or cargo whose exit requires quality control or regulation [8], [9]. The prevalent way to identify cargo for a cargo receptor entails testing selected individual proteins in transport assays or in vitro COPII budding reactions, as was utilized to pair glycosylphosphatidylinositol (GPI) -anchored proteins with their cargo receptors—the p24 family of proteins [10]–[12]. Because of the complexity of these approaches, only a few additional cargo receptors have since been identified (Table S1). Moreover, despite their important function in ER exit and their potential for regulating the flow of traffic in the entire secretory pathway, there is still little information about the entire spectrum of cargos for a given cargo receptor or what defines its cargo specificity. Importantly, no attempt has yet been made to pair large sets of possible cargos with their cargo receptors in a manner that is systematic and unbiased. The lack of systematic data has hindered the identification of the determinants shared by specific sets of cargo that allow their recognition by a particular cargo receptor. Identification of such determinants might also shed light on the purpose and mechanism of action by which a given cargo receptor operates. Here we describe a systematic approach that aims to complement the traditional methods of cargo and cargo receptor discovery which we call “PAIRS” (pairing analysis of cargo receptors). The PAIRS approach utilizes robotic methodologies to genetically manipulate Saccharomyces cerevisiae libraries containing green fluorescent protein (GFP) -tagged cargo [13]–[15], followed by automated microscopy to identify mutated backgrounds that cause ER retention of cargo. Using PAIRS we have increased the number of known cargos for a set of nine cargo receptors. Since our approach probes a large set of proteins for their receptor requirements it defines both groups that are dependent and that are independent of any given cargo receptor. Combined, this should help to define the rules of specificity for each of the cargo receptors. We demonstrate the utility of this approach by using the set of cargo uncovered for the cargo receptor Erv14 to formulate a hypothesis on its mode of substrate recognition. The large group of cargo that require Erv14 as their cargo receptor do not share a detectable functional similarity or sequence motif. However, all identified cargo resides in late secretory pathway membranes that are populated with proteins of longer transmembrane domains (TMDs) than TMDs of ER resident proteins [16]. This raises the hypothesis that cargo specificity of Erv14 is determined by TMD length. Following up on one substrate, Mid2, we show this to indeed be the case. Thus Erv14 may be able to recognize a diversity of cargo by recognizing a shared physical property rather than a specific sequence. This also suggests a resolution for conflicting findings on the effect of TMD length on protein retention in the ER or Golgi [17]–[22].
To pair as many cargo proteins as possible with their corresponding cargo receptors in a systematic, non-biased approach, we devised a methodology we call PAIRS. PAIRS is based on the idea that when a cargo receptor is missing, then its cargo accumulates in the ER and that this can be visualized by using fluorescently tagged cargo. The PAIRS approach can be used for two purposes. First, it can be used to uncover the cargo receptor for a specific cargo of interest by expressing that specific cargo fused to GFP on the background of mutations in trafficking-related proteins. Second, it can be used to uncover the spectrum of cargo for a putative cargo receptor by visualizing a large number of strains with various GFP-tagged cargo on the background of mutations in that cargo receptor. The approach relies on systematic creation of genetically modified strains using the synthetic genetic array (SGA) methodology [13], [14], [23], which is followed by acquisition of fluorescent images of all strains using a high-throughput automated microscopy platform. Finally, manual examination of the resulting images uncovers strains in which the mutation causes ER retention of cargo, implying a cargo receptor/cargo pair (Figure 1). To determine whether our methodology can indeed facilitate the identification of a cargo receptor for an arbitrary cargo, we chose Tpo4, a plasma membrane multidrug transporter involved in polyamine transport, whose ER exit had not been shown to rely on a particular cargo receptor. We first created an SGA compatible query strain expressing Tpo4-GFP from its endogenous promoter. We then collected strains of mutants in trafficking related proteins from either the deletion library (for non-essential genes) [11], or from the decreased abundance by mRNA perturbation (DAmP) library (for hypomorphic alleles of essential genes) [24] (for a full list of strains used and the proposed function of their corresponding protein see Table S6). Using the SGA approach [14], we crossed the Tpo4-GFP into the mutant library creating a new library of haploid yeast strains each expressing Tpo4-GFP on the background of a mutation in a single gene. Visualization of these strains demonstrated that all but one of the strains did not alter Tpo4-GFP' s localization (Figure 2). Only the Δerv14 strain displayed ER accumulation of Tpo4-GFP (red arrows in Figure 2). Although Erv14 is a known cargo receptor [25]–[27], it has not been previously implicated in trafficking of Tpo4. Our analysis suggests that Erv14 is the cargo receptor for Tpo4 and demonstrates that the PAIRS methodology can be used to find a cargo receptor for a given cargo of interest. We next wanted to utilize the PAIRS methodology to map the spectrum of cargos for a cargo receptor of interest. We therefore created nine query strains, each carrying a deletion or a DAmP hypomorphic allele of a putative cargo receptor: Δerv14, Δerv15, Δerv26, Δerv29, Δemp24, Δemp47, Δgsf2, Δchs7, and Shr3-DAmP. To generate a library of GFP-tagged cargo we used the Yeast GFP Fusion Localization Database to identify fusion proteins that were reported to reside in post-ER compartments (Golgi, puncta, vacuole, or cell periphery) [15]. These were filtered to remove all those without TMDs or a signal peptide giving rise to 157 strains. The strains and controls were assembled and all of these were crossed into each of the nine query strains thus generating nine new libraries of GFP-tagged cargo proteins, each lacking an individual putative receptor. Inspection of the strains showed that the majority of cargos (126 out of 157) managed to exit the ER in all deletion backgrounds, suggesting that they do not solely rely on any one of the nine cargo receptors studied here for their ER exit. This suggests that most proteins can either bind the COPII coat directly, depend on redundant mechanisms for ER exit, rely on as yet undiscovered cargo receptors, or that they are exported out of the ER by spontaneous “bulk flow. ” However, for all but one cargo receptor, Erv15, we could find at least one cargo that depended on it. The annotation of Erv15 stems from its high homology to the cargo receptor Erv14 (63%), and it appears to be required to augment the activity of Erv14 in transporting particular cargo in sporulating cells but not under normal growth conditions [25], [26]. The full set of novel cargos found for each of the other seven cargo receptors is shown in Figure 3 (for previously characterized cargo that were verified by the screen see Figure S1). It appears that there is never a complete blockage of ER exit; this may simply reflect the proteins leaking out of the ER once they have accumulated to high levels, but there may also be some functional redundancy in the ER exit machinery. Since this is the first time that all cargo receptors have been studied in the same system and under the same conditions in a systematic manner, the spectrum of cargo uncovered for each cargo receptor could also be used to start defining the functional rules guiding the recognition mode. For example, all cargo for Erv26 comprised of Golgi-localized mannosyltransferases (Figure 3C) as had previously been suggested [28], [29]. However Erv26 seems to be specific to a subset of this functional group as several additional mannosyltransferases (Mnn1-GFP, Mnn11-GFP, Mnn10-GFP, Anp1-GFP, and Hoc1-GFP) did not accumulate in the ER in this background (unpublished data). Another example for specificity is our finding of only a single novel cargo for Shr3 and the identity of this cargo as an amino acid permease (Figure 3E) as are all previously identified cargo, supporting the notion that Shr3 is a dedicated cargo receptor for amino acid permeases [8], [9], [30], [31]. A similar picture emerges for Gsf2 whose novel cargo all fall into the same functional category of sugar transporters (Figure 3B) as reported previously [32]. Moreover, the sugar transporter Hxt2-GFP previously shown to be independent of Gsf2, is indeed properly localized to the plasma membrane (unpublished data), supporting the notion that Gsf2 is involved in exit of only particular sugar transporters from the ER. Other cases are less clear, such as the three seemingly unrelated cargo that we uncovered for Erv29 (Figure 3A). Previous reports identified three soluble proteins as requiring Erv29 for efficient ER exit (PrA, CPY, and α-factor) [33], [34], and although one of the new proteins is a soluble protein (Pry1), two others (Ear1 and Mam3) are membrane proteins of the vacuole or endosome. Perhaps the most striking finding is the large number of proteins that require Erv14 for efficient ER exit. Erv14 was identified as being enriched in COPII vesicles and shown initially to be required for the ER exit of the plasma membrane protein Axl2 [26]. Recent work has shown that mutants lacking Erv14 also show ER accumulation of the proteins Sma2 [25], Mid2, Gap1, Hxt1, and Hxt2 [35]. Our PAIRS approach identified that Erv14 is required for the ER exit of 32% of the plasma membrane proteins checked (18 of 57) (Figures 3F and S2A and S2B). Among these proteins are permeases (e. g. , Mep2-GFP), transporters (e. g. , Hxt2-GFP and Nha1-GFP), multidrug transporters (e. g. , Snq2-GFP and Tpo4-GFP), lipid flippases (e. g. , Cdc50-GFP and Dnf1-GFP), eisosome components (Sur7-GFP), and proteins involved in cell polarity or cell wall regulation (e. g. , Mid2-GFP and Axl2-GFP). Some have a single TMD whilst others are polytopic with up to 12 TMDs. Hence there is no obvious functional or structural similarity between the proteins affected by Erv14. Consistent with previous work, Erv14 was not required for ER exit of soluble proteins and non-conventional membrane tethered proteins such as GPI-anchored and tail-anchored proteins (Figure S3) [26], [35]. To strengthen the predictions made by our PAIRS methodology we analyzed the physical interactors of Erv14 under the assumption that direct cargo should physically interact with its cargo receptor. To this end, we immunoprecipitated HA-tagged Erv14 (which completely retains the function of the endogenous Erv14 [36]) from microsomes and analyzed the precipitated proteins by mass spectrometry. Using this approach we could corroborate eight out of the 23 cargo predicted by PAIRS as physically interacting (Figure S4A). We also found five interacting proteins that could be cargo; however, they were not examined in our original screen because of mislocalization of the C-terminal tagged fusion. To verify these proteins as cargo we made strains expressing N-terminal GFP fusion proteins and found that two of them are indeed retained in the ER in Δerv14 (Figure S4B). Live cell imaging confirmed that Erv14' s absence decreased the kinetics of ER exit of predicted cargo (Figure S5), raising the question of how it could accelerate the ER exit of such a defined set of diverse proteins. We performed in-depth sequence analysis of Erv14 cargo but could not uncover any identifiable sequence motifs (unpublished data). However, the fact that all cargos of Erv14 are membrane proteins destined to reside in the membranes of the late secretory pathway suggested that inherent characteristics of the membrane-spanning region might be responsible for the recruitment of Erv14. Indeed, a comprehensive comparison of TMDs of bitopic proteins from different compartments has shown that TMDs from post-Golgi compartments are significantly longer, suggesting that the bilayer is thicker [16]. Thus a larger hydrophobic portion, adapted for the apparently thicker bilayer of the plasma membrane, may be the trait that determines potential cargo for Erv14. We thus investigated the dependence of ER exit of an Erv14-regulated cargo on its TMD length. To assay the effect of TMD length on protein sorting by Erv14 we used the plasma membrane cargo protein Mid2 as a reporter. Mid2 is a non-essential type I membrane protein with a signal peptide and a single, 26–amino acid-long, TMD (Figure 4A). One advantage of using Mid2 is that its maturation along the secretory pathway can be monitored owing to the presence of luminal modifications by a single N-linked glycan and multiple O-linked glycans. Since the extension of the O-linked glycans occurs in the Golgi and results in reduced mobility on SDS gels, this can be used to assay the extent of ER exit [37]. To remove any possible interference that may stem from specific sequences in the TMD we replaced the endogenous TMD with a stretch of 26 leucines (Mid2L26M). The residues at either end of the TMD were modified to be basic to provide sharp ends to the hydrophobic region (Figure 4B), and because basic residues are the most common charged residues at both the cytoplasmic and luminal ends of the TMDs of yeast plasma membrane proteins [16]. To ascertain that these changes did not alter the basic cargo properties of Mid2, we expressed a GFP-tagged form of Mid2L26M in yeast and observed that it localizes to the plasma membrane in a manner identical to the wild-type protein (Figure 4C). This finding indicates that our synthetic reporter is correctly localized and that Mid2 does not require specific motifs within its TMD for its trafficking through the secretory pathway. We next generated variants of Mid2 in which the polyleucine TMD was shortened in increments of two residues to give a set of variants spanning the range of 14–26 residues. To examine the trafficking of these variants, each one was expressed under the control of a galactose-inducible, glucose-repressible, promoter. By inducing transcription with galactose for 90–120 min followed by termination upon return to glucose, it was possible to generate a pulse of protein whose progress through the secretory pathway could be followed by blotting and microscopy. When we compared the TMD-length variants by blotting we found that they behaved differently (Figures 4D and S6A). The majority of the longest TMD form was exported from the ER even during the pulse of induction, and the remaining material was rapidly matured during the chase. However, as TMDs became shorter in the variants, the ER form took a longer time to disappear. Similar results were obtained with an independent set of TMD variants in which the polyleucine stretch was flanked by a tryptophan at either end, a residue sometimes enriched in this position (Figure S6B). In fact, in the shorter TMD variants we could not observe Golgi forms accumulating but instead we observed an accumulation of a band corresponding to free GFP. This free GFP likely reflects degradation of the short TMD variants in the vacuole, as it is not present in a strain expressing the 16-leucine variant (Mid2L16M) and lacking the vacuolar hydrolase Pep4 (Figure 4E). Therefore we assume that once out of the ER and in the Golgi, the shorter TMD variants are directed to the vacuole and degraded, whereas the longer TMD forms are trafficked correctly to the plasma membrane. Thus, the steady state pool of the Golgi-modified, but not yet degraded, form of the shorter variants will be low. Quantitation of the ER form confirmed that the longer TMD variants exited the ER much more rapidly, with 20 leucines being the point of transition (Figure 4F). We next asked what effect TMD length has on Erv14-dependent exit. By repeating the pulse chase experiments in a Δerv14 strain we found that the ER exit of the long form of Mid2 was drastically slowed down, demonstrating a dependence on Erv14 (Figure 4G). However, we could not observe any change in the slow exit rate of the short TMD form Mid2L16M, indicating that this shorter form exits the ER in an Erv14-independent manner (Figure 4H). Taken together with the fact that the polyleucine TMD did not abolish Erv14 dependence of Mid2L24M, this strongly supports the idea that the length of the TMD and not its sequence are a determinant of Erv14-mediated sorting. The above results are consistent with the idea that Erv14 selectively acts as a cargo receptor on proteins with a long TMD. However it is also possible that a short TMD acts as an ER retention signal by discouraging entry into COPII vesicles or ER exit domains, and hence this effect over-rides the ability of Erv14 to extract the cargo from the ER. To investigate this possibility we asked whether the short TMD forms of Mid2 could exit more rapidly if directed to COPII vesicles by a different mechanism. Thus we created new versions of the Mid2 reporter that were fused with the cytoplasmic tail of the Golgi-localized Sys1 protein that contains a DXE motif for direct binding to Sec24 (Figure 5A) [7], [38], [39]. When these constructs were expressed in wild-type cells we found that the ER exit of the short forms was significantly accelerated in comparison to the form that lacked the DXE motif and was now closer to the rate of the longer TMDs (Figure 5B and 5D). In addition, we observed a reduction in the levels of free GFP, indicating that once the constructs had left the ER, they were no longer rapidly degraded. When the DXE motif in the Sys1 tail was mutated to AXE the rate of ER exit was reduced, confirming that the effect is mediated by COPII binding (Figure 5C and 5E). Interestingly, the stabilization of the shorter TMD variants (i. e. , the reduction in accumulation of free GFP during the chase) conferred by the Sys1 tail was retained even after DXE was mutated, suggesting that other sequences in the Sys1 tail are responsible for this effect. Taken together, these results indicate that the reduction in ER exit rate seen upon TMD shortening is a feature of an Erv14-dependent cargo, but not a feature of a cargo protein that is concentrated in COPII vesicles by other mechanisms. This suggests that Erv14 specifically directs the ER exit of proteins with longer TMDs. In the above experiments we found that the Sys1 tail enabled all of the polyleucine TMD variants to efficiently exit the ER. This allowed us to investigate the effect of TMD length on later trafficking steps without complications from variations in ER exit rate. We thus examined the distribution of the different TMD length forms at the end of the galactose pulse (Figure 6A). As expected, the longer TMDs showed predominantly plasma membrane staining. However, the shorter TMD variants accumulated in intracellular puncta that were also labeled with the late Golgi protein Sec7 (Figure 6A). This finding suggests that TMD length affected Golgi exit as well as ER exit. To confirm that this difference was not due to kinetic effects, we re-examined the distribution of TMD variants expressed constitutively under the control of the endogenous Mid2 promoter. As observed for the galactose-induced versions, the longer TMD variants show a clear plasma membrane localization, but the variants with TMDs of less than 22 residues again accumulated in puncta with little if any cell surface staining (Figure 6B). These puncta co-localized with the late Golgi marker Sec7, but not with the early Golgi marker Rud3, confirming that the protein was accumulating in trans Golgi compartments (Figure 6B and 6C). The same Golgi accumulation of shorter TMDs was observed in cells lacking the End3 protein that is required for endocytosis [40], negating the possibility that the shorter TMD variants are simply travelling to the surface and being more rapidly endocytosed (Figure 6D) [41]. To determine whether this TMD length-dependent sorting at the Golgi also required Erv14 we expressed the 16 and 24 leucine constructs with Sys1 tails in Δerv14. Both variants displayed the same ER exit rates; however, they still differed in location (Figure 7A and 7B). This indicates that the difference in localization was not dependent on any residual differences in ER exit rate conferred by the Erv14 protein, or by Erv14 chaperoning the protein through the Golgi and having a second role at Golgi exit. Taken together, these results demonstrate that the length of the TMD can determine the rate at which Mid2 leaves the Golgi to travel to the plasma membrane and, unlike the effect of TMD length on ER exit rate, this sorting does not require Erv14.
The PAIRS methodology aims at pairing cargos with their cargo receptors and to elaborate on the existing body of knowledge on cargo extraction from the ER. Our analysis did not include all possible cargos since the proteins in the GFP library are all C-terminally tagged and under their endogenous promoter. This resulted in a number of strains whose proteins were mislocalized due to the tag or had fluorescence levels below detection [15]. This may explain why we could not identify some of the previously recognized cargo/cargo receptor pairs. Despite these caveats, the power of the PAIRS analysis is that it is not biased within the set of pre-selected proteins, allowing us a broad overview of cargo receptors function. This has allowed us to gain insight to the rules governing their specificity and to present the first step towards creation of a cellular “traffickome. ” We demonstrate the value of the PAIRS approach by identifying putative new cargo proteins for most of the known cargo receptors of yeast. This represents 31 of the 157 proteins tested, of which 27 had not been previously linked to a cargo receptor. This is likely to be a slight underestimate of the success rate as some of the proteins stated to have a vacuolar localization in the GFP Database probably reflect ER residents displaced by the tag and hence would not be expected to have a dedicated cargo receptor. One general trend seen across all hits is that removal of a receptor did not result in a complete block of ER exit of its GFP-tagged cargo. This is consistent with previous studies of known cargo/receptor pairs including GPI-anchored proteins [10]. It may be that cargo receptors typically act to accelerate the exit of a particular cargo, and that some bulk flow always occurs, with the volume of this flow increasing for a particular protein if it accumulates in the ER. The increased knowledge of the range of cargo for each specific cargo receptor should make it easier to generate hypotheses as to what determines the selective recognition of particular cargo by individual cargo receptors. Indeed, examination of the spectrum of cargo relying on the cargo receptor Erv14 suggested that this large and non-homogenous group is recognized on the basis of the length of its TMD. By assessing this hypothesis using one particular cargo, Mid2, we found that TMD length is a major determinant for allowing Erv14 to accelerate exit from the ER. It is still feasible that Erv14 recognizes a more specific motif in Mid2 adjacent to the TMD, and alterations in the length of the TMD affect the position of this region relative to the bilayer and thus prevent Erv14 binding. However, this seems unlikely given the wide range of bitopic and polytopic proteins that are affected when Erv14 is deleted, and their lack of shared sequence motifs. Interestingly, there have been two recent studies reporting that shortening the TMD of a mammalian protein reduces its exit rate from the ER [18], [19]. The mechanism for these effects was not determined, but one reporter used was VSV-G that has been found to depend on the Erv14 paralogue, CNIH4, for normal ER exit [42]. Whether Erv14 enables exit of long TMD-containing proteins from the ER by performing more than just COPII coupling is yet to be uncovered. One option is that it could also act as a chaperone to protect protruding hydrophobic residues on cargo proteins thus enabling them to assume a correct conformation in the shorter ER membranes. Another option is that it sorts long TMD containing proteins into areas of the ER that have thicker membranes thereby enabling their recruitment to vesicles. Regardless, it seems that speed of ER exit may be a major determinant in Erv14' s function. In yeast, one of the substrates, Axl2, has been shown to require very rapid ER exit [26], as it must be inserted into the forming yeast bud at a particular point in the cell cycle (AXL2 mRNA is under cell-cycle control) [43]. One of the substrates for the Drosophila paralogue of Erv14, Cornichon, is Gurken, a TGFα-like bitopic protein [44]–[46]. The ability of Gurken to polarize Drosophila oocytes depends on its rapid exit from a restricted region of the oocyte ER following translation from a pool of mRNA that is spatially restricted for a short time during development [47], [48]. Indeed, also the action of Cornichon on Gurken requires that the latter has a TMD [47], [49]. Our analysis of the substrate recognition mode for Erv14 reveals that TMD length-dependent sorting may be a more general principle in cellular trafficking than previously appreciated. Using our DxE-containing Mid2 variant we noticed that the TMD variants also underwent a TMD length-dependent sorting in the Golgi apparatus. This is consistent with a previous study examining the effects of lengthening the TMD of the yeast ER protein Ufe1, although this is the first time it has been shown to occur with a homogenous synthetic TMD rather than a native TMD, which may contain additional cryptic sorting motifs [50]. How might length-dependent sorting occur in the Golgi if it does not involve Erv14? One option is that a dedicated cargo receptor exists at this compartment that has not yet been identified. However, an alternative option is that the vesicle composition itself plays a major role in this step with lipids and/or cargo proteins directing a change in bilayer properties [16], [51]–[53]. In summary, our unbiased approach allowed the formulation of a simple hypothesis for the underlying commonality allowing cargo identification by Erv14. Using Mid2 with a synthetic TMD has allowed us to indeed observe such TMD length-dependent steps both in the ER and the Golgi. The notion that TMD length is used by the cell to sort proteins is appealing [51], [53], [54], since many and diverse membrane proteins must be continuously extracted from the ER following synthesis. If these proteins share a generic feature that reflects their normal environment being different to that present in the ER, in this case TMD length, then it would provide a simple means of sorting of many different proteins without the need for specific linear signals. More generally, the conceptual methodology that we have put forward here could be applied in a wider context to uncover protein localization changes that occur in the absence of any specific gene in the genome. The notion of the “traffickome” could be extended to other trafficking events such as retrograde Golgi to ER traffic, Golgi to plasma membrane traffic, or Golgi to vacuole traffic. Hence, by pairing high-throughput genetic manipulations with a microscopic output it is now possible to study basic questions of specificity and promiscuity in cell biology that have previously been difficult to tackle.
Cultures were grown at 30°C in either rich medium (1% Bacto-yeast extract [BD], 2% Bacto-peptone [BD], and 2% dextrose [Amresco] or synthetic [S] minimal medium [0. 67%] yeast nitrogen base without amino acids [Conda Pronadisa) ] and 2% dextrose) containing the appropriate supplements for plasmid selection. When necessary, dextrose was replaced by galactose (2%; Amresco) or raffinose (2%; Amresco). For galactose induction, overnight cultures in SD, SD-LEU, or SD–URA cells were diluted 1/10 and grown at 30°C to early log phase in SD or SD–URA medium, then washed and resuspended in 2% galactose-containing SG or SG–URA medium for 2 h. For pulse-chase lysates, the first time point was obtained directly from the galactose culture. Cells were resuspended in glucose-containing medium for chase time points. When needed as selection markers, G418 (200 µg/ml; Calbiochem) or Nourseothricin (Nat) (200 µg/ml WERNER BioAgents) were added. In cases where G418 was required in a SD-based medium, yeast nitrogen base without ammonium sulfate (Conda Pronadisa) was added and supplemented with mono-sodium glutamate (Sigma) as an alternative nitrogen source. Manipulations of plasmid DNA were performed in Escherichia Coli strains DH5α and TOP10. A complete list of plasmids used in this study can be found in Table S2. All yeast strains in this study are based on the BY4741 laboratory strain [55]. General laboratory strains and strains created in this study are listed in Table S3. Unless otherwise stated, strains harboring a deletion in a specific ORF were taken from the yeast deletion library [11], while strains harboring a hypomorphic allele of an essential gene were taken from the DAmP library [24]. Strains harboring an ORF endogenously tagged with GFP in its C terminus were taken from the yeast GFP library [15]. Genomic modifications and introduction of plasmid DNA were done as previously described [56]. YMS792, YMS793, and YMS954 were created by targeting the erv14, erv15, and emp24 genes, respectively, for disruption with the kanR gene with pFA6a-KanMX6 [57]. MID2 was cloned by PCR in frame with a GAL1 promoter and monomerized GFP (A207K) into a modified version of pRS416 between HinDIII and Xho I. AclI, SpeI, and BglII sites were introduced into the MID2 sequence to facilitate cloning of overlapping oligonucleotides encoding polyleucine stretches to replace the Mid2 TMD. Integration plasmids to express TMD chimeras of Mid2-EGFP under the GAL1 promoter or MID2 promoter were constructed as follows: MID2 promoter-NatMX-GAL1 promoter-MID2 (chimera) -EGFP-MID2 terminator - in pBluescriptII (KS-). Homologous recombination was performed using a unique SnaB1 site in the MID2 gene for expression from the endogenous promoter, or using the unique sites HpaI and Blp1 for expression from the GAL1 promoter. All constructs were sequenced. All genetic manipulations were performed using the Traffo method for transforming yeast strains [56], and deletions were verified using check PCR to assay for loss of the endogenous gene copy. For a complete list of primers used see Table S4. For this work we assembled two “mini libraries” by choosing strains of interest from the above commercially available yeast libraries. First we chose 379 strains that represent a variety of possible cargo (the cargo library) from the GFP library in which each ORF is C-terminally tagged with GFP, thus enabling the visualization of the sub-cellular localization of a protein under control of its own promoter [15]. To assemble the library, we hand-picked all possible cargo proteins—those which had been visualized as being localized to either the plasma membrane, Golgi apparatus, vacuolar membrane, vacuolar lumen, COPI vesicles, COPII vesicles, peroxisomes, adiposomes, or endosomes. In addition, we added all proteins that had an undefined punctate localization (a full list of selected cargo strains is available in Table S5). The initial array was visualized and only strains displaying a strong and correctly localized GFP signal were put into the final array. The second library contained strains mutated in ER to Golgi trafficking proteins (the trafficking library), either from the yeast deletion library that contains deletions of all non-essential proteins [11] or from the DAmP library that contains hypomorphic alleles of the essential ones (for a full list of strains included see Table S6) [24]. All genetic manipulations were performed using SGA techniques to allow efficient introduction of a trait (mutation or marker) into systematic yeast libraries. SGA was performed as previously described [13], [14], [23], [58]. Briefly, using a RoToR bench-top colony arrayer (Singer Instruments) to manipulate libraries in high-density formats (384 or 1,536), haploid strains from opposing mating types, each harboring a different genomic alteration, were mated on rich media plates. Diploid cells were selected on plates containing all selection markers found on both parent haploid strains. Sporulation was then induced by transferring cells to nitrogen starvation plates. Haploid cells containing all desired mutations were selected for by transferring cells to plates containing all selection markers alongside the toxic amino acid derivatives canavanine and thialysine (Sigma-Aldrich) to select against remaining diploids. Each SGA procedure was validated by inspecting representative strains for the presence of the GFP-tagged cargo and for the correct genotype using check PCR (primer sequences can be found in Table S4). Microscopic screening was performed using an automated microscopy set-up as previously described [14]. Briefly, cells were moved from agar plates into liquid 384-well polystyrene growth plates using the RoTor arrayer. Liquid cultures were grown overnight in SD medium, with appropriate auxotrophic selections where applicable, in a shaking incubator (LiCONiC Instruments) in 30°C. A JANUS liquid handler (Perkin Elmer), which is connected to the incubator, was used to back-dilute the strains into plates containing the same medium, after which plates were transferred back to the incubator and were allowed to grow for 3. 5 h at 30°C to reach logarithmic growth. The liquid handler was then used to transfer strains into glass bottom 384-well microscope plates (Matrical Bioscience) coated with Concanavalin A (Sigma-Aldrich) to allow formation of a cell monolayer. Wells were washed twice in medium to remove unconnected cells and plates were transferred into an automated inverted fluorescent microscopic ScanR system (Olympus) using a swap robot (Hamilton). The ScanR system is designed to allow auto focus and imaging of plates in 384-well format using a 60× air lens and is equipped with a cooled CCD camera. Images were acquired at excitation at 490/20 nm and emission at 535/50 nm (GFP). After acquisition images were manually reviewed using the ScanR analysis program. Images were processed by the Adobe Photoshop CS3 program for slight contrast and brightness adjustments. Manual Microscopy was performed using either one of two systems: for Figures S3 and S4 we used an Olympus IX71 microscope controlled by the Delta Vision SoftWoRx 3. 5. 1 software with ×100 oil lens. Images were captured by a Phoetometrics Coolsnap HQ camera with excitation at 490/20 nm and emission at 528/38 nm (GFP) or excitation at 555/28 nm and emission at 617/73 nm (mCherry/RFP). Images were transferred to Adobe Photoshop CS3 for slight contrast and brightness adjustments. For Figures 4–8 we used a 100×1. 49 NA objective on a Nikon Eclipse TE2000 epifluorescent microscope using a CCD camera (CoolSNAP-HQ2, Roper Scientific) and RFP and GFP filters (Chroma Technology). Images were acquired and analyzed using MetaMorph and ImageJ, and normalized using Adobe Photoshop. For some co-localization studies with Golgi markers both channels were imaged simultaneously using a beam splitter (Cairn Research). For fusions expressed under the MID2 promoter strains were grown in synthetic complete medium to reduce background fluorescence. For protein purification during the galactose-induced pulse chases we first added 3 OD600 of cells to NaN3 (t = 0). For subsequent time points 1 ml of cells were collected. Cells were resuspended in 500 µl NaOH solution (0. 2 M NaOH, 0. 2% β-mercaptoethanol) and precipitated in 5% trichloroacetic acid. Pellets were resuspended in sample buffer and 10 µl Tris base. After electrophoresis on 4%–20% gradient gels (Novex, Invitrogen), immunoblots were blotted with mouse anti-GFP (7. 1/13. 1, Roche), HRP anti-mouse, and ECL (Amersham). For quantitation purposes, gel lane profiles were obtained from scanned autoradiograms and peak areas were determined using ImageJ. The ratio of the ER peak to the sum of the ER, post-ER, and free-GFP peaks was calculated and normalized so that it was 1. 0 at the start of the chase (t = 0). | All cells sense their environment, respond to it, and communicate with neighboring cells. To perform these functions, cells use an impressive array of proteins that they display on their surface membranes and secrete into their external environment. Newly synthesized proteins destined for the surface of nucleated cells, or to be secreted into the environment must enter the secretory pathway through the endoplasmic reticulum. Those that reside there remain behind, but most leave for their next destination as cargo proteins in lipid vesicles. To be packaged into vesicles, many of them require a “cargo receptor, ” which recognizes and tethers specific cargo proteins in the vesicles. Our study takes a systematic approach to identify the range of cargo proteins that bind to each of the known receptors in yeast. By using this approach, we both discover new cargo for known cargo receptors and delineate the rule that governs cargo selection for one cargo receptor, Erv14. Thus, our study demonstrates a novel approach to identify the cargo for any receptor or to discover new cargo receptors. | Abstract
Introduction
Results
Discussion
Materials and Methods | systems biology
biochemistry
genetics
biology
proteomics
molecular cell biology
genetics and genomics | 2012 | A Systematic Approach to Pair Secretory Cargo Receptors with Their Cargo Suggests a Mechanism for Cargo Selection by Erv14 | 9,926 | 246 |
Chagas Disease (CD) is an anthropozoonosis caused by Trypanosoma cruzi. With complex pathophysiology and variable clinical presentation, CD outcome can be influenced by parasite persistence and the host immune response. Complement activation is one of the primary defense mechanisms against pathogens, which can be initiated via pathogen recognition by pattern recognition molecules (PRMs). Collectin-11 is a multifunctional soluble PRM lectin, widely distributed throughout the body, with important participation in host defense, homeostasis, and embryogenesis. In complex with mannose-binding lectin-associated serine proteases (MASPs), collectin-11 may initiate the activation of complement, playing a role against pathogens, including T. cruzi. In this study, collectin-11 plasma levels and COLEC11 variants in exon 7 were assessed in a Brazilian cohort of 251 patients with chronic CD and 108 healthy controls. Gene-gene interactions between COLEC11 and MASP2 variants were analyzed. Collectin-11 levels were significantly decreased in CD patients compared to controls (p<0. 0001). The allele rs7567833G, the genotypes rs7567833AG and rs7567833GG, and the COLEC11*GGC haplotype were related to T. cruzi infection and clinical progression towards symptomatic CD. COLEC11 and MASP2*CD risk genotypes were associated with cardiomyopathy (p = 0. 014; OR 9. 3,95% CI 1. 2–74) and with the cardiodigestive form of CD (p = 0. 005; OR 15. 2,95% CI 1. 7–137), suggesting that both loci act synergistically in immune modulation of the disease. The decreased levels of collectin-11 in CD patients may be associated with the disease process. The COLEC11 variant rs7567833G and also the COLEC11 and MASP2*CD risk genotype interaction were associated with the pathophysiology of CD.
Chagas Disease (CD) is a neglected anthropozoonosis in which classically the primary infection with the protozoan Trypanosoma cruzi, transmitted by blood-sucking bugs, can occur during early childhood and may continue clinically silent for decades [1,2]. Approximately 30% of chronically infected individuals develop cardiac and/or digestive alterations, however, the majority of infected individuals remains asymptomatic [1,3]. In the last decades, increasing migration from endemic to non-endemic countries resulted in altered epidemiological scenarios, turning CD into a global health concern [4,5]. The complex pathophysiology of CD is influenced by several factors, in particular the parasite’s genetic variability and the degree of host immune response, both playing a critical role in the disease outcome [3,6]. Parasite persistence is dependent on its ability to evade the host defense mechanisms. Here, host genetic background plays an important role in infection establishment and clinical presentation of CD [7–9]. The complement system is a central component of the innate immune response and one of the first line of defense against pathogens, in which carbohydrates or acetylated patterns on the pathogen surface are recognized by pattern recognition molecules (PRMs), such as lectins [10]. The initiators of the lectin pathway are ficolins (ficolin-1, ficolin-2 or ficolin-3) and collectins (mannose-binding lectin—MBL—and collectin-11 also known as collectin kidney 1—CL-11 alias CL-K1) [11]. Deficiency in the components of this pathway can critically impact the immune competence and therefore may lead to susceptibility to infectious diseases [12,13]. Moreover, the genetic variation on collectins may alter protein structure and, thereby affecting their ability to recognize parasites, including T. cruzi, contributing to parasite persistence. Indeed, genetic variation in PRMs of the lectin pathway has been associated with disease establishment and clinical progression of CD [13–15]. The infective T. cruzi metacyclic trypomastigote has a broad range of carbohydrates on its surface, including mannose, N-acetyl-D-glucosamine, galactose, and fucose on glycosylated proteins [16]. These glycoconjugates act as pathogen-associated molecular patterns (PAMPs), allowing the PRMs to interact with them [9,17]. Initially, the collectins associated with MBL-associated serine proteases (MASPs) bind to glycosylated molecules on the surface of T. cruzi in the presence of Ca2+ activating the proteolytic cascade [18,19]. This cascade carries forward the activation of complement, which may also result in the elimination of pathogens [10,20]. Collectin-11 is a multifunctional soluble PRM lectin with important participation in host defense, homeostasis, and embryogenesis [19,21,22]. It is expressed by a wide range of tissues, with the adrenal glands, kidneys and liver being the sites of highest abundance [10,23]. Recently, collectin-11 has been found to circulate in the form of a heteromeric complex with collectin-10 [23]. Collectin-11 has binding affinity to sugars such as fucose, mannose, N-acetyl-D-galactosamine, and N-acetyl-D-glucosamine [21–23]. Similar to MBL, the monomer is composed of a collagen-like domain and a carbohydrate recognition domain (CRD), linked by a helical neck [10,24]. The gene encoding collectin-11, COLEC11, is located on chromosome 2p25. 3 (OMIM 612502) and comprises 7 exons that transcribe the canonical protein [23]. COLEC11 variability was shown to interfere with expression and also with the binding of calcium and carbohydrates, possibly affecting protein folding [23]. Three distinct genetic variations in exon 7 of COLEC11 have been associated to the 3MC (Carnevale, Mingarelli, Malpuech, and Michels) developmental syndromes [25]. Two variations result in single amino acid substitutions, p. Ser169Pro and p. Gly204Ser, and the third in a deletion (p. Ser217del). All variations alter the primary structure of the CRD [24]. Homozygous individuals for the mutation p. Gly204Ser do not present detectable collectin-11 in serum [24]. Moreover, the variant p. His219Arg (rs7567833A>G) in exon 7 was associated with higher prevalence of urinary schistosomiasis [26]. Collectin-11 shows a strong binding affinity to fucose-proteins [27], as found in the Tc-85 protein family, expressed on the surface of T. cruzi metacyclic trypomastigotes. Those proteins are involved in the entry of the parasite to host cells [28]. In addition, collectin-11 is structurally similar to MBL and it has been shown that both MBL levels and MBL2 genetic variants were associated with disease susceptibility and pathophysiology of CD [29]. Considering these observations, collectin-11 plasma levels and COLEC11 variants in exon 7 were assessed to investigate their potential role in the chronic CD. Moreover, on account of the interaction between collectin-11 and MASPs for complement activation, gene-gene interaction between COLEC11 and MASP2 was assessed to evaluate the additive genetic effect of the two loci and their role in the pathophysiology of this chronic disease.
A cohort of 251 patients with chronic CD attending the outpatient department for Chagas Disease of Hospital de Clínicas, Federal University of Paraná, was investigated [mean age 57 years; 140 (56%) females, 111 (44%) males, 190 (75. 7%) Euro-, 48 (19. 1%) Afro-Brazilian, 2 (0. 8%) Asian, 11 (4. 4%) Amerindian]. CD serodiagnosis was performed utilizing two serological tests: ELISA (Architect Plus Chagas, Abbott, Illinois, USA—sensitivity: 100%; 95%CI 97. 90–100% and specificity: 99. 93%; 95%CI 99. 80–99. 99%) and indirect immunofluorescent (IMUNO-Con Chagas, WAMA diagnóstica, São Paulo, Brazil (sensitivity and specificity 100%) assays. Clinical assessments were obtained through medical records and interviews, whereas patients younger than 18 years old, with recent infection, or suspected non-chagasic cardiomyopathy were excluded. Ancestry was self-referred by the patient in the first interview. Demographic and clinical characteristics of the distinct CD forms are shown in Table 1. Patients with cardiomyopathy were graded according to the cardiac insufficiency classification of the American Heart Association, adapted for CD [30]: A, altered electrocardiogram (ECG) and normal echocardiogram (ECHO), absence of cardiac insufficiency (CI); B1, altered ECG, left ventricular ejection fraction (LVEF) > 45%, absence of CI; B2, altered ECHO, LVEF < 45%, absence of CI; C, altered ECG and ECHO, compensable CI; D, altered ECG and ECHO, refractory CI. The digestive forms of Chagas disease were identified by alterations in esophagography and barium enema radiological exams, used to diagnose megaesophagus and/or megacolon. Chronic asymptomatic individuals (with the indeterminate form) presented reactive serology and/or positive parasitological examination for T. cruzi but did not present clinical symptoms specific to CD and had normal results of ECG and radiological chest, esophagus and colon exams [30]. A total of 108 healthy Brazilians [mean age 51 years; 52 (48. 1%) females, 56 (51. 9%) males, 95 (88%) Euro-, 10 (9. 3%) Afro-Brazilian, 2 (1. 8%) Asian, 1 (0. 9%) Amerindian] was used as control group. All individuals from the control group were selected consecutively from a blood bank in the same geographic region as patients with chronic CD. Following Brazilian health regulations, the blood donors were screened for CD, syphilis, hepatitis B, hepatitis C, HIV and human T-cell lymphotropic viruses 1 and 2 using high sensitivity assays. Additionally, self-referred ancestry and information about autoimmune diseases and cancer background was obtained during the pre-selection interview. The study protocol was approved by the local Ethics Committee (CEP/HC-UFPR n. 360. 918/2013-08), and all adult patients and controls provided written informed consent on their behalf in accordance with the Declaration of Helsinki. No children were enrolled in this study. Collectin-11 plasma levels were determined in 233 patients and 102 controls using a commercial high-sensitivity ELISA kit [Human Collectin-11 (COLEC11) /abx517452, Abbexa Ltd, Cambridge, UK] in accordance with the manufacturer’s instructions. The limit of detection was 78 pg/ml. Plasma from 18 patients and six controls was not available. In total, 186 patients and 95 controls had overlapping samples between the genetic and ELISA analysis. Additionally, protein levels of C-reactive protein (CRP) [13,31], pentraxin 3 (PTX3) [31], MASP2 [14] and complement receptor 1 (CR1) [32] generated by previous studies in the same cohort were used for correlation analysis with collectin-11. In order to assess the distribution of the three COLEC11 variants in exon 7 (Fig 1), rs148786016G>A (g. 3643816G>A, p. Gly172Ser), rs7567833A>G (g. 364395A>G, p. His219Arg) and rs114716171C>T (g. 3644079C>T, p. Thr259 =), the entire COLEC11 exon 7 including its intron-exon boundaries was directly sequenced in 204 patients with chronic CD and 101 healthy control individuals. DNA from 47 patients and seven controls could not be isolated in sufficient amount; therefore, these individuals were excluded from further genetic analyses. Genomic DNA was extracted from buffy-coats using the QIAamp Blood mini kit (Qiagen GmbH, Hilden, Germany) following the manufacturer' s instructions. The COLEC11 reference sequence (ENST00000349077. 8) was retrieved from the Ensembl database (www. ensembl. org), primers targeting exon 7 of COLEC11 gene were those utilized by Antony et al. [6] and were synthesized commercially (Eurofins Genomics, Ebersberg, Germany). PCR amplifications were carried out in a 25 μl volume of reaction mixture containing 10x PCR buffer, 2 mM MgCl2,0. 125 mM of dNTPs, 0. 2 μM of each primer, 1 unit of Taq polymerase (Qiagen, Germany), and 20 ng of genomic DNA on a Mastercycler Nexus Gradient (Eppendorf, Germany). Cycling parameters were initial denaturation at 94°C for 3 minutes, followed by 35 cycles of denaturation at 94°C for 30 seconds, annealing at 59. 8°C for 30 seconds and elongation at 72°C for 1 minute, and a final elongation step at 72°C for 10 minutes. PCR fragments were stained with SYBR Safe DNA Gel Stain (Invitrogen, Carlsbad, USA) and visualized in a 1. 5% agarose gel. PCR products were purified using Exo-SAP-IT (USB-Affymetrix, Santa Clara, USA) and the purified products were directly used as templates for sequencing using the BigDye terminator cycle sequencing kit (v. 3. 1; Applied Biosystems, Texas, USA) on an ABI 3130XL DNA Analyzer. DNA polymorphisms were identified by assembling the sequences with the reference sequence of the COLEC11 (ENST00000349077. 8) using the Geneious v11. 0. 3 software (Biomatters Ltd, Auckland, New Zealand) and reconfirmed visually from their respective electropherogram. Previously assessed MASP*CD genotypes from the same cohort group were retrieved from a previous study performed by our research group and utilized in the gene-gene interaction analysis [14]. In silico analysis of possible functional effects of rs7567833 (p. His219Arg) on protein function/structure was performed. The SIFT tool (Sorting Intolerant from Tolerant) is a multi-step sequence alignment comparison algorithm, which infers whether an amino acid substitution may have an impact on protein function considering the premise that highly conserved amino acids are more intolerant to substitution than those less conserved (http: //sift. bii. a-star. edu. sg/) [33]. PolyPhen-2 utilizes a trained Naive Bayes classifier to evaluate physical and comparative considerations to predict the functional significance of a mutation on the structure and function of a protein (http: //genetics. bwh. harvard. edu/pph2/) [34]. Ensembl Variant Effect Predictor (VEP) infers the effect of variants on protein sequence using SIFT and PolyPhen-2 predictions in the extensive collection of genomic annotation of Ensembl database (http: //www. ensembl. org/vep) [34–36]. SNAP2 is a neural network-based classifier that utilizes a backpropagation algorithm resulting in predictions regarding the functionality of mutated proteins (https: //www. rostlab. org/services/snap/) [37–39]. Combined annotation dependent depletion (CADD) is a tool for scoring the deleteriousness of single nucleotide variants in the human genome (https: //cadd. gs. washington. edu/snv) [40]. Collectin-11 plasma levels were tested for normality using Shapiro-Wilk and compared between groups using nonparametric Kruskal-Wallis and Mann-Whitney tests using GraphPad Prism software (version 5), with dispersion graphics displaying median and percentiles values. For all the analysis, CD patients were compared among the clinical forms as indeterminate/asymptomatic, cardiac (A+B1/2+C+D groups), digestive, and cardiodigestive, and also grouped as symptomatic patients (cardiac + digestive + cardiodigestive forms). Also, patients with cardiac form were grouped as with cardiomyopathy (B2+C+D), without ECHO alterations (A), with ECHO alterations (B1/2+C+D), without heart failure (A+B1/2) and with heart failure (C+D). Multiple logistic regression was executed in a multivariate model using a backward selection including variants with p<0. 20 in the univariate analysis. Age, sex, and ethnic group were always included as covariables (age as a continuous covariable). Significant p values were corrected using Benjamini-Hochberg method. Continuous data were described as means, and categorical variables were presented as numbers and percentages. Odds ratios (OR) and their 95% confidence intervals (95% CI) were calculated using the STATA software (v. 12. 0, StataCorp, College Station, Texas, USA). Correlation analysis was performed by non-parametric Spearman´s rank test. Genotype and allele frequencies were obtained by direct counting. Haplotype was estimated by expectation-maximum algorithms and the significance of deviation from Hardy-Weinberg equilibrium was tested using the random-permutation procedure as implemented in the Arlequin v. 3. 5. 2. 2 software (http: //cmpg. unibe. ch/software/arlequin35/). Linkage disequilibrium (LD) between COLEC11 polymorphisms was measured by the relative LD coefficient (D’) and the correlation coefficient (r2) in the Arlequin v. 3. 5. 2. 2 software (http: //cmpg. unibe. ch/software/arlequin35/). Possible associations of COLEC11 alleles, genotypes, and haplotypes with different clinical forms were evaluated with two-tailed Fisher exact tests. Gene-gene interaction between COLEC11 (rs7567833A>G, p. His219Arg) and MASP2 (rs17409276, g. 1961795C>T and rs12711521C>A, p. D371Y) variants were calculated using a two-stage strategy for identifying relevant interactions [41]: (i) calculation of the additive effect among COLEC11 and MASP2 with two-tailed Fisher exact tests and (ii) validating the gene interactions using the Model-Based Multifactor Dimensionality Reduction (MB-MDR) adjusted model, considering a conservative risk threshold of 0. 1 and 100 permutations assessed with the mbmdr package of R Studio (www. r-project. org) [42,43]. P-values <0. 05 were considered significant (unless for genetic disease association analysis, where the threshold was corrected to p = 0. 0478, using the Benjamini-Hochberg method). A post hoc analysis to compute statistical power, given alpha (0. 05), sample size, and effect size (considering the respective odds ratio) with the software G*Power v. 3. 1. 9. 4 for Mac (http: //www. gpower. hhu. de) was performed for each significant genetic association found in this study.
The mean of collectin-11 plasma levels observed in healthy Brazilian individuals (237. 8±183. 3 ng/ml) agrees with the levels found in healthy individuals from other populations, including Nigerian (246±155 ng/ml) [26], Japanese (340±130 ng/ml) [44], and Danish (284±180 ng/ml) [45]. Collectin-11 plasma levels were significantly lower in CD patients compared to controls (p<0. 0001; 172. 5 ng/ml, 95% CI 130. 8–214. 2 vs 237. 8 ng/ml, 95% CI 201. 8–273. 8) (Fig 2). When comparing controls to each clinical form separately, statistically significant differences were also observed in collectin-11 levels between controls (n = 102) and the indeterminate form (n = 90) (p<0. 0001; 141. 0 ng/ml, 95% CI 102. 3–179. 7), cardiac form (n = 85) (p<0. 0001; 220. 2 ng/ml, 95% CI 115. 3–325. 1), digestive form (n = 24) (p = 0. 006; 156. 7 ng/ml, 95% CI 107. 5–205. 8), and cardiodigestive form (n = 34) (p<0. 0001; 149. 9 ng/ml, 95% CI 87. 07–212. 7) (Fig 2). Comparison of collectin-11 levels between asymptomatic (indeterminate form) and symptomatic (n = 143) patients, even when grouped according to each clinical form, presented no statistical difference. In addition, Collectin-11 plasma levels presented a negative correlation with LVEF index (p = 0. 0419, r = -0. 15) (Fig 3). No significant correlation was found between plasma levels of collectin-11 and protein levels of CRP, PTX3, ficolin-2, CR1, and MASP2. The distribution of COLEC11 genotypes has not violated Hardy-Weinberg equilibrium in both control (rs148786016, not applicable–monomorphic locus; rs7567833, p = 0. 38; rs114716171, p = 1. 00) and patient (rs148786016, p = 1. 00; rs7567833, p = 0. 16; rs114716171, p = 1. 00) groups, as well as in the asymptomatic group (rs148786016, not applicable–monomorphic locus, rs7567833, p = 0. 67; rs114716171, p = 1. 00). No association was found between the analyzed genetic variants and collectin-11 plasma levels. The frequency of COLEC11 variant rs7567833G (p = 0. 005; OR 2. 3,95% CI 1. 2–4. 2) was significantly higher in chronic CD patients. It also occurred more frequently among patients with the cardiodigestive form (p = 0. 002; OR 3. 9,95% CI 1. 7–8. 8), compared to controls (Table 2). Also, the frequencies of COLEC11 genotypes AG and GG of rs7567833 were significantly higher in chronic CD patients (p = 0. 028; OR 2. 2,95% CI 1. 1–4. 4) than in controls (Table 2). In addition, carriers of the G allele (AG and GG of rs7567833) were rather present among patients presenting the cardiodigestive form of CD (p = 0. 002, OR 5. 1,95%CI 1. 9–14. 2) in relation to controls (Table 2). No significant difference was found between the allelic and genotypic frequencies of controls and patients for COLEC11 variants rs148786016 and rs114716171 (Table 2). The G allele (p = 0. 006, OR 2. 5,95% CI 1. 3–4. 8) and the genotypes AG and GG of rs7567833 (p = 0. 023, OR 2. 5,95% CI 1. 1–5. 7) were more frequent in patients with cardiomyopathy than in controls (Table 3). Considering the different stages of cardiac pathology, the G allele and AG and GG genotypes of rs7567833 were significantly higher in patients with cardiomyopathy with ECHO alteration (p = 0. 01, OR 2. 5,95% CI 1. 2–4. 9; and p = 0. 03, OR 2. 5,95% CI 1. 1–5. 9, respectively) in comparison to controls. In addition, the minor allele G and carriers of the G allele (AG and GG of rs7567833) were more frequent among patients with heart failure than in controls, although not statistically significant after logistic regression. No association was found when analyzing patients presenting only pathology of the digestive tract (Table 3). In silico analysis predicted that the non-synonymous variant rs7567833A>G might have a functional impact on collectin-11 (SNAP2; score 69) with a likely deleterious effect on protein function (CADD, score 20. 5). However, this variant was predicted to present a tolerated effect on protein function by the SIFT tool, being considered benign regarding the structure and function of the protein by PolyPhen-2. The rs148786016A and rs7567833G, rs148786016A and rs114716171C, rs7567833G and rs114716171C occurred in absolute linkage disequilibrium (LD) in patients (D’ = 1); as well as rs7567833G and rs114716171C in controls. LD could not be measured between rs148786016 and rs7567833 and rs148786016 and rs114716171 in controls; since the rs148786016 was monomorphic in this group. A total of four COLEC11 haplotypes was reconstructed from the three COLEC11 variants (rs148786016, rs7567833, and rs114716171) investigated in this study. No association was found between the analyzed haplotypes and collectin-11 plasma levels. The frequency of COLEC11*GGC haplotype, carrying the rs7567833G allele, was significantly increased among CD patients (p = 0. 022, OR 2. 0,95% CI 1. 1–3. 8) and cardiodigestive form of CD (p = 0. 009, OR 3. 4,95% CI 1. 4–8. 7) in comparison to controls (Table 4). The COLEC11*GAC haplotype, carrying the rs7567833A allele, was associated with protection when compared patients (p = 0. 010, OR 0. 4,95% CI 0. 2–0. 8), asymptomatic/indeterminate (p = 0. 047, OR 0. 5,95% CI 0. 2–0. 9), symptomatic patients (p = 0. 02, OR 0. 4,95% CI 0. 2–0. 9) and those presenting cardiodigestive form (p = 0. 005, OR 0. 3,95% CI 0. 1–0. 7) with controls. A trend towards lower frequency of COLEC11*GAC haplotype was found when comparing patients with cardiac form with controls (p = 0. 05, OR 0. 5,95% CI 0. 2–1. 0). Patients with cardiomyopathy had a higher frequency of COLEC11*GGC haplotype, carrying the rs7567833G allele (p = 0. 022, OR 2. 2,95% CI 1. 1–4. 5) than controls. In addition, COLEC11*GGC was significantly associated with patients presenting ECHO alterations and with heart failure (p = 0. 044, OR 2. 2,95% CI 1. 0–4. 6; and p = 0. 033, OR 2. 5,95% CI 1. 1–5. 6 respectively) in comparison to controls (Table 5). Also, the COLEC11*GAC haplotype, carrying the rs7567833A allele, was associated with protection against cardiomyopathy (p = 0. 008, OR 0. 39,95% CI 0. 2–0. 8), presenting ECHO alteration (p = 0. 008, OR 0. 37,95% CI 0. 2–0. 8) and heart failure (p = 0. 006, OR 0. 32,95% CI 0. 1–0. 7) compared to controls. Considering the biological relevance between collectin-11 and MASP2 and that MASP2 genetic variants were associated with high risk of cardiomyopathy in chronic CD [14], the genetic interaction between COLEC11 and MASP2 variants were analyzed. For this, the combined effect of cardiac commitment risk genotypes for COLEC11 (rs7567833AG and rs7567833GG) and MASP2 (MASP2*CD carriers, g. 1961795C>A, p. D371Y) (S1 Table) in chronic chagasic cardiomyopathy was calculated. The frequency of the risk genotypes in both loci (COLEC11 AG+GG and MASP2*CD+ carriers) was higher in patients with cardiodigestive form (21%) and cardiomyopathy (13%), than healthy controls (2%), (p = 0. 005, OR 15. 2,95% CI 1. 7–137; p = 0. 014, OR 9. 3,95% CI 1. 2–74, respectively) (Table 6). As recommended for gene-gene interaction in case-control association studies, a dimension reduction method (MB-MDR) was applied to check the association of COLEC11 AG+GG and MASP2*CD genotypes with a risk phenotype for CD. With this approach, COLEC11 and MASP2 risk genotypes presented high risk interaction for CD, which remained significant even after adjustment (considering 100 permutations) for patients with cardiomyopathy when compared to controls (adjusted permutation p = 0. 05) and for patients with cardiodigestive form compared to asymptomatic but infected individuals (adjusted permutation p = 0. 04).
Pathogen recognition is a critical step in host defense against pathogens. The lectin pathway activates the complement system based on the recognition of surface microbial carbohydrate patterns by PRM such as collectin-11. This recognition can lead to pathogen lysis through the membrane attack complex formation and may support the control of the parasite burden [46]. Previous reports have shown that the PRMs MBL, ficolins, and collectin-11 can recognize and bind to specific glycoproteins on the surface of pathogens, including T. cruzi [47–49]. Association studies have also demonstrated that the lectin proteins ficolin-2 [13] and MBL [15] are involved in disease progression of chronic CD, however, these results were not yet tested in vitro or in vivo experimental models. In this study, individuals chronically infected with T. cruzi presented decreased levels of collectin-11 compared to healthy controls, however, this was not associated with the genetic variants analyzed in this study. It is important to mention that other causal variants responsible for modulating COLEC11 expression were not investigated in this study, such as rs13417396 (in intron 4), rs11895384 (intron 5), rs10185914 (intron 6), and rs10166336 (intron 6) (https: //ldlink. nci. nih. gov/). Polymorphisms in the promoter region do not appear to play a role in collectin-11 expression [26,50]. Alternatively, the lower collectin-11 levels found in patients may be due to consumption of collectin-11 during T. cruzi chronic infection. Moreover, no difference in protein levels was found between both groups indeterminate/asymptomatic and symptomatic patients, indicating that the different CD phenotypes are not directly induced by collectin-11. However, this lack of difference may be due to the limited number of patients per clinical group and/or the difficulty to detect minimal changes in asymptomatic patients using conventional medical examinations. Lower levels of collectin-11 have been associated with other infectious disease including Schistosoma haematobium infection [26] and tuberculosis [51]. In line with recent studies, collectin-11 plasma levels presented no correlation with CRP and PTX3 levels, reinforcing that it is not an acute phase protein [50]. The weak negative correlation of collectin-11 levels with LVEF (r = -0. 15, p = 0. 0419) may indicate that low levels of the protein could be associated with an increased risk of cardiac commitment in patients with chronic CD. Nevertheless, additional studies are necessary to confirm this hypothesis. The positive association of AG and GG genotypes and the G allele in variant rs7567833 observed in patients with chronic CD may be related to the functional properties of the collectin-11 molecule. Interestingly, G (the minor allele) is indeed the ancestral allele [52] and its reduction indicates that this polymorphism may have experienced selection pressures over the time [53]. Although the genetic drift resulting from human migration may be an alternative explanation. This variant (rs7567833A>G) results in an amino acid change (p. His219Arg) in the carbohydrate recognition domain of the protein which probably interferes with its binding affinity to carbohydrates and thereby alters the potential of collectin-11 to activate the lectin pathway (Fig 1) [24,52]. Indeed, collectin-11 p. His219Arg (rs7567833A>G) was predicted by in silico analysis to have a functional impact with a likely deleterious effect on protein function (SNAP2, CADD). Nevertheless, p. His219Arg did not affect collectin-11 plasma levels either in CD patients or controls, which is in agreement with the finding of Bayarri-Olmos and collaborators [52]. As seen for ficolin-2, amino acid substitutions in the pathogen recognition domain could affect the binding affinity of the variant molecule towards its ligand and thus the complement activation potential [52]. Two non-synonymous polymorphisms in FCN2 positioned near the binding site markedly alter its binding capacity [54]. Interestingly, the substitution FCN2*258S affecting the binding affinity of ficolin-2 was associated with the development of the cardiodigestive form in chronic CD [13]. This was also observed for the COLEC11 variant rs7567833A>G, where the G allele, the carriers of G allele (AG and GG genotypes) as well as the COLEC11*GGC haplotype were associated with cardiodigestive form of CD, indicating that this variant might predispose to clinical progression of chronic CD. Additionally, in a study that evaluated another C-type collectin, alleles causing MBL deficiency were associated with clinical progression of CD and MBL2 genotypes causing MBL deficiency were associated with heart damage [29]. Also, the minor allele G (rs7567833G), its genotypes (rs7567833AG and rs7567833GG) and COLEC11*GGC haplotype were associated with cardiomyopathy. Here the analyzed COLEC11 genetic variant does not lead to protein deficiency, but it may alter protein function, being associated with the development of infection and pathophysiology of CD. Nevertheless, functional studies on both p. 219His and p. 219Arg collectin-11 conformations must be performed in order to define their effect type on the interaction of collectin-11 to its ligands. It is known that collectin-11 binds to PAMPs and activates MASP-2 to initiate the activation of lectin pathway, stimulating immune processes [20]. Here, the results indicated that COLEC11 (rs7567833G>A) and the diplotype MASP2*CD (g. 1961795C>A, p. D371Y) presented gene-gene interaction. Patients carrying both risk genotypes were shown to have a 15. 2-fold increased risk of developing cardiodigestive form of CD and a 9. 3-fold increased risk of cardiomyopathy. This additive or synergic interaction may contribute to the immune modulation of the disease. Nevertheless, the increased risk of developing the cardiodigestive form should be interpreted carefully due to the low sample size in this study. Analysis of a larger population would be required to confirm the role of this genetic interaction. The mechanisms by which these two genes interact with each other in the pathophysiology of CD is not clear; but interplay of both proteins, collectin-11 and MASP-2, occurs during activation of the lectin pathway. In addition, previously, results showing that MASP2*CD genotypes are associated with high risk of CD cardiomyopathy [g. 1961795C, p. 371D diplotype was more frequent in symptomatic patients (p = 0. 012, OR 3. 11) as well as in patients with cardiomyopathy (p = 0. 012, OR 13. 53) compared to asymptomatic patients] [14], corroborates these results. This is the first study analyzing the impact of gene-gene interaction in markers of innate immunity in CD. The combined genetic analysis used in this study may provide further insight into the complex pathogenesis of this disease. The low number of patients in some groups, especially those with the cardiodigestive form, presents a limitation for this study and is partly due to the unequal distribution and stratification of the patients according to the different clinical forms. This may affect the statistical power by reducing it (<70%) (S2 Table), requiring careful interpretation of the results for the clinical forms, especially the cardiodigestive form. For these reasons, more studies, including analysis of a larger population and functional approaches, are necessary to understand better the role of collectin-11 in the pathophysiology of CD. In addition, the ancestry was self-referred by the participant/patient, which result in bias regarding the ancestry data. Nevertheless, the fact that the same results were reproduced in different comparisons, leads us to suggest that the associations are indeed reliable. In conclusion, this study reports that the analyzed COLEC11 variants and collectin-11 levels are associated with T. cruzi infection. Nevertheless, the decreased collectin-11 levels were not associated with the studied polymorphisms and may be related to the disease process. COLEC11 rs7567833G and MASP2*CD risk genotype may act synergistically increasing the risk of developing chagasic cardiomyopathy. This pioneering study provides insights on the role of collectin-11 and also on combinational genetic analysis (COLEC11 and MASP2) of two initiators of the complement response in the clinical presentation of chronic CD. Future functional studies are required to unveil the interaction of collectin-11 with T. cruzi as well as to investigate the additive/synergic effect of COLEC11 and MASP2 genes in the development and clinical expression of CD. | The heterogeneity of clinical progression during chronic Trypanosoma cruzi infection and the mechanisms determining why some individuals develop symptoms whereas others remain asymptomatic are still poorly understood. The pathogenesis of chronic Chagas Disease (CD) has been attributed mainly to the persistence of the causing parasite and the character of individual host immune responses. Collectin-11 is a host immune response molecule with affinity for sugars found on the T. cruzi’s surface. Together with mannose-binding lectin-associated serine proteases (MASPs), it triggers the host defense response against pathogens. Genetic variants and protein levels of MASP-2 and the mannose-binding lectin (MBL), a molecule structurally similar to collectin-11, have been found to be associated with susceptibility to T. cruzi infection and clinical progression to cardiomyopathy. This prompted us to investigate collectin-11 genetic variants and protein levels in 251 patients with chronic CD and 108 healthy individuals, and to examine the effect of gene interaction between COLEC11 and MASP2 risk mutations. We found an association to CD infection with COLEC11 gene variants and reduced collectin-11 levels. The concomitant presence of these genetic variants and MASP2 risk mutations greatly increased the odds for cardiomyopathy. This is the first study to reveal a role for collectin-11 and COLEC11-MASP2 gene interaction in the pathogenesis of CD. | Abstract
Introduction
Methods
Results
Discussion | cardiomyopathies
medicine and health sciences
variant genotypes
tropical diseases
parasitic diseases
parasitic protozoans
genetic mapping
protozoans
neglected tropical diseases
cardiology
proteins
protozoan infections
trypanosoma cruzi
biochemistry
haplotypes
trypanosoma
eukaryota
lectins
chagas disease
heredity
genetics
biology and life sciences
human genetics
genetics of disease
organisms | 2019 | Human collectin-11 (COLEC11) and its synergic genetic interaction with MASP2 are associated with the pathophysiology of Chagas Disease | 9,139 | 331 |
Loa loa infection is endemic in limited areas of West-Central Africa. Loiasis has been associated with excess mortality, but clinical studies on its treatment are scant, particularly outside endemic areas, due to the rarity of cases diagnosed. With this retrospective TropNet (European Network for Tropical Medicine and Travel Health) study, we aimed at outlining the treatment schedules followed by different reference centers for tropical medicine across Europe. We gathered information about 238 cases of loiasis, 165 of which had follow up data. The regimens followed by the different centers were heterogeneous. The drugs most frequently administered were: diethylcarbamazine alone (74/165,45. 1%), ivermectin alone (41/165,25%), albendazole + ivermectin (21/164,11. 6%), ivermectin + diethylcarbamazine (16/165,9. 7%). The management of loiasis substantially differs across specialized travel clinics in Europe. These discrepancies could be due to different local protocols as well as to (un) availability of the drugs. An harmonization of clinical protocols for the treatment of loiasis would be suggested across reference centers for tropical medicine in Europe.
Loa loa is a filarial worm transmitted to the human hosts by the tabanid flies of the genus Chrysops. Human loiasis occurs only in Africa, where the transmission is confined to the rainforests areas from south-eastern Benin in the west, South Sudan and Uganda in the east, North Angola in the south[1]. Adult worms moving in subcutaneous tissues can live for more than 15 years: moreover, after a prepatent period of a minimum of six months, they can produce microfilariae (mff), circulating in peripheral blood with a diurnal periodicity[2]. The most common clinical manifestations due to adult worms and/or mff are" Calabar swelling" (transient angioedema of allergic nature) and pruritus. Moreover, adult worms may be noticed when they pass under the conjunctiva of the eye (" eyeworm" ). Albeit rarely, loiasis can cause damage to other organs[2]. Loiasis had been considered as a benign disease until a retrospective study recently demonstrated an excess of mortality in patients with high microfilaremia[3], and advocated more studies on this disease[4]. Currently three drugs may be used for the treatment of loiasis: diethylcarbamazine (DEC), ivermectin (IVM) and albendazole (ALB). DEC is the drug of choice because of its macro- and microfilaricidal activity, that causes a rapid decrease in the Loa loa microfilaremia, although sometimes multiple courses of DEC are required to achieve clinical and parasitological cure[5]. DEC is considered as contra-indicated in patients with a high microfilarial density (>8. 000 mff/ml), because of the risk of encephalopathy[6]. Moreover, this drug is not currently available in Europe[7]. In fact, in an inquiry involving 69 TropNet (European Network for Tropical Medicine and Travel Health) centers in Europe, DEC was immediately available in 25 centers, available within a few days in 11, and not available in 33[8]. Ivermectin has a marked microfilaricidal effect (Loa microfilaremia decreases by 70–80% within the first 3 days after a single dose of 150 μg/kg) [9], but is probably not active on macrofilariae[6]. Besides this, this drug should be administered with caution in case of microfilaremia > 8,000/ml and can also induce an encephalopathy in people with very high microfilarial densities (>30,000/ml), contra-indicating its use above that level[6,10]. Short courses of ALB have little effect on Loa loa[11,12], but when given at a dose of 200 mg twice a day for 21 days, the drug has probably an embryotoxic effect (i. e. , it interrupts embryogenesis in the uterus of the adult female worm), and possibly also a macrofilaricidal effect[13]. This being said, the treatment strategy depends firstly on the risk of adverse events, which is related to the patient' s Loa microfilarial density. In fact, given the risk of serious adverse events after DEC or IVM treatment, Boussinesq proposed the following strategy: ALB for microfilarial loads higher than 8,000/mL, followed by IVM when microfilarial density is between 2,000 and 8,000/mL, DEC when microfilarial density is below 2,000/mL[6]. Two series published in 1986[14] and in 1996[15], reported that a single course of treatment with DEC for 21 days achieves a cure rate of 55% and 66%, respectively. Four recent papers, a case series of 47 imported cases in France[16], a review of 101 cases reported in non-endemic countries[17], a case series of 100 cases in a single centre in Italy[18], and a case series of 50 cases in London[19], revealed a wide heterogeneity of treatment regimens and follow -up patterns over the last three decades across Europe, highlighting that the management of imported loiasis needs standardization. A randomized controlled trial (RCT) comparing different drugs in non-endemic countries, is problematic, due to the relatively small number of cases diagnosed per travel clinic and the differences in drug availability between countries. The primary objective of this study was to describe the different drugs regimens used for imported loiasis in different TropNet sites. Secondary objectives include the description of the treatment outcome and tolerability of the drugs used.
The study protocol was submitted to the Ethics committee of the coordinating centre (Comitato Etico per la Sperimentazione Clinica delle Province di Verona e Rovigo), and obtained a waiver of informed consent on the 13th July 2016. All individuals who had been traveling or living in an endemic country AND were diagnosed with loiasis (according to the case definition given below) AND were treated with either DEC alone, IVM alone, ALB alone, ALB + DEC, IVM + DEC or ALB + IVM (at the dosages specified below) AND had at least one follow up visit ≥ 1 month after treatment AND had not traveled to any endemic country before the last follow up evaluation. Eligible drug regimens for patient inclusion: DEC, 6 mg/kg/day for 21 days; IVM, 150–200 μg/kg as a single dose; ALB, 200–400 mg twice a day for 21–28 days. All patients treated with any other drug regimen. Patients retreated after first treatment failure. A case of loiasis was defined in this study as the presence -or history (in the last two months) —of eyeworm OR the demonstration of Loa loa microfilaremia OR the presence -or history (in the last two months) - of a Calabar swelling associated with eosinophilia (defined as > 450 eosinophils/μL). Clearance of symptoms was defined as the lack of re-occurrence of the clinical signs/symptoms present at inclusion (Calabar swellings, eyeworm) in the time frame going from treatment to the last available follow up visit. Parasitological outcome was assessed on the basis of values of microfilaremia and eosinophil count at the last available follow up visit. Circulating microfilariae were detected in 9 mL of peripheral blood collected on daytime using a modified Knott technique followed by Giemsa staining for species identification. For each patient, information on clinical history, laboratory examinations, and treatment was extracted from the medical records, and entered into a Google Drive case report form (CRF). An anonymous code was assigned to each patient. The analyses were performed using Epi Info version 3. 5. 1 (Centers for Disease Control and Prevention, Atlanta, GA, USA) and Stata vers. 15 (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC). Categorical variables were reported as frequencies and percentages whilst discrete or continuous variables as medians and interquartile ranges (IQR).
No serious AE was registered. Patients treated with I and A alone had no AE. Two patients in the AI group (2/21,9. 5%) reported itching; other two patients in the DA group reported itching (2/8,25%), and one patient in the same group reported dizziness. Itching was also reported by three patients of 16 in the DI group (3/16,18. 7%), and one patient in the same group reported dyspnea. The number of AE registered for patients in the D group was 19 (19/74,25. 7%): 8 patients reported itching (8/74,10. 8%), 5 fever (5/74,6. 7%), 2 had a syncope (2/74,2. 7%), and single patients reported one of the following symptoms: thoracic pain, abdominal pain, dizziness, skin rash. None of the patients had to stop the treatment due to AE.
The main limitation of this study is represented by its retrospective design and the substantial proportion of missing data. Also, different drug dosages and durations were used in the study sites, thus limiting the comparison among the regimens. A statistical inference on the different outcomes after each regimen was not performed, as the recruitment was not random and the groups differed markedly before treatment. Indeed, clearance of symptoms, reduction of microfilaremia and eosinophilia are the parameters usually considered to follow up patients with Loa loa infection, but there are no clear indications for the timing of follow-up and the cut-off values defining cure. Finally, the clinical outcome may be difficult to assess, for example in case of persistence of symptoms not clearly attributable to loiasis (such as edemas reported by the patients, but not observed by the health staff). Despite these limitations, we were able to provide useful clinical information on a neglected disease for which complete follow up data are hardly available[16–19].
This paper shows that, in absence of specific guidelines, different reference centers for tropical diseases in Europe use different treatment schedules for loiasis. Our study suggests that some alternative recommendations may be possible for the treatment of loiasis in non-endemic areas, also in consideration of the unavailability of some drugs (namely, DEC). Ideally, a randomized clinical trial would provide a much more robust base of evidence to support management guidelines, but its feasibility in non-endemic countries is questionable. Nevertheless, an observational, prospective study, with well-defined criteria for inclusion and for definition of cure, would be a valuable option in order to evaluate the proposed management indications. | Loa loa is a worm which infects millions of people living in wide forested areas of central Africa. The infection is rarely diagnosed outside Africa, and cases are mainly referred to referral centers on tropical medicine. Aim of this study was to describe the treatment and management of patients diagnosed with loiasis in some referral centers for tropical medicine in Europe. The results showed that different health centers treat the patients with different drugs/drug combinations. On the basis of the availability of the drugs and on the data observed, common protocols might be recommended for the participating centers. | Abstract
Introduction
Methods
Results
Discussion
Conclusions | invertebrates
medicine and health sciences
pruritus
tropical diseases
geographical locations
social sciences
parasitic diseases
animals
neuroscience
parasitology
nematode infections
pharmaceutics
drug administration
neglected tropical diseases
eosinophilia
loiasis
loa loa
hematology
people and places
psychology
eukaryota
nematoda
biology and life sciences
sensory perception
drug therapy
europe
organisms | 2018 | Comparison of different drug regimens for the treatment of loiasis—A TropNet retrospective study | 2,527 | 123 |
The toxicity of available drugs for treatment of leishmaniasis, coupled with emerging drug resistance, make it urgent to find new therapies. Antimicrobial peptides (AMPs) have a strong broad-spectrum antimicrobial activity with distinctive modes of action and are considered as promising therapeutic agents. The defensins, members of the large family of AMPs, are immunomodulatory molecules and important components of innate immune system. Human neutrophil peptide-1 (HNP-1), which is produced by neutrophils, is one of the most potent defensins. In this study, we described anti-parasitic activity of recombinant HNP-1 (rHNP-1) against Leishmania major promastigotes and amastigotes. Furthermore, we evaluated the immunomodulatory effect of rHNP-1 on parasite-infected neutrophils and how neutrophil apoptosis was affected. Our result showed that neutrophils isolated from healthy individuals were significantly delayed in the onset of apoptosis following rHNP-1 treatment. Moreover, there was a noteworthy increase in dying cells in rHNP-1- and/or CpG–treated neutrophils in comparison with untreated cells. There is a considerable increase in TNF-α production from rHNP-1-treated neutrophils and decreased level of TGF-β concentration, a response that should potentiate the immune system against parasite invasion. In addition, by using real-time polymerase chain reaction (real-time PCR), we showed that in vitro infectivity of Leishmania into neutrophils is significantly reduced following rHNP-1 treatment compared to untreated cells.
AMPs are small, cationic proteins which are found in a wide variety of organisms and function as key components of the innate immune system [1], [2]. They exhibit broad-spectrum anti-bacterial, anti-viral, anti-fungal and anti-parasitic activities and have low cytotoxicity to mammalian cells at concentrations required to kill microorganisms [3], [4], [5], [6], [7], [8], [9]. Positive charge together with amphipathicity enable AMPs to interact with negatively-charged microbial surface membranes leading to permeation, disruption and ultimately cell death [2]; a mechanism considerably different from currently available anti-leishmanial drugs. These qualities combined with low susceptibility to resistance makes AMPs good candidates as anti-leishmanial agents [3]. Defensins belong to a large family of AMPs including cysteine-rich peptides with three or four intra-molecular disulfide bonds. They are classified as α-, β- and θ-defensins [10] where the first two are the most common human antimicrobial peptides. The α-defensins found in neutrophils (polymorphonuclear cells, PMNs) are called human neutrophil peptides (HNP) -1, -2, -3 and -4 [11]. Neutrophils constitutively express α-defensins with an increase in production level during infections [12], [13], [14]. In addition to being potent antimicrobial agents, HNPs may act as immunomodulatory molecules since they induce cytokine production and immune cell activation [15]. In cutaneous leishmaniasis (CL), neutrophils are like a double-edged sword. On the one hand, they are believed to be the first recruited effector cells right after infection [16]. Their increased accumulation (as a result of defensins and/or other stimulators) may evoke an immune response set to control Leishmania infection. On the other hand, they may be exploited by the parasites and the presence of neutrophils has been found to facilitate infection experimentally [17]. Thus, understanding the neutrophil-parasite interaction may be an important step towards understanding the underlying mechanisms controlling the parasite. We addressed this interaction by characterization of changes in the production of different cytokines, including tumor necrosis factor-α (TNF-α), Interleukin-8 (IL-8) and transforming growth factor-β (TGF-β). TNF-α is a pro-inflammatory cytokine that causes differentiation and activation of dendritic cells (DC) and macrophages [18] and contributes to intracellular parasite elimination by neutrophils [19]. IL-8 is an important neutrophil chemotactic attractants [18] that induces degranulation of toxic granules allowing neutrophils to kill invading microorganisms [20]. TGF-β, on the other hand, is an immunosuppressive cytokine, beneficial for parasite persistence within neutrophils and a suppression of T-helper 1 type (Th1) response; the protective response against CL. Being abundant in azurophilic granules of human neutrophils; the primary effector cells against cutaneous infection with Leishmania [21], [22] and being considered as the most active human α-defensins [23], we decided to investigate the in vitro activity of recombinant HNP-1 (rHNP-1) against L. major. Using a prokaryotic expression system and an in vitro folding procedure, we succeeded to produce substantial amount of active peptide. Beside in vitro evaluation of rHNP-1 effect on stationary-phase promastigote and amastigote forms of L. major, we assessed the effect of rHNP-1 in combination with CpG motif on human neutrophil' s lifespan and cytokine production pattern. We have previously shown that CpG motifs of class A is superior to class B, in induction of TNF-α production [24]. Therefore, in this study, we harnessed class A CpG motif besides rHNP-1 for investigating their immunomodulatory effects. In vitro cytokine assay showed a considerable increase in TNF-α production from rHNP-1-treated neutrophils and decrease of TGF-β concentration in rHNP-1- and CpG-treated cell cultures. Furthermore, a decreased infectivity following rHNP-1 treatment of human neutrophils was another important effect on the parasite. Moreover, we found that rHNP-1 changes the lifespan of neutrophils in a way anticipated to favor parasite clearance. Our results favor further investigation on AMPs as a new class of anti-leishmanial agents.
Class A CpG motif (ggT GCA TCG ATG CAG ggg gg) used in all experiment was synthesized by TIB MOLBIOL Syntheselabor GmbH (Germany). Bases in capital letters were modified with phosphorodiester and those in lower-case letters with phosphorothioate respectively. L. major (MHRO/IR/75/ER) was cultivated in vitro in M199 medium (Sigma, Germany) supplemented with 5% heat-inactivated fetal calf serum (FCS) (GIBCO BRL, Germany), 40 mM HEPES, 0. 1 mM adenosine, 0. 5 µg/ml hemin and 50 µg/ml gentamicin (all from Sigma, Germany) and was maintained at 26°C. Stationary-phase promastigotes (6-day-old) were harvested and washed in phosphate buffered saline (PBS) prior to use. All blood samples from healthy volunteers were collected following written informed consent, according to institutionally approved procedures (Pasteur Institute of Iran ethical committee, approved on 2nd of October 2009) Whole blood (20 ml) was obtained from healthy volunteers (11 women and 9 men, age 25–60 years). Polymorphprep (Axis-Shield Poc Ac, Norway) was used to isolate polymorphonuclear granulocytes from whole blood, following instructions from manufacturer. The isolated granulocyte population contained 98% neutrophils and 2% eosinophils and there are no lymphocytes or monocytes as determined microscopically after Kimura staining. This staining method enables the discrimination between neutrophils and eosinophils [25]. Isolated neutrophils had 98% viability as assessed by trypan blue dye exclusion. Isolated cells were cultured in RPMI 1640 medium (Sigma, Germany) supplemented with heat-inactivated FCS (10%), HEPES (1%), L-glutamine (1%, Sigma, Germany) and 100 µg/ml gentamicin. RNeasy mini kit (Qiagen, Germany) was used for isolation of total RNA from purified PMNs based on manufacturer' s protocol. The quality of the RNA was assessed by gel electrophoresis on a 1% agarose gel. RNA was then reverse-transcribed into cDNA using oligo-dT primers based on instructions from Omniscript Reverse Transcription kit (Qiagen, Germany). According to HNP-1 nucleotide sequence data registered in GenBank (NM_004084. 3) and by means of primer express program, specific complementary set of forward (5′-TATGGATCCGTCGACATGGCCTGCTAT-3′) and reverse (5′-AATGAGCTCGGTACCGCAGCAGAATGC-3′) oligomers flanked with BamHI and SacI restriction enzyme sites were designed. 50 ng of cDNA and specific primers (10 pmol per reaction) plus 0. 5 µl Taq DNA polymerase (Roche, Germany) were used in the PCR reaction mixture (30 µl). Amplification steps included 95°C for 5 min followed by 40 cycles of 95°C for 15 sec, 63°C for 20 sec, 72°C for 40 sec and 72°C for 10 min as the final step. PCR products were analysed by 2% agarose gel electrophoresis with ethidium bromide staining. Product of the PCR assembly containing HNP-1 coding sequence flanked with BamHI and SacI was purified by a gel extraction kit (QIAquick Gel Extraction kit, Qiagen, Germany) and ligated into the cloning vector pGEM-2 (Promega, USA). The recombinant vectors were transformed into competent DH5α cells. In order to identify colonies including pGEM-HNP-1, PCR and enzymatic digestion were carried out. After isolation and purification, the obtained clone was confirmed by DNA sequencing analysis. After verifying DNA sequence, HNP-1 was ligated into the expression vector pQE-30 (Qiagen, Germany). The resulting recombinant vector (pQE-HNP-1) was transformed into the E. coli M15 cells. After confirmation by PCR and enzymatic digestion, a clone was chosen (pQE-HNP-1) and used in overnight cultivation at 37°C in Luria-Bertani (LB) broth medium supplemented with ampicillin (0. 1 mg/ml, Jaber Ibn Hayan, Iran) and kanamycin (0. 025 mg/ml, Sigma, Germany). The overnight culture was then used to inoculate the fresh LB medium (supplemented with antibiotics) at 37°C. Peptide expression was induced by adding 1 mM isopropyl-β-D-thiogalactoside (IPTG, Roche, Germany), at optical density of 0. 75 and the culture was grown for another 4 hours. 17. 5% standard sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot analysis using anti-His antibodies (Qiagen, Germany) was used to check peptide expression. In order to determine HNP-1 solubility, after 4-hour post IPTG induction, the bacterial pellet was suspended in a buffer containing NaH2PO4 (50 mM), NaCl (300 mM) and imidazole (10 mM) (all from Merck, Germany), and then lysozyme (1 mg/ml, Roche, Germany) was added to the suspension. After 30 min incubation on ice (to destroy the cell wall), the sample was sonicated and centrifuged. Supernatant (soluble extract) and pellet (insoluble extract) were separately evaluated by SDS-PAGE to determine the solubility of recombinant peptide. The 4-hour bacterial pellet distinguished to harbor the highest amount of expressed peptide was dissolved in lysis buffer containing Tris-base (Sigma, Germany) solution (50 mM) and NaCl solution (100 mM), incubated on ice for an hour and sonicated by ultrasonic processor (Cole-Parmer, CPX 750, USA). Then, the suspension was centrifuged at 500 g for 20 min followed by centrifuging supernatant at 12000 g for 30 min. The resulted pellet was resuspended in a denaturating buffer containing 8M urea (Sigma, Germany) and 20 mM imidazole and placed on rocker for 1 hour. Denatured peptide was further purified on Ni-NTA resins through His-tag residues (encoded by pQE-HNP-1 construct). After centrifuging at 12000 g for 30 min, the supernatant was loaded on Ni-NTA column, HiTrap chelating HP (Amersham Biosciences, Sweden). In this FPLC system, two buffers containing 0 mM and 500 mM imidazole along with 8M urea were used for stepwise extraction. 30 mM imidazole buffer was first applied to wash the column. Then, the concentration of imidazole was gradually increased to 350 mM. At this concentration, the HNP-1 dissociated from the column and appeared in the effluent and was collected for further use. In all steps, pH was adjusted to 8. The purified peptide was characterized by MALDI-TOF/TOF mass spectrometry (4700 Proteomics Analyzers, Applied Biosystems, UK). For peptide folding, we followed the method described by Rehder and Borges [26]. Since the peptide was purified in 8M urea, urea was removed to obtain proper folding. Briefly, purified, unfolded HNP-1 was loaded into an Amicon Ultra-15 (3 kDa MWCO) centrifugal concentration unit (Millipore, France) containing 4 ml of 0. 1M acetic acid. The sample was centrifuged in swing-bucket rotor for 40 min at 3000 g. The retentate was then rediluted with 4 ml of 0. 1M acetic acid. Repeating the cycle a total of seven times, resulted in urea dilution more than 106-fold. Then a solution of 0. 1M acetic acid containing 50 µM CuSO4 was prepared and adjusted to pH 8 with a saturated solution of Tris-base. An aliquot of the retentate was diluted to 75 µg/ml with the prepared solution and left to spontaneously refold at room temperature in the presence of air for 16 hours. In order to confirm effective folding, its anti-bacterial activity was investigated in comparison with commercial human alpha-Defensin-1 (commercial HNP-1, Peptide Institute, Japan) against E. coli (ATCC 25922) as a susceptible target. The anti-bacterial activity was determined using a standard protocol against E. coli cells described by Pazgier and Lubkowski [27]. In this assay, the ability of rHNP-1 (with different concentrations ranging from 0 to 60 µg/ml) to kill stationary-phase promastigotes of L. major was evaluated. Furthermore for comparing the activity of rHNP-1 with its commercial form, the effect of two concentrations (10 and 40 µg/ml) of commercial HNP-1 was evaluated on L. major. 50 µl stationary-phase promastigotes (107/ml) were resuspended in sterile PBS and incubated in duplicate with rHNP-1 and its commercial form for 16 hours in a 96-well plate to evaluate the parasiticidal activity of the peptides. Untreated control parasites received mock treatment with sterile Milli-Q water (Milli-Q System, Millipore, France). 25-µl aliquot of each sample was diluted in 275 µl PBS and stained with 3 µl propidium iodide (PI, Biovision, USA) based on manufacturer' s protocol, followed by flow cytometric (FASC) analysis (Partec CyFlow, Germany) to determine the percentage of PI-stained promastigotes in the population. 50000 events were analyzed for each sample using Flomax software (Partec, Germany). LM1 cell line (kind gift of M. Olivier, McGill University, Department of Microbiology and Immunology), immortalized bone marrow-derived macrophages from motheaten mice, was seeded in a 12-well plate at a density of 2×105/ml in serum free DMEM medium (Sigma, Germany), and incubated for 16 hours at 37°C in a humidified atmosphere containing 5% CO2. Non-adherent cells were removed and 1 ml of fresh DMEM medium supplemented with 5% FCS was added to each well. Cells were then infected by stationary-phase promastigote of L. major at 5∶1 parasite to macrophage ratio and incubated for further 16 hours. After that, each well was washed three times by PBS to remove free parasites. Then, cells were treated by 20 µg/ml of rHNP-1 or unfolded HNP-1 followed by a further incubation for 16 hours in the same condition. Culture as control (without rHNP-1 treatment) was included. Cells were collected and subject to genomic DNA extraction using DNA Extraction Kit (Qiagen, Germany) according to manufacturer instructions. The obtained genomic DNA was quantified by a spectrophotometer (Nanodrop, ND-1000, USA) and 150 ng of each sample was subject to real-time PCR using a 7500 Real-Time PCR System (Applied Biosystems, USA) in order to quantify DNA polymerase II (DNApol), the target gene of the parasites, and TATA-binding protein (TBP), one of the house keeping genes of macrophages by use of specific forward and reverse primers. The primers used for DNApol amplification were sense 5′-CGCCTTGTTGTGGACTCCTACT-3′ and antisense 5′-TGTTGCTGCCCTTTGTAATCC-3′. Those used for TBP amplification were sense 5′-AGTTGTCATACCGTGCTGCTA-3′ and antisense 5′-TTCTCCCTCAAACCAACTTGTCA-3′. A PCR volume of 25 µl included 2. 5 µl DNA template, 9 µl QuantiFast SYBR Green PCR master mix (Qiagen, Germany), 1 µl of sense and antisense primer (10 pmol per reaction) and 11. 5 µl DNase free water. The PCR program was initiated at 95°C for 10 min, followed by 40 thermal cycles of 15 sec at 95°C, 30 sec at 60°C and 40 sec at 72°C. Alongside each real-time PCR assay, a series of standards was prepared by performing 10-fold serial dilutions of recombinant pDrive cloning vector (Qiagen, Germany) encoding DNApol or TBP and covering a range of 2500 to 2. 5×107 copies per PCR reaction. This series of standards served as a calibrator and the amounts of DNApol and TBP in each sample was calculated from the generated standard curve at the end of the assay (the regression line for the standards had always an R2>0. 99). As a negative control, reactions without DNA template were also included. For each sample, PCR was performed in duplicate, and each experiment was performed three times. DNApol quantity was normalized to TBP quantity of each sample. In order to determine the cytotoxic effect of rHNP-1 on neutrophils and LM-1 cell line, we performed methylthiazol tetrazolium (MTT) assay 24 hours after incubation of cells with rHNP-1. Briefly, 1. 5×106 neutrophils per well were seeded in a 96-well plate in complete RPMI 1640 medium and treated with different concentrations of rHNP-1 (in the range of 0 act as control up to 60 µg/ml) followed by incubation for 24 hours at 37°C in a humidified atmosphere containing 5% CO2. In the case of LM-1 cell line, cells were seeded in a 96-well plate at the density of 5×104 per well in DMEM medium (without FCS), and incubated for 16 hours at 37°C in a humidified atmosphere containing 5% CO2. Non-adherent cells were removed and then cells were treated with different concentrations of rHNP-1 (as mentioned for neutrophils) in complete DMEM medium followed by incubation for 24 hours in the same condition. Cells were then washed twice with PBS, and 180 µl of fresh medium (without FCS) and 20 µl of MTT (Sigma, Germany) solution (5 mg/ml in PBS) were added to each well, followed by incubation for an additional 4 hours. The supernatants were removed, and 150 µl of DMSO (Sigma, Germany) was added to each well. After complete dissolution of formazan crystals, the optical density (OD) of the solution was measured by an absorbance microplate reader (BioTek, USA) at 570 nm using a reference wavelength of 630 nm. The percentage of cell viability was determined based on following equation: (ODtreated group/ODcontrol group) ×100. Experiments were performed three times in triplicate. Neutrophils isolated from healthy volunteers were seeded in a 12-well plate at a density of 2×106 cells per well in RPMI 1640 medium supplemented with 10% FCS. Half of the wells were infected by stationary-phase promastigotes of L. major at a parasite-to-PMN ratio of 5∶1 and the other half remained uninfected followed by 4-hour incubation at 37°C in a humidified atmosphere containing 5% CO2. Cells were then washed three times with PBS and 1 ml fresh medium was added to each well. All cells (uninfected and infected groups) were treated by 20 µg/ml of rHNP-1 (in the presence or absence of 20 µg/ml of class A CpG motif) followed by a further incubation for 18 hours in the same condition. Apoptosis was assessed by flow cytometry using fluorescein isothiocyanate (FITC) –annexin V and PI (Biovision, USA), according to the manufacturer' s instructions. After 10-min incubation in dark, cells were subject to flow cytometric analysis on a Partec CyFlow cytometer. FSC and SSC were used to identify the PMN population. 15000 events were counted per sample and the percentages of apoptotic, dead or viable cells were determined. Apoptotic neutrophils were defined as annexin V-positive but PI-negative cells, dead neutrophils were defined as annexin V-positive and PI-positive cells and viable neutrophils were defined as annexin V-negative and PI-negative cells. This assay was performed on isolated PMNs from 10 healthy individuals (in ten separate experiments) and in duplicate. The effect of different concentrations of granulocyte-macrophage colony stimulating factor (GM-CSF) on cytokine production by neutrophils was investigated in our previous study [24] and a concentration of 50 ng/ml was considered as optimal. To determine cytokine production from rHNP-1- and/or CpG-treated cells, isolated neutrophils (2×106 cells per well) from 20 healthy volunteers were seeded in 12-well plates in RPMI 1640 medium supplemented with 10% FCS and incubated in the presence of 50 ng/ml of recombinant human GM-CSF (RELIATech GmbH, Germany) at 37°C in a humidified atmosphere containing 5% CO2 for 90 minutes. The supernatants were then removed and half of the wells were infected by stationary-phase L. major promastigotes at a parasite-to-PMN ratio of 5∶1 and the other half remained uninfected followed by additional 4-hour incubation. After washing the cells with PBS three times, all cells (uninfected and infected groups) were treated by 20 µg/ml of rHNP-1 (in the presence or absence of 20 µg/ml of class A CpG motif) followed by a further incubation at 37°C for 18 hours (some wells containing infected neutrophils were added unfolded HNP-1, as control). Culture supernatants were then collected and stored at −80°C until use for quantification of TNF-α, IL-8 and TGF-β by enzyme-linked immunosorbent assay (ELISA). All cytokines were quantified using DuoSet ELISA kits (R&D Systems, Germany) according to the manufacturer' s instructions. After collecting supernatants, untreated, infected neutrophils and those treated by rHNP-1 and unfolded HNP-1 were collected and were subject to genomic DNA extraction using DNA Extraction Kit (Qiagen, Germany) according to manufacturer' s instruction. For further study, neutrophil samples related to 10 individuals (out of 20) were randomly chosen and subject to genomic DNA extraction. Real-time PCR assay was run to test the potential of rHNP-1 on reducing infectivity rate of Leishmania-infected PMNs in vitro. Extracted genomic DNA from each sample was quantified (Nanodrop, ND-1000, USA) and 100 ng of each sample was subject to real-time PCR in order to quantify DNApol, the target gene of the parasite, by means of specific forward and reverse primers (previously mentioned in amastigote detection assay section). PCR condition was the same as mentioned. For each sample, PCR was performed in duplicate, and each experiment was performed twice. The percentage of infectivity rate reduction was determined based on following equation: [1− (DNApol quantitytreated group/DNApol quantitycontrol group) ]×100. GraphPad Prism (version 5, GraphPad Software, USA) was used for all statistical analyses. The results are shown as mean ± standard deviation (SD). A p-value of less than 0. 05 was considered as statistically significant difference.
As shown in Figure S1A, isolated RNA exhibited a desirable quality. Purity of total RNA was evaluated with Agilent 2100 Bioanalyzer according to our previous study [24]. RNA was then reverse-transcribed into cDNA. 2% agarose gel electrophoresis exhibited an expected single band of 123 bp (Figure S1B) in accordance with its calculated length. DNA sequencing of pGEM-HNP-1 clone showed that the gene sequence of interest was in-frame correctly. Expression of peptide was evaluated on 17. 5% SDS-PAGE (Figure 1A) and confirmed by Western blot analysis using anti-His antibodies as shown in Figure 1B. The molecular weight of obtained peptide is about 6. 5 kDa. 4 hours post IPTG induction by comparing the soluble and insoluble extracts on SDS-PAGE, we found the majority of peptide to be presented in the insoluble phase. For further purification of HNP-1 in insoluble extract, FPLC was performed as described in M&M. The obtained isolated fractions were evaluated on 17. 5% SDS-PAGE (Figure 1C) and deemed to be of sufficient purity for further use. Moreover, the mass spectrometry analysis and Triptic digestion showed 63% sequence coverage with theoretic sequence (score = 206). The purified HNP-1 was folded and kept at −70°C for further use. To confirm effective folding of the rHNP-1, anti-bacterial activity of rHNP-1 was investigated against E. coli in comparison with commercial HNP-1 using standard assay of Pazgier and Lubkowski [27]. The results of these experiments are shown in Figure S2. As the bacteria inhibition curve illustrates in this Figure, the growth of E. coli was dramatically suppressed with the increasing concentrations of rHNP-1 and its commercial form. The potency of commercial HNP-1 against E. coli at highest (16 µg/ml) and lowest concentrations were 7-fold and 1. 1-fold more than recombinant form respectively (as illustrated in Figure S2, p<0. 05). The ability of rHNP-1 to kill stationary-phase promastigotes of L. major was evaluated and compared with commercial HNP-1. Stationary-phase promastigotes, incubated with different concentrations of recombinant (ranging from 0 to 60 µg/ml) and commercial HNP-1 (two distinct concentrations; 10 and 40 µg/ml), were stained with PI and subject to FACS analysis in order to determine changes in the PI uptake. As shown in Figure 2B, the higher the concentration of rHNP-1, the higher frequency of necrotic cells. We found that the commercial HNP-1 (at concentrations of 10 and 40 µg/ml) was 1. 4-fold more potent than rHNP-1 (as illustrated in Figure 2C, p<0. 01). The ability of rHNP-1 to kill L. major amastigote was evaluated by quantifying the DNApol gene by real-time PCR in rHNP-1-treated LM-1 cells and compared with parasites treated with unfolded HNP-1 and untreated cells. As shown in Figure 2D, real-time PCR analysis revealed that the normalized DNApol quantity in rHNP-1-treated LM-1 cells was significantly less than untreated ones (24018±3857gene copies versus 53867±8040gene copies, p<0. 05) confirming anti-amastigote activity of rHNP-1. However there was no significant difference between DNApol quantities in unfolded HNP-1-treated and untreated groups which was obviously representative of a direct correlation between antimicrobial activity and peptide folding. The concentrations of rHNP-1 (10,20,40 and 60 µg/ml) used, had no cytotoxic effect on neutrophils. On the contrary, they tended to increase the cell viability compared to control. As shown in Figure 3, the highest effect was related to the concentration of 20 µg/ml, so was selected as optimal for further studies. In the case of LM-1 cells, we found that the concentrations above 20 µg/ml of rHNP-1 were toxic, but lower concentrations (10 and 20 µg/ml) did not affect cell viability compared to control (data not shown). The lifetime of neutrophils is relatively short and they undergo spontaneous apoptosis, which can be inhibited by various pathogen- and host-derived substances. Here, we investigated whether rHNP-1 (as a host derived substance) and/or CpG (as a potent immunostimulator) could influence neutrophil apoptosis. Most neutrophils (85%), either infected or uninfected, were Annexin V+ (i. e. apoptotic and dead population) after 18 hours incubation (Figure 4A). Following rHNP-1 treatment of uninfected neutrophils, the percentage of Annexin V+ population was significantly reduced compared to untreated, uninfected and infected neutrophils (66±5% versus 88±1% and 89±3%, respectively, p<0. 001, Figure 4A and Figure S3). This indicates that rHNP-1 increases the lifespan of uninfected neutrophils. Except for rHNP-1 treatment of uninfected neutrophils, the other treatments did not have any effect on the percentage of viable cells (Figure 4A). Furthermore, the percentages of Annexin V+, PI− population (apoptotic cells) and Annexin V+, PI+ population (dead cells) in all treated groups were significantly different from controls (Figure 4B). In untreated neutrophils, 79. 9±2. 8% and 76±6. 5% for uninfected and infected groups, respectively, were Annexin V+, PI− (apoptotic). Following different treatments, parallel with decreasing in apoptotic cells, an increase in the frequency of dead cells (Annexin V+, PI+ population) was seen (irrespective of being infected or not (Figure 4B) ). Adding CpG motif or rHNP-1 separately to neutrophils (either infected or not) caused significant decrease in apoptotic population (40±3% and 46±7% for uninfected groups and 30±11% and 44±8% for infected groups, respectively, p<0. 001) and adding the combination of them caused the maximum decrease in the frequency of apoptotic cells (22±1% and 15±1% for uninfected and infected groups, respectively, p<0. 001), showing their synergistic effect on neutrophil death. The addition of rHNP-1 increased viability of neutrophils (34±5, p<0. 001) confirming MTT assay result, while in all other groups, the percentage of viable cells was the same as control groups (≈11. 5%). To further investigate the activities of rHNP-1 and/or CpG motif on human neutrophils, we evaluated their abilities to induce production of the cytokines TNF-α, IL-8 and TGF-β under different conditions. To increase cell viability and delay apoptosis, neutrophils were first treated by 50 ng/ml of GM-CSF. This procedure will also amplify cytokine production to the detection range of ELISA kits used. GM-CSF-treated neutrophils (either infected or not) were stimulated by rHNP-1 and/or CpG motif for 18 hours. The cytokines of interest were then quantified in cell-free supernatants by ELISA. As shown in Figure 5A, rHNP-1 markedly induced production of TNF-α in uninfected neutrophils, (310±188 pg/ml versus 20±10 pg/ml, p<0. 001). Furthermore, rHNP-1 exerted its highest effect on infected neutrophils leading to a 3-fold increase in TNF-α concentration as compared with its effect on uninfected neutrophils (843±257 pg/ml versus 310±188 pg/ml, p<0. 001). Although, treatment of uninfected and infected cells by CpG motif alone increased and decreased TNF-α production respectively (p<0. 001), the value of this cytokine in the cultures of these groups was less than 50 pg/ml. Co-administration of rHNP-1 with CpG motif did not cause any significant change in TNF-α concentration compared to neutrophils treated by rHNP-1 alone (p>0. 05). rHNP-1 treatment of uninfected neutrophils (regardless of CpG motif treatment) primed the release of IL-8 and reduced IL-8 production from infected neutrophil culture (Figure 5B, p≤0. 01). CpG motif alone had a decreasing effect on IL-8 release from infected neutrophils (1279±230 pg/ml versus 1621±214 pg/ml, p<0. 001) and had no significant effect on uninfected neutrophils. The reduction effect of CpG motif on IL-8 production from infected neutrophils was significantly higher than rHNP-1 (p<0. 001). The highest concentration of TGF-β was detected in untreated, parasite-infected neutrophils (Figure 5C), equal to 452±95 pg/ml (p<0. 001 versus uninfected neutrophils). Whereas rHNP-1 enhanced TGF-β release from uninfected neutrophils (207±94 pg/ml versus untreated neutrophils; 151±57 pg/ml, p≈0. 04), it attenuated the TGF-β release from infected ones (326±77 pg/ml versus 452±95 pg/ml, p<0. 001). Further stimulation with CpG motif and rHNP-1 had similar effect as rHNP-1 alone; increasing TGF-β release in uninfected group (172±55 pg/ml versus 151±57 pg/ml, p<0. 001), while decreasing a release from infected group (322±75 pg/ml versus 452±95 pg/ml, p<0. 001). CpG treatment reduced TGF-β up to 60% in both uninfected and infected groups (compared to uninfected or infected control, p<0. 001). To summarize, rHNP-1 (or its combination with CpG motif) markedly induced production of TNF-α, while an inhibitory effect was seen on both IL-8 and TGF-β productions in Leishmania-stimulated cultures. Treatment of infected neutrophils by unfolded HNP-1 was acting as negative control and concentrations of different cytokines in this culture did not differ from those in untreated, infected neutrophil culture. To evaluate if rHNP-1 had effect on the Leishmania infectivity rate of PMNs, the DNApol gene of rHNP-1-treated, Leishmania-infected neutrophils was quantified by real-time PCR and compared to untreated group. We found that infectivity rate was reduced in 8 out of 10 rHNP-1-treated neutrophil samples, with a decrease of 20 to 80% (Figure 6). Infectivity rate reduction between rHNP-1-treated and unfolded HNP-1-treated groups was significantly different (46±18% versus 12±4%, p≈0. 001), showing a correct folding dependence of antimicrobial activity of HNP-1.
All currently used first-line and second-line drugs for the treatment of leishmaniasis have drawbacks in terms of toxicity, cost and resistance. Because of low susceptibility to resistance and being nontoxic to host cells at effective concentrations against pathogens, AMPs are attractive candidates. Despite the extensive documentation on the activities of AMPs against bacteria and fungi, studies investigating the effects of AMPs against protozoa are rare and to the best of our knowledge there is no report on HNP-1 tested against L. major. Pharmacological studies of defensins needs relatively large quantities of pure functional peptides. Since the purification process of natural peptides is difficult and chemical synthesis of these peptides is expensive, recombinant strategies are a promising alternative. By using bacterial expression system, we succeeded in obtaining substantial amounts of peptide, which was purified on Ni-NTA column via FPLC. In contrast to the commercial form of HNP-1, the purified peptide needed folding in order to become bioactive. By using the method of Rehder and Borges for in vitro folding of proteins with disulfide bonds [26], we succeeded in producing active recombinant peptide with a functional anti-bacterial activity, albeit lower than that of commercial HNP-1. We found that our rHNP-1 was active against stationary-phase promastigotes of L. major. By flow cytometric analysis, we observed that increased levels of rHNP-1 caused necrosis of parasites, which reached to 50% at a concentration of 60 µg/ml (Figure 2B). While the commercial HNP-1 was slightly more potent in parasite killing (1. 4-fold higher), the microbicidal activity of in house recombinant HNP-1 was deemed to be sufficient and due to cost-effectiveness preferred for further studies. The concentration needed to kill 50% of parasites (about 60 µg/ml) is higher than that needed to kill 50% of bacteria (less than 7 µg/ml). This may be due to the complicated membrane structure of the parasite; extremely rich in glycosylphosphatidylinositol-anchored molecules such as leishmanolysin (the major surface-metalloprotease) and LPG, which are considered as protective agents for parasites. Since the activities of AMPs may vary against amastigotes; the pathological form of Leishmania in vertebrate host [28], the anti-amastigote activity is important to be determined for further drug development. To determine the effect of rHNP-1 on L. major amastigotes, we measured the amastigote laden in rHNP-1-treated and untreated LM-1 macrophages. The quantification was by using DNApol as the target gene of parasite. Prina et al. have shown that Leishmania DNA degradation follows rapidly after parasite death [29]. Thus the quantified DNA was related to the number of viable amastigotes. Based on our results and previous documents that denatured HNP-1 does not display antimicrobial activity [30], as a control, we tested the effect of unfolded HNP-1 on the parasite load in LM-1 macrophages. While rHNP-1 treatment significantly reduced the amastigote load (24018±3857 gene copies versus 53867±8040 gene copies, p<0. 05), unfolded HNP-1 had no significant effect, similar to control (Figure 2D), indicating that anti-parasitic effect of rHNP-1 (like its anti-bacterial effect) relies on a correct tertiary structure. The mechanism by which rHNP-1 reduces amastigote load in macrophages is unknown, but may be a direct effect of rHNP-1 on parasites. In this respect, Arnett et al. have shown that HNP-1 is taken up by bacteria-infected macrophages and reduces bacterial load [31]. Neutrophils play an important role as host cells in the early phase of L. major infection. Therefore, agents that modulate neutrophil functions may alter the outcome of infection and/or vaccination. In addition to its potent antimicrobial activity, HNP-1 has immunomodulatory effects; stimulating various immune cells to production of cytokines. HNP-1 has been shown to act as a chemotactic factor for both macrophages and T lymphocytes [11] and HNP-1 treatment enhances leukocyte accumulation at the site of infection in K. pneumonia-infected mice [30] and provokes TNF-α release from lymphocytes [32]. Further, HNP-1 treatment of macrophages has been shown to inhibit proliferation of the intracellular bacteria [31]. Another immunomodulatory molecule is CpG motif. It has been shown that CpG motif administration to mice causes neutrophil accumulation at the site of infection [33], [34] and induction of IL-8 secretion [35]. In our previous study, we showed considerable level of TNF-α production by class A CpG-treated neutrophils following GM-CSF treatment [24]. Further, class A CpG motif is considered as a potent stimulator of Th1 response [36]. To follow up on the anti-parasitic effect of rHNP-1, we tested the immunomodulatory effect of rHNP-1 on parasite-infected neutrophils combined with class A CpG motif co-stimulation. Treatment of neutrophils with rHNP-1 was found to delayed the apoptosis of uninfected neutrophils, which is in concordance with the study by Nagaoka et al. [37]. However, rHNP-1 treatment had no effect on viability when neutrophils had been infected with L. major. Nagaoka et al. showed that the anti-apoptotic effect of rHNP-1 may be related to its effect on the expression of truncated Bid and Bcl-xL (considering as pro-apoptotic and anti-apoptotic proteins) and consequently caspase 3 activity (a key executor of apoptosis) [37]. Why rHNP-1 had no anti-apoptotic effect on infected neutrophils is not clear, but it could be possible that parasite antagonizes the signaling pathway through which rHNP-1 exerts its anti-apoptotic effect. We suggest that rHNP-1, rather than blocked, delayed the neutrophil apoptosis for up to 24 hours. In support of this, we observed that the number of viable treated cells dropped during the second day of culture (data not shown). Contrary to reports suggesting that CpG motif delays neutrophil apoptosis [35], [38], we found no such effect of CpG motif alone or in combination with rHNP-1 (Figure 4A). This controversy over CpG results may be due to the different kinds of CpG motif used in our experiments and mentioned studies; they used E. coli DNA for stimulating neutrophils, while we used class A CpG motif. It was reported that short oligonucleotides are generally less potent immunostimulators than E. coli DNA [39], [40]. A noteworthy similarity among all treated neutrophils is reduction in the ratio of apoptotic to dead cells in comparison with control (Figure 4B). Two possible mechanisms could be considered for this reduction in the ratio of apoptotic to dead cells. First, rHNP-1 and/or CpG treatment stimulated the signaling pathways required for transition from early apoptosis to secondary necrosis and consequently accelerated this step (as compared with untreated groups). Second possible mechanism is that such treatments partly blocked apoptosis pathway and drove cell death by necrosis rather than apoptosis. It seems that these dead cells are usually better at triggering inflammatory responses and altering the lifespan of neutrophils may influence the fate of intracellular Leishmania parasite. While necrotic and apoptotic neutrophils are engulfed by macrophages to similar extents [41], neutrophils that die by necrosis, are usually better at triggering inflammatory responses. It has been shown that phagocytosis of apoptotic neutrophils by macrophages, contrary to necrotic bodies, fails to trigger antimicrobial effector functions [22], [42], [43]. The implication of these finding on our results are not clear, as necrosis as well as prolonged neutrophil survival, which maybe represent enhanced neutrophil function, could be envisaged to have beneficial effects for the host. We therefore sought to investigate how rHNP-1 and CpG motif affected cytokine/chemokine production by neutrophils. For this purpose, we measured TNF-α, IL-8 and TGF-β concentrations in culture supernatants. rHNP-1 had significantly higher effect than CpG motif on induction of TNF-α production by neutrophils and there was no additive effect of CpG motif (Figure 5A). An additive effect on TNF-α production was seen when rHNP-1 was added to L. major-infected neutrophils; parasite itself caused a 1. 6-fold increase in TNF-α production over background, while rHNP-1 treatment of infected neutrophils exhibited a 25. 6-fold increase (p<0. 001). TNF-α is a pluripotent cytokine, which in the context of Leishmania infection is primarily linked to protection. Thus, enhanced TNF-α would be expected to be beneficial to the host. IL-8 is a chemokine important in recruitment of inflammatory cells. We found, in concordance with Laufs et al. and Zandbergen et al. [44], [45], a small (1. 5-fold) but significant (p<0. 001) increase of IL-8 in parasite-infected neutrophil culture in comparison with uninfected control. While both rHNP-1 and CpG motif showed a reduction effect on IL-8 production in parasite-infected neutrophils, the latter had more potent effect. Similar to the effect on IL-8 production, we found that parasite uptake by neutrophils stimulated TGF-β production, an immunosuppressive cytokine associated with susceptibility to Leishmania infection [46]. Interestingly, rHNP-1 and CpG treatment of infected neutrophils reduced the level of TGF-β up to 28% and 38% respectively (p<0. 001). The reduced Leishmania infectivity rate, observed following rHNP-1 treatment of neutrophils, maybe correlated with the marked effect of rHNP-1 on TNF-α production, a pro-inflammatory cytokine potentiating neutrophil microbicidal capacity. Moreover, This reduction may be explained by direct leishmanicidal effect of rHNP-1 on entrapped promastigotes in neutrophils, since Leishmania parasites do not fully transform into amastigotes inside neutrophils [22], [42]. However, for this direct effect, HNP-1 must be transported into the neutrophils and trafficked to the parasitophorous vacuole (containing parasites) or interact with the releasing parasites from the parasitophorous vacuole to carry out direct leishmanicidal activity. Alternatively, it is possible that HNP-1, as may be suggested by the induction of TNF-α production, triggers mechanisms in neutrophils and thereby indirectly potentiates their ability to kill parasites. In conclusion, HNP-1 and similar peptides may represent a promising alternative in the search for new nontoxic, broad-spectrum antibiotic agents. Due to defensins' potency to promote development of protective immune responses, they can be regarded as adjuvants. By optimizing simple methods for producing active form of AMPs including HNP-1, new horizons for further investigation on functional, structural and pharmacological features are open. Better understanding of the properties and mechanisms underlying immunomodulatory effects of rHNP-1 on neutrophils may benefit the design of new and improved leishmanicidal derivatives with high selectivity for the pathogens and/or with potent immune-adjuvant effects. Engagement of neutrophils may enhance the killing of Leishmania, and facilitate initiation of a proper immune response against Leishmania infection. | In Iran, cutaneous leishmaniasis (CL) is a widespread and highly endemic disease in young individuals. To date, treatment strategy is based on chemotherapy accompanied with high incidence of toxicity and drug resistance. Distinctive mode of action of defensins (members of antimicrobial peptides) with low susceptibility to resistance and low toxicity to mammalian cells makes them suitable candidates for anti-leishmanial agents. The most active human defensin is human neutrophil peptide-1 (HNP-1) produced by neutrophils; the first effector cells during Leishmania infection. In this work, we used recombinant HNP-1 (rHNP-1) against both the promastigote and amastigote forms of Leishmania (L.) major. Furthermore, immunomodulatory effect of rHNP-1 on Leishmania-infected neutrophils was investigated. Our result showed that rHNP-1 has anti-parasitic effect against L. major promastigotes and amastigotes and also reduces infectivity rate of Leishmania-infected neutrophils. Moreover, assessment of cytokine production from Leishmania-infected neutrophils reveals an increase in TNF-α and a decrease in TGF-β production after rHNP-1 treatment; a cytokine pattern anticipated to facilitate control of parasites. The immunomodulatory effect of rHNP-1 on cytokine production from parasite-infected neutrophils besides its direct effect on free parasites is considered as promising step towards developing new anti-leishmanial agents. | Abstract
Introduction
Materials and Methods
Results
Discussion | 2013 | Human Neutrophil Peptide-1 (HNP-1): A New Anti-Leishmanial Drug Candidate | 12,171 | 384 |
|
Using a transgenic mouse model harboring a mutation reporter gene that can be efficiently recovered from genomic DNA, we previously demonstrated that mutations accumulate in aging mice in a tissue-specific manner. Applying a recently developed, similar reporter-based assay in Drosophila melanogaster, we now show that the mutation frequency at the lacZ locus in somatic tissue of flies is about three times as high as in mouse tissues, with a much higher fraction of large genome rearrangements. Similar to mice, somatic mutations in the fly also accumulate as a function of age, but they do so much more quickly at higher temperature, a condition which in invertebrates is associated with decreased life span. Most mutations were found to accumulate in the thorax and less in abdomen, suggesting the highly oxidative flight muscles as a possible source of genotoxic stress. These results show that somatic mutation loads in short-lived flies are much more severe than in the much longer-lived mice, with the mutation rate in flies proportional to biological rather than chronological aging.
Aging is a complex process with an as yet unclear etiology and a - for most species - poorly described phenotype [1]. Accumulation of somatic mutations, due to imperfect repair and maintenance, has since long been implicated as a universal, major cause of aging [2], [3], but proved difficult to study in higher organisms. Most somatic DNA mutation assays are indirect and based on alterations in phenotypic characteristics, such as the mouse or Drosophila spot tests [4], [5]. In the past, we have generated transgenic mouse models harboring chromosomally integrated lacZ-plasmid constructs that can be recovered in E. coli for the subsequent quantification and sequence characterization of a broad range of spontaneous mutations [6]. The results with this system indicate that somatic mutations accumulate in virtually all organs and tissues of the mouse albeit at different rates and varying in the types of mutational events [7]. Very little is known about spontaneous DNA mutation burdens in somatic tissues during aging of different higher organisms. For mammalian species there is evidence that germ line mutation rates in different phylogenetic groups are generally higher in short-lived Drosophila or rodents than in longer lived primates or birds [8]. This has been ascribed to either differences in generation time or selection of increasingly efficient mechanisms of DNA replication and repair during the evolution of long-lived mammals [9]. However, DNA mutation loads in somatic tissues of invertebrates and vertebrates have never been directly compared. We recently generated several lines of D. melanogaster harboring a lacZ-plasmid construct identical to the one used to generate the mouse models [10]. This allowed us, for the first time, to directly compare spontaneous somatic mutation frequencies and spectra between a mammal and an insect as a function of age. The results indicate a significantly higher somatic DNA mutation frequency in flies than in mice with especially the fraction of the more toxic genome rearrangements much higher in the former. Like in the mouse, also in flies mutation frequencies increase with age, but at a much higher rate at higher temperature, which correlates with an increased aging rate and shorter life span.
Figure 1A schematically depicts the lacZ-plasmid model for mice and flies. The systems are identical for the two species except for the copy number; while in mice there are approximately 10 copies per integration site, with integration sites on chromosomes 3 and 4, each fly line only contained one copy. Since mutation frequencies are calculated as the number of mutant lacZ copies per total copy number of lacZ-plasmids recovered from a given DNA sample, this has no influence on the mutation frequencies observed. It should also be noted that in general mutation frequencies do not depend on the integration site. While in the past we have observed rare mouse lines with much higher mutation frequencies [11], and there was also at least one fly line that differed significantly from most of the others [10], overall very similar results were obtained for different integration sites. Hence, we felt confident that a direct comparison between the two species would show, for the first time, natural levels of spontaneous mutations in somatic tissue of an invertebrate and a vertebrate. Figure 1B shows the spontaneous DNA mutation frequencies in the two species, measured as the number of mutant lacZ copies per total number of lacZ copies isolated from mouse liver or whole fly DNA. For mouse liver, mutation frequency is expressed as the average over multiple individual animals and for Drosophila as the average over multiple batches of 50 flies. The results indicate a spontaneous mutation frequency in young (1–2 days), male flies of 12. 6×10−5, which is about 3-fold higher than the mutation frequency in the liver of a young (3 months), male mouse, i. e. , 3. 8×10−5 (Figure 1B). The mutation frequency in the mouse is very similar as what we reported previously and, at young age, does not vary much between tissues [7]. The lacZ-based mutation reporter system allows the characterization of mutations simply by restriction digestion and/or sequence analysis of the plasmids from the E. coli colonies on the selective plates (Figure 1A). In the mouse we found that most mutations are either point mutations (basepair substitutions or small deletions) or genome rearrangements (with one breakpoint in the lacZ gene and the other elsewhere in the mouse genome), with very few intra-lacZ deletions [12]. In the fly we could make the same classification, but while in mouse tissues genome rearrangements were less than 50%, in the fly such events were predominant (Figure 1B). The small fraction of point mutations was dominated by GC to AT transitions and 1-bp deletions (results not shown). Hence, Drosophila cells and tissues tolerate not only a 3-fold higher mutation frequency than mouse cells, but also a substantially higher fraction of genome rearrangements. Mutations are generally assumed to be associated with cell division when they arise, for example, as errors in repairing damage to DNA. Drosophila is a postmitotic organism and contains very little cell proliferative activity as adults. Hence, one would predict that in contrast to the mouse, Drosophila would not accumulate mutations during its adult life. Lifespan in Drosophila is temperature-dependent and longevity decreases exponentially with increasing temperature between 12 and 30°C [13]. This may be caused by an increase in the rate of metabolic processes, presumably speeding up the aging process through the accumulation of damage. As shown in Figure 2A and 2B, life span of both female and male lacZ transgenic flies is indeed temperature-dependent. While at 29°C the flies live very short with no survivors after 30 days, at 18°C they can live three times as long. Mutation frequencies at the lacZ locus appeared to be both age and temperature-dependent (Figure 2C). Indeed, this is very clear from Figure 2D where we plotted side by side the age-related increase of lacZ mutation frequencies in male lacZ flies at 18 and 29°C. Of note, these results were obtained in an experiment independent from the experiment that gave rise to the data presented in Figure 2C. Similar results were obtained when using another lacZ fly line, line 5 with the lacZ reporter integrated elsewhere on the same chromosome (Figure S1; see also ref. [10]). From Figure 2D it is obvious that the rate of the age-related increase differs dramatically between 18 and 29°C. Indeed, when plotting the mutation frequencies in Figure 2C as a function of the remaining life span, as obtained from the survival curves in Figure 2A and 2B, highly significant correlations were observed both for males and females (Figure 2E). This suggests that mutations accumulate as a function of biological rather than chronological age. If true, this would predict that when for each temperature mutation frequencies are plotted against chronological rather than biological time they would be statistically significantly different from each other. As we show in Figure S2, this was indeed the case, for both line 11 and line 5. Next, we considered the possibility that increasing the temperature could alter the type of mutations. Hence, we compared the mutation spectrum in flies of 4 weeks old at 18 and 29°C. The results show that in both cases the far majority of the mutations were genome rearrangements, the accumulation of which with age and temperature was statistically significant (Figure 3A). The point mutation loads are fairly constant, but their numbers were too small to draw the conclusion that only genome rearrangements accumulate. However, it is clear that the temperature does not dramatically affect the mutation spectrum as we have seen, for example, after treatment of the flies with the known point mutagen ethyl nitrosourea [10]. Next, as we have done previously for the mouse [14], we physically characterized the rearrangements to assess if the mechanism by which these events arise might be different between the two extreme temperatures (Table S1). Like in the mouse, in flies we observed both intra-chromosomal and extra-chromosomal events (Figure 3B). The latter were defined as translocations of which there were very few in the fly. In the fly, virtually all intra-chromosomal events were either internal lacZ deletions or events that involved a breakpoint close to the integration site of the reporter gene. The latter could be subdivided, based on the direction of the sequence that was recovered from the mutant plasmid and mapped to chromosome 3, as deletions, inversions or more complex intra-chromosomal recombinations, which included transpositions (not to be confused with the P element transpositions causing hybrid dysgenesis in Drosophila) and complex, undefined events. While the types of genome rearrangements observed were similar to what we previously reported for the mouse, the number of rearrangements close to the integration site was much higher for the fly. In mouse tissues we found many more very large events, including many translocations [14]. This is to be expected because Drosophila has less and much shorter chromosomes than the mouse and therefore less opportunity for accommodating extra-chromosomal recombinations or very large intra-chromosomal events. As shown in Figure 3B, no significant differences between 18 and 29°C were observed. Of note, at neither temperature we observed obvious sequence homologies at the break points. Hence, most likely these mutational events are a consequence of erroneous non-homologous end joining of DNA double-strand damage. These results indicate that increasing temperature merely accelerates mutation accumulation without significantly altering the spectrum. This would be expected when temperature affects the basic rate of aging. The similar mutation spectrum between flies aged at the two extreme temperatures strongly suggests that similar types of DNA damage drive the generation of these mutations. Since our mutation frequency determinations were based on whole flies we were not able to tell where in the fly the mutations were generated. Mutations are often considered to arise as errors during DNA replication. While Drosophila is generally considered a postmitotic organism, mitotically active cells are present in the abdomen, i. e. , the gonads and the gut. Hence, we analyzed mutation frequencies at the lacZ locus in abdomen, thorax and heads of young (1 week) and old (3 weeks), female (Figure 4A) and male (Figure 4B) flies at 29°C. Somewhat surprisingly, the results indicate that most age-related mutations had accumulated in the thorax, not the abdomen. The higher mutation frequency in thorax as compared to abdomen of old flies was statistically significant for males, but not for females; females merely showed a higher mutation frequency in thorax as compared to the heads (also true for males). No significant differences in mutation spectrum were found between body parts; virtually all mutations were genome rearrangements (not shown).
Somatic DNA mutations are random and almost always have adverse effects or are neutral. In cancer, random mutations are selected by providing the host cell with attributes allowing it to escape normal growth constraints, invade tissues and evade host defense systems, such as the immune response. However, increasing loads of random mutations also adversely affect cellular fitness, as has been demonstrated conclusively in E. coli [15]. In higher organisms, such as mice and flies, somatic mutations may exert adverse effects mainly by deregulating gene control. Hence, one would expect somatic mutation frequencies to reach levels that are still compatible with life without compromising reproduction at early age. Using a similar reporter system for Drosophila as we previously used in mice, we now show significantly higher and potentially more severe DNA mutation loads at the lacZ locus in tissues of the fly as compared to the mouse. While the lacZ reporter is not expressed, acting as a neutral target gene, mutation detection requires inactivation or partial inactivation of the lacZ-encoded β-galactosidase activity [16]. Therefore, our present data underestimate the total somatic mutation rate in the organism. Germ line mutation rates, which can now be derived from direct sequence comparison between individual animals [17], can vary between 1 and 10×10−6 per average gene per generation [9], [18]. However, somatic mutation rate is not subject to the same selection constraints as germ line mutation rate. Our current data indicate that the frequency of somatic mutations per locus in flies is 3-fold that of a mammal (Figure 1B). While the fly has a smaller genome than the mouse (about 16-fold), it is more compact in the sense that its gene density is much higher. Hence, random mutations should be more likely to have an adverse effect on the fly genome than on that of the mouse. Moreover, the fly shows a much higher proportion of genome rearrangements - mutational events that generally have a much higher functional impact than point mutations - than the mouse. Our results in Drosophila also indicate, for the first time, that the rate of somatic mutagenesis is a function of both age and temperature, with temperature as the main factor. Since increased temperature reduces life span of flies, it is tempting to speculate that somatic mutation rate is causally related to the rate of aging. For the mouse this is supported by our previous results indicating reduced spontaneous mutation frequencies in mice subjected to caloric restriction [19], an intervention leading to extended life span [20]. We did not find a similar result for the fly, however [21], which is in keeping with results by Mair et al. , showing that DR extends life span in Drosophila entirely by reducing the short-term risk of death [22]. This difference in response to DR between mice and flies may reflect a difference in the modes of dietary restriction, which is inherently more complicated in flies than in rodents [23]. Interestingly, Mair et al. [22] also demonstrated that reduced temperature, in contrast to the effect of DR, may well increase the life span of flies by reducing the accumulation of aging-related damage. While we have no direct evidence regarding the source of somatic mutations in Drosophila, elevated temperatures increase respiration rate in ectotherm animals and thus the production of mitochondrial reactive oxygen species (ROS) [24]. ROS is considered as a major cause of aging [25] and produces a multiplicity of alterations in DNA, including chromosomal aberrations, probably as a consequence of DNA double-strand breaks [26]. Treatment of the lacZ Drosophila lines with paraquat, a widely used herbicide that produces ROS in cells, resulted in a significant elevation of somatic mutations [21], most or all of which were genome rearrangements (results not shown). Hence, it is conceivable that the genome rearrangements observed in flies are a result of increasing amounts of ROS generated as a function of the temperature. This would explain why the amount but not the type of mutation depends on both age and temperature. Our results indicate that while mutations are generally attributed to errors during replication and thus require cellular proliferation most of the mutations in Drosophila accumulate in the thorax, a postmitotic organ (Figure 4). In this respect we speculate that high levels of oxidative stress are generated by the flight muscles, which have a very high metabolic rate. This would induce DNA double-strand breaks. Subsequent errors during the repair of the damage, for example, mis-annealing of different DNA ends by non-homologous end-joining or errors during homologous recombination, using the homologous chromosome or sequences elsewhere on the same chromosome as exchange partner rather than its sister chromatid [27]. Interestingly, after crossing the lacZ flies with a line harboring a defect in the BLM gene (DmBlm) [28] - which is required for accurate repair of DNA double-strand gaps by homologous recombination - the age-related accumulation of genome rearrangements was significantly elevated. The DmBlm mice lived significantly shorter than the wildtype control animals (Garcia, A. , et al. , submitted). Hence, genome rearrangements can evidently accumulate during adult life, possibly as a consequence of errors during DNA double-strand break repair. This is not dependent on cell proliferation. As we demonstrated previously in cultured mouse cells, large rearrangements are easily induced in non-dividing cells by hydrogen peroxide, an oxidative agent [29]. The tolerance of Drosophila for severe genomic mutation loads is remarkable and may reflect a reduced level of gene regulatory intricacy as compared to mammals [30]. In the mouse, sizable fractions of genome rearrangements could disrupt the many long-distance gene regulatory interactions and might be unsustainable. The evolution of more complex species with longer life spans and more numerous cell divisions has been associated with more complex mechanisms for gene regulation, which most likely required increasingly sophisticated systems for replication and repair to prevent the deleterious effects of genome rearrangements.
Male lacZ mice of line 60 were used with integration sites on chromosomes 3 and 4 at approximately 10 copies per integration site [31]. They were maintained in the animal facilities of the University of Texas Health Science Center at San Antonio on a 14-h light/10-h dark cycle at a standard temperature of 23°C. Standard lab chow (Harlan Teklad, Madison WI) and water were supplied ad libitum. Animals were sacrificed by cervical dislocation following CO2 inhalation. Ethical approval to carry out this work on animals was provided by the IACUC of the University of Texas Health Science Center. The transgenic lines 11 and 5, harboring the lacZ reporter gene on chromosome 3R, was used in this study and generated by P-element transformation using a w1118 background [10]. Flies were raised on standard cornmeal-molasses-agar-yeast medium with propionic acid added to the food as an anti-fungal agent. All mating, egg laying and hatching were done at 25°C and 60% humidity. Once hatched, one to two day old flies were transferred to cages containing 200–300 flies and maintained at 18°C, 25°C and 29°C, respectively, throughout the experiment. Males and females were maintained together in the same cage throughout the experiment but they were separated by sex during collection time. Food was changed every other day and samples of 50 flies for each sex were collected at 5 days, 2 weeks, 4 weeks and 6 weeks. For the longevity study of line 11 at 18°C additional male samples were also collected at 8,9 and 10 weeks. Samples were stored at −80°C. For the body parts study, flies were obtained as described above and 1–2 days old flies were transferred to a 29°C incubator. Samples of 100 flies for each sex were collected at 1 week (young) and 3 weeks (old) of age and stored at −80°C. The dissection was done under the microscope (Stemi SV 6 Zeiss) in the presence of dry ice. Flies were dissected into head, thorax and abdomen and mutant frequency determinations were carried out for each sex and age group. For the survival determination, fly deaths were recorded every 2–3 days and dead flies were removed from the cages. The total number recorded disregards rare escape or accidental death of flies. All life span determinations were performed independently of the mutation frequency assessments. Each sample consisted of 50 pooled male or female flies. Flies were homogenized in 600 µl of lysis buffer (10 mM Tris-HCl, pH 8. 0; 10 mM EDTA; 150 mM NaCl) in 2 ml eppendorf tubes using a battery-operated pestle. To the homogenate, 12 µl of Proteinase K (25 mg/ml), 60 µl 10% SDS and 10 µl RNAse A (20 µg/ml) were added and samples were incubated at 65°C while rotating during 30 min. Genomic DNA was subsequently extracted from these samples using phenol/chloroform. The mutation frequency was determined as described in detail elsewhere [31]. Briefly, isolated DNA (either from mouse liver tissue or from 50 flies was digested for one hour at 37°C with Hind III (40 U) in the presence of magnetic beads coated with lacI-lacZ fusion protein. The lacZ plasmid was then eluted from the beads by incubation with IPTG, circularized by ligation with T4 DNA ligase (Biolabs), precipitated with ethanol and used to electrotransform E. coli (ΔlacZ, galE -). Each mutant frequency determination point was based on at least 3 replicates of the same sample, i. e. , three 50-fly groups from the same population, with a minimum of 100,000 colonies for each rescue. The mutation frequency is the ratio of colonies growing on the selective plate vs. the total number of recovered plasmids from the DNA sample (as measured on the titer plate). Hence, mutation frequencies as determined with this system reflect a ratio and do not depend on the amount of DNA. They are expressed on a per locus basis as the number of mutant lacZ copies for a given number of lacZ copies isolated from the in vivo situation. LacZ plasmids from mutant colonies were further characterized as described in detail elsewhere [31]. Sequence reactions of purified mutant plasmids were outsourced to Sequetech corporation (Mountain View, CA). The returned chromatograms were analyzed with Sequencher (Gene Codes, Ann Arbor, MI). Analysis of large rearrangements consisted of non-lacZ sequences were carried out using the fly genome database (http: //www. flybase. org). After alignment with the D. melanogaster sequence, the chromosomal origin of the flanking sequences was determined and the orientation and the type of chromosomal rearrangements deduced as described [14]. Statistical tests were performed using R (http: //www. r-project. org). Linear regressions were used to model mutation frequency as a function of chronological and biological age. Differences in mutation frequency between body parts were evaluated using the Welch two-sample t-test. | DNA mutations are changes in the DNA sequence, including basepair substitutions and genome rearrangements. Mutations are inevitable and are a consequence of errors made in replicating DNA during cell division or repairing damage. Aging is a complex process of functional decline and increased disease risk that has often been ascribed to the accumulation of DNA mutations. In the past we generated a mouse model that allows quantification and characterization of mutations in somatic tissue in a transgene that can be removed and studied in Escherichia coli. The results indicated that mutations accumulate with age in a tissue-specific manner. Using a recently generated similar model for Drosophila melanogaster, we now show that mutations also accumulate with age in this invertebrate organism. However, the frequency of mutations, already at young age, was found to be much higher than in mice, with a much higher fraction of genome rearrangements. Mutations accumulated much more quickly at higher temperature, a condition which in flies is associated with decreased life span. Hence, we conclude that the burden of DNA mutation is much heavier in short-lived flies than in mammals and increases with age as a function of biological rather than chronological age. | Abstract
Introduction
Results
Discussion
Materials and Methods | genetics and genomics | 2010 | Age- and Temperature-Dependent Somatic Mutation Accumulation in Drosophila melanogaster | 5,385 | 275 |
Anopheles balabacensis of the Leucospyrus group has been confirmed as the primary knowlesi malaria vector in Sabah, Malaysian Borneo for some time now. Presently, knowlesi malaria is the only zoonotic simian malaria in Malaysia with a high prevalence recorded in the states of Sabah and Sarawak. Anopheles spp. were sampled using human landing catch (HLC) method at Paradason village in Kudat district of Sabah. The collected Anopheles were identified morphologically and then subjected to total DNA extraction and polymerase chain reaction (PCR) to detect Plasmodium parasites in the mosquitoes. Identification of Plasmodium spp. was confirmed by sequencing the SSU rRNA gene with species specific primers. MEGA4 software was then used to analyse the SSU rRNA sequences and bulid the phylogenetic tree for inferring the relationship between simian malaria parasites in Sabah. PCR results showed that only 1. 61% (23/1,425) of the screened An. balabacensis were infected with one or two of the five simian Plasmodium spp. found in Sabah, viz. Plasmodium coatneyi, P. inui, P. fieldi, P. cynomolgi and P. knowlesi. Sequence analysis of SSU rRNA of Plasmodium isolates showed high percentage of identity within the same Plasmodium sp. group. The phylogenetic tree based on the consensus sequences of P. knowlesi showed 99. 7%–100. 0% nucleotide identity among the isolates from An. balabacensis, human patients and a long-tailed macaque from the same locality. This is the first study showing high molecular identity between the P. knowlesi isolates from An. balabacensis, human patients and a long-tailed macaque in Sabah. The other common simian Plasmodium spp. found in long-tailed macaques and also detected in An. balabacensis were P. coatneyi, P. inui, P. fieldi and P. cynomolgi. The high percentage identity of nucleotide sequences between the P. knowlesi isolates from the long-tailed macaque, An. balabacensis and human patients suggests a close genetic relationship between the parasites from these hosts.
Anopheles species of the Leucosphyrus group have been identified as medically important vectors in Southeast Asia region [1,2]. The Leucosphyrus group has three main subgroups; Hackeri, Leucosphyrus and Riparis subgroups [3], with the Leucosphyrus subgroup further divided into Dirus complex and Leucosphyrus complex [2,4]. In Peninsular Malaysia, three species of the Leucosphyrus group namely An. hackeri, An. cracens and An. introlatus had been incriminated as primary vectors for P. knowlesi [5–7]. However, in East Malaysia, An. latens in Sarawak and An. balabacensis in Sabah had been confirmed as primary vectors for P. knowlesi [8,9]. A study in Cambodia in 1962 has shown that An. balabacensis (identified as An. dirus later [10]) preferred biting human compared to monkeys placed at the ground level, but preferred monkeys at canopy level to monkeys on the ground [11]. A study in Sabah comparing human landing catch (HLC) and monkey baited trap (MBT) at ground level showed that more An. balabacensis were caught using HLC than MBT [12]. Recent studies showed that this species is more active during the early night with a peak biting time between 7 pm to 8 pm [9,13], and also prefers to bite outdoors than indoors [13]. Such biting behaviors coupled with an abundant source of simian malaria parasites in the reservoir long-tailed macaques (Macaca fascicularis) contribute to An. balabacensis becoming an effective vector for transmitting P. knowlesi malaria in Sabah. Previous studies in Malaysia have shown that the long-tailed macaques harbor at least five species of simian Plasmodium [14,15], all of which have also been detected in An. balabacensis [9,16]. In Sabah, besides P. knowlesi, other simian malaria parasites recorded in An. balabacensis are P. coatneyi, P. inui, P. fieldi and P. cynomolgi [9,13]. Apart from recording these parasites in the mosquitoes, there is limited study on the phylogenetic relationship among these simian malaria parasites found in An. balabacensis, macaques and human. In this study, we compare the partial nucleotide sequences of SSU rRNA of simian malaria parasites isolated from An. balabacensis caught in Kudat district of Sabah, from macaques as well as human patients with other published sequences of human and simian malaria parasites available in the GeneBank database. Building a phylogenetic tree of these malaria parasites will give us a clearer picture about their genetic relationship especially for P. knowlesi isolated from long-tailed macaque, An. balabacensis and human.
Kudat district, located at the northern tip of Borneo under the Kudat Division, is about 153 kilometers from Kota Kinabalu, the state capital of Sabah. Paradason village where the study was conducted is located in Kudat District and about 50 kilometers from Kudat town (Fig 1). Most of the villagers belong to the Rungus ethnic group who are dependent on small-scale farming (paddy), oil palm and rubber plantations as their primary source of income. Anopheles mosquitoes were sampled monthly from October, 2013 to December, 2014 using human landing catch (HLC) method. A total 70 nights of sampling were performed starting from 1800 to 0600 hours (12 hours). Two pairs of volunteers were assigned working in shifts at a randomly selected habitat during each night of sampling. Anopheles was lured by the volunteers exposing their legs. The mosquitoes landing on the legs were caught by the volunteers using plastic specimen tubes (2 cm diameter X 6 cm) aided by a flashlight. The next morning, the Anopheles mosquitoes were killed by keeping them in the freezer (-20°C) for a few minutes, then gently pinned onto Nu poly strip using ultra-thin micro-headless pins. Species identification was done under a compound microscope using published keys [2,17,18]. After identification, each individual specimen was stored separately in a new microfuge tube and transported to Faculty of Medicine & Health Sciences, Universiti Malaysia Sabah for further processing. Each individual Anopheles specimen was placed separately inside a sterilized mortar and the tissue homogenized using a sterile pestle. The total DNA was extracted from the tissues using DTAB-CTAB method [19] with some modifications (for example: incubation time was reduce to 30 minute instead of overnight and at the final step of precipitation before adding TE buffer, DNA pellet was incubated at 45°C to completely evaporate any residue of ethanol). First, 600 μl of DTAB solution was added into the mortar and the tissue was ground using pestle until homogenized. Then, the homogenized tissue was transferred into a clean 1. 5 ml microfuge tube and incubated at 68°C for 30 min. Subsequently, 600 μl of chloroform was added into the microfuge tube which was inverted ten times to mix the contents and centrifuged at 13,000 rpm for 5 min. Then, 400 μl of the upper aqueous layer was carefully transferred into a new clean 1. 5 ml microfuge tube and mixed with 900 μl sterile dH2O and 100 μl CTAB solution by gently inverting the microfuge tube for several times and allowed it to sit at room temperature for 5 min. The mixture was then spun at 13,000 rpm for 10 min. The supernatant was discarded and the DNA pellet was re-suspended in 300 μl of 1. 2 M NaCl solution. Total DNA was precipitated by adding 750 μl of absolute ethanol and centrifuged at 13,000 rpm for 5 min. The supernatant was discarded, the DNA pellet washed with 500 μl of 70% ethanol and centrifuged at 13,000 rpm for 2 min. The DNA pellet was incubated at 45°C for 10 min and re-suspended in 30 μl Tris-EDTA (pH8. 0) buffer and stored at -30°C. Presence of malaria parasites in the mosquitoes was detected using nested PCR by targeting the small subunit ribosomal RNA (SSU rRNA) gene of Plasmodium. A PCR primer pair, rPLU1 and rPLU5, was used in first PCR reaction, while another pair (rPLU3 and rPLU4) was used in the second PCR reaction [20]. For internal control, another set of nested PCR was performed separately to amplify the cytochrome c oxidase subunit II (COII) gene of Anopheles [12]. When a mosquito was confirmed positive for malaria parasites, the Plasmodium species was determined using species specific primers. Both PCR reactions were performed with 25. 0 μl final volume. The reaction components were prepared by mixing 5. 0 μl of 5X PCR buffer (Promega), 0. 5 μl of (10 mM) dNTPs (Promega), 3. 0 μl of (25 mM) MgCl2,1. 0 μl of (10 μM) forward and reverse primers, 0. 3 μl of (5. 0 U/μl) Taq DNA polymerase (Promega), 2. 0 μl of DNA template and sterile dH2O to make up to 25. 0 μl final volume. After completion of the first PCR, 2. 0 μl of the PCR product was used as DNA template in the second PCR. The reaction was carried out using a thermal cycler (T100 Thermal Cycler, BioRad) with an initial denaturation at 95°C for 5 min followed by 35 cycles of denaturation at 94°C for 1 min, annealing for 1 min and extension at 72°C for 1 min and one final extension step at 72°C for 5 min. The annealing temperature was set at optimal temperature for each set of primers (see S1 Table). The PCR products were analyzed on 1. 5% agarose gel electrophoresis stained with RedSafe nucleic acid staining solution (iNtRON Biotechnology), and visualized with an UV transilluminator. The SSU rRNA gene of the five simian malaria parasite species extracted from An. balabacensis caught in Paradason were cloned and sequenced. In addition, we included in the study blood samples from two P. knowlesi patients and two long tail macaques, one infected with P. knowlesi while the other with P. inui. To make the data set larger, we included simian malaria parasites obtained from mosquitoes caught in three other villages (Tomohon, Mambatu Laut and Narandang) in Kudat district from another study. A new universal forward primer (UMSF) combined with species-specific primers were used to amplify the SSU rRNA gene of Plasmodium. Details of the primers are provided in S2 Table. Preparation of the reaction mixture and the PCR conditions programmed are as described above. After the PCR was completed, the PCR products were purified to remove impurity and excess reaction mixture using MEGA quick-spin PCR & Agarose Gel DNA Extraction System (iNtRON Biotechnology, Korea) according to manufacturer’s procedure. Cloning the SSU rRNA gene was done using pGEM-TEasy vectors (Promega, USA) and the plasmids were extracted from the transformed E. coli (JM109) using DNA-spin Plasmid DNA Purification Kit (iNtRON Biotechnology, Korea), all according to the manufacturer’s protocol. The extracted plasmid vectors were restricted using EcoRI restriction enzyme (Promega, USA) and sent to AITBIOTECH, Singapore for sequencing. Sequencing was carried in both directions using forward and reverse M13 primers. The nucleotide sequences of SSU rRNA of 21 Plasmodium isolates in this study were aligned and compared with other SSU rRNA sequences available at the GeneBank database to determine the percentage identity using Basic Local Alignment Search Tool (BLAST) available online at https: //blast. ncbi. nlm. nih. gov/Blast. The SSU rRNA sequences were standardized to a fixed region for analysis based on the UMSF and UNR primers binding sites. Further analysis was performed using MEGA software, version 4. 1 [21]. The nucleotide sequences were multi-aligned using ClustalW method [22] incorporated in the software and the number of variable nucleotides within each of the five Plasmodium spp. determined. Phylogenetic tree was constructed using neighbor-joining method [23] and the evolutionary distances computed using maximum composite likelihood model with a bootstrap test of 1000 replicates [24] and pairwise deletion option. This method was adopted as it takes into account the different rates of evolution or substitution between nucleotides. The selected region for constructing the phylogenetic tree was nucleotides numbered nt81 to nt1041, based on the published P. knowlesi sequence (AY327551) isolated in Kapit Sarawak where there was a large focus of infected people [25]. This region includes the binding sites for universal forward (UMSF, used in this study) and reverse primers (UNR, [26]) of SSU rRNA. In constructing the phylogenetic tree, Theileria spp. (AF162432) was used as the outgroup. Details of the other 66 nucleotide sequences that were used in constructing the phylogenetic tree are given in S3 Table. Both Plasmodium simium (AY579415) and P. brasilianum (AF130735, KT266778) were not included in the sequence analysis as the selected sequence used in this study was not available in GeneBank database. A second phylogenetic tree was constructed using the consensus sequences of five Plasmodium species found in Sabah to show the relationship between Plasmodium isolates found in the macaque, An. balabacensis and human. This project was approved by the National Medical Ethics Committee (NMRR), Ministry of Health Malaysia (Ref. NMRR-12-786-13048). All volunteers who carried out mosquito collections signed informed consent forms and were provided with antimalarial prophylaxis during the study period. Blood spots on Whatman filter paper were collected from adult patients by Kudat hospital personnel, after they had signed informed consent forms. This human blood sample collection was also approved by the NMRR (Ref. NMRR–11–4539471). Blood spots on filter paper were collected by wild life department personel from ten wild macaques captured for relocation purposes and kept in cages following the guidelines in the Animals (Scientific Procedures) Act 1986 Code of Practice for the Housing and Care of Animals Used in Scientific Procedures (UK), with the approval from the London School of Hygiene and Tropical Medicine Animal Welfare and Ethical Review Body (AWER, Ref. 2012/8N). Fecal samples were not used then as the protocol for storing the samples had not yet been established by primatology group of the research team.
A total of 1,599 Anopheles individuals belonging to ten species were caught during 14 months of sampling (Table 1). Anopheles balabacensis was the dominant species in Paradason village comprising 89. 87% of the total catch, followed by An. barbumbrosus (5. 75%), An. maculatus (1. 38%) and An. donaldi (1. 19%). A total of 1,586 Anopheles mosquitoes (of which 1,425 were An. balabacensis) were tested for presence of malaria parasites using the PCR method. Only 23 An. balabacensis (1. 61%) were found to have malaria parasites in them, being infected with one (78. 3%) or two simian Plasmodium spp. (Table 2). The single infection was mostly by P. inui (n = 11). BLAST analysis of 21 SSU rRNA sequences of Plasmodium spp. isolated from An. balabacensis, human and long tail macaques (3 samples of P. coatneyi, 1027–1029 bp; 4 samples of P. cynomolgi, 1015 bp; 3 of P. fieldi, 1039 bp; 6 of P. inui, 1039 bp and 5 of P. knowlesi, 1050 bp) showed high percentage of identity with the simian Plasmodium nucleotide sequences published in the GeneBank database. The Plasmodium species in Sabah show a high percentage identity within the same species groups (98. 4%–99. 6%) but less between different species groups. The highest percentage identity (99. 6%) was observed between the P. cynomolgi samples isolated from Tomohon, Membatu Laut and Paradason villages, while the least was for P. coatneyi isolates (98. 4%) obtained from Narandang and Paradason villages. The SSU rRNA sequences of Plasmodium spp. from Sabah also show high percentage identity with the same species from other Asian regions. Plasmodium coatneyi sequences showed 99% identity with P. coatneyi isolated from M. fascicularis in Kapit, Sarawak (FJ619094), as well as with CDC (AB265790) and Hackeri (CP016248) strains. Plasmodium cynomolgi sequences showed 99%–100% identity with P. cynomolgi isolated from M. fascicularis in Kapit, Sarawak (FJ619084), and from other macaque species viz. M. radiata (AB287290) of southern India and M. nemestrina (AB287289) from unspecified South-east Asian nation. Similarly, P. fieldi has high percentage identity with P. fieldi isolated from M. fascicularis in Kapit, Sarawak (KC662444). Of interest is P. inui, which not only has high identity (99%–100%) with those isolated in Kapit (FJ619074) but also with P. inui isolated from M. fascicularis from South China (HM032051), Southern Thailand (EU400388) and strain Taiwan II isolate from M. cyclopis (FN430725). The P. knowlesi samples of Sabah showed 99% identity with P. knowlesi isolated from both human (AY327551) and M. fascicularis (FJ619089) in Kapit, Sarawak, as well as with that from a Swedish traveler who was infected during his visit to Sarawak (EU807923) [27]. The number of nucleotides in the analyzed region for the various Plasmodium spp. are: P. knowlesi 961 bp, P. inui 946, P. coatneyi 942, P. cynomolgi 935 and P. fieldi 934 respectively. Sequence alignment indicated that P. coatneyi has the highest number of variable nucleotides among the isolates (n = 3 isolates; 15 variable nucleotides) followed by P. knowlesi (n = 5; 9), P. inui (n = 6; 7), P. fieldi (n = 3; 5) and P. cynomolgi (n = 4; 4). Further analysis of the P. knowlesi group using consensus sequences showed that there were three variable nucleotides between P. knowlesi isolated from the long-tailed macaque and human, two between long-tailed macaque and An. balabacensis isolates but none between An. balabacensis and human isolates (Fig 2). In the phylogenetic tree generated for 13 Plasmodium species infecting monkeys and humans (Fig 3), all the 21 Plasmodium isolates obtained in the study were placed in the correct species group. P. knowlesi group was positioned below P. coatneyi group whereas P. inui, P. fieldi and P. cynomolgi were placed at the upper branches. In the phylogenetic tree depicting relationship between the five Plasmodium species found in Sabah using consensus sequences, a similar tree topology was also observed (Fig 4). All Plasmodium group except for P. knowlesi group has two branches, each representing the host from which Plasmodium was isolated. However, P. knowlesi group has three branches with the isolates from both An. balabacensis and macaque closer to each other than to the isolates from humans.
In this study, we analyzed 21 nucleotide sequences of partial SSU rRNA of five Plasmodium spp. isolated from An. balabacensis collected in Kudat district of Sabah, infected humans and a long-tailed macaque together with other nucleotide sequences downloaded from the GeneBank database. The results suggest that in Sabah, there is a close genetic relationship between the P. knowlesi specimens in the long-tailed macaques, An. balabacensis and human. Plasmodium inui appears to be a common simian malaria parasite found in 61% (14/23) of the infected An. balabacensis specimens. This was also the case in other investigations [9,28]. Hitherto, this simian malaria has not become zoonotic to humans yet although it has been proven experimentally to be infective to monkey through the bites of An. dirus [29]. The infection rate of P. knowlesi in An. balabacensis is low (0. 14%, 2/1,425) with only two mosquitoes being infected along with other Plasmodium species. Nevertheless P. knowlesi is the dominant Plasmodium species recorded among the human cases in Sabah [30]. These cases were recorded mainly in the rural areas near to forests and also among the workers in the agricultural sector viz. in oil palm estates and vegetable farms [13,31]. Sequence data of the SSU rRNA of Plasmodium confirm that the five species of simian Plasmodium commonly harbored by the wild macaques in Malaysia are also found in An. balabacensis. BLAST results of Sabah’s Plasmodium sequences showed high identity with other simian Plasmodium sequences published in the GeneBank database, especially with the simian malaria parasites in long-tailed macaques in Kapit, Sarawak (FJ619069 and FJ619089). This could suggest that a similar or closely related cluster of simian Plasmodium is circulating among the monkey populations and Anopheles mosquitoes in both Sabah and Sarawak. This is highly plausible as these two states share a common boundary, and there is a continual movement of humans between these two states. The total number of nucleotides in the analyzed region was different for the five simian Plasmodium spp. in Sabah, with P. knowlesi having a higher number. The differences in total number of nucleotides in the SSU rRNA gene confer a unique signature to each Plasmodium species. Furthermore the presence of conserved and variable sequences in the gene makes it suitable for species identification and phylogenetic study [32,33]. The percentage of identity between consensus sequences of SSU rRNA of P. knowlesi isolates from the monkey, mosquito and man was high (Fig 2). For example, 100% identity was observed between P. knowlesi isolates from An. balabacensis and human, 99. 8% between An. balabacensis and the long-tailed macaque, and 99. 7% between long-tailed macaque and human. This indicates a great genetic similarity in P. knowlesi found in the long-tailed macaque, An. balabacensis and human populations. However, it is not certain if this would indicate the same cluster of P. knowlesi is circulating between these hosts, since we did not dissect the mosquitoes’ salivary glands to detect for sporozoites, or carry out RT-PCR targeting the specific mRNA transcripts of the sporozoite stage. Thus further study is needed to determine this, using more P. knowlesi positive Anopheles balabacensis and analyzing other polymorphic markers or microsatellite loci of the parasite. Different P. knowlesi haplotypes have been observed in the macaque and human populations in Kapit Sarawak [14] as well as in the human population in Thailand [34]. Overall, the 13 Plasmodium species in the phylogenetic tree can be grouped into two main clusters, one containing the P. vivax/simian malaria parasites while the other human malaria parasites (Fig 3). Although P. simium (AY579415) and P. brasilianum (AF130735, KT266778) were not included in our analysis as their nucleotide sequences in the GeneBank database do not contain the same analyzed region, P. simium is closely related to P. vivax [32]and can be placed in the first cluster, while P. brasilianum is closely related to P. malariae and can be placed in the second cluster. It may be noted that P. cynomolgi, P. fieldi and P. simiovale were not clearly resolved as some of the isolates were grouped in different branches. This could be due to the high percentage of nucleotide identity (99. 6%) among these three species. The consensus tree (Fig 4) of Plasmodium species found in Sabah showed a very close relationship between the Plasmodium isolates from monkey as the reservoir, An. balabacensis as the vector, and human as the case. This is supported by P. knowlesi isolates from these three organisms having high nucleotide identity (99. 7–100%). Currently in Sabah, An. balabacensis is the only species found to carry P. knowlesi. The phylogenetic analysis here indicates that the vector picks up the malaria parasites from monkeys and transmits them to humans when it feeds on them. However, there is a lot more about the transmission dynamics of P. knowlesi that is still unknown and needs to be unpacked. A clearer picture on the interrelationship of simian malaria parasites found in An. balabacensis will help us to understand more about Plasmodium itself. Future research may focus more on the host-vector relationship that requires longer nucleotide sequence analysis so that new informed alternatives for malaria elimination strategy targeting on P. knowlesi as well as other simian malaria parasites may be formulated. | Anopheles balabacensis has been incriminated as the primary vector of zoonotic simian malaria, P. knowlesi in Malaysian Borneo with a high prevalence recorded in the states of Sabah and Sarawak. In this study, Anopheles spp. were sampled using human landing catch (HLC) method at Paradason village in Kudat district of Sabah. Total DNA was extracted from these specimens, followed by sequencing the SSU rRNA gene of Plasmodium using polymerase chain reaction (PCR) for the detection and identification of Plasmodium. PCR results showed that only 1. 61% (23/1,425) of the screened An. balabacensis had either single or double Plasmodium spp infections. The simian malaria parasites isolated from An. balabacensis were P. coatneyi, P. inui, P. fieldi, P. cynomologi and P. knowlesi. Sequence analysis of these Plasmodium isolates showed high percentage of identity within the same Plasmodium sp. group. Consensus sequences phylogenetic tree of P. knowlesi isolates from An. balabacensis, human patients and a long-tailed macaque from the same locality had 99. 7%–100. 0% nucleotide identity. This study suggests a close genetic relationship between the parasites isolated from these hosts. | Abstract
Introduction
Materials and methods
Results
Discussion | taxonomy
invertebrates
parasite groups
medicine and health sciences
evolutionary biology
plasmodium
tropical diseases
parasitic diseases
animals
parasitic protozoans
parasitology
apicomplexa
phylogenetics
data management
protozoans
phylogenetic analysis
insect vectors
cellular structures and organelles
research and analysis methods
sequence analysis
infectious diseases
computer and information sciences
malarial parasites
bioinformatics
biological databases
evolutionary systematics
disease vectors
insects
ribosomes
arthropoda
biochemistry
rna
mosquitoes
eukaryota
dna sequence analysis
sequence databases
ribosomal rna
nucleic acids
cell biology
database and informatics methods
biology and life sciences
species interactions
malaria
non-coding rna
organisms | 2017 | Phylogenetic analysis of simian Plasmodium spp. infecting Anopheles balabacensis Baisas in Sabah, Malaysia | 6,621 | 344 |
AMPA receptors (AMPARs) and their associations with auxiliary transmembrane proteins are bulky structures with large steric-exclusion volumes. Hence, self-crowding of AMPARs, depending on the local density, may affect their lateral diffusion in the postsynaptic membrane as well as in the highly crowded postsynaptic density (PSD) at excitatory synapses. Earlier theoretical studies considered only the roles of transmembrane obstacles and the AMPAR-binding submembranous scaffold proteins in shaping receptor diffusion within PSD. Using lattice model of diffusion, the present study investigates the additional impacts of self-crowding on the anomalousity and effective diffusion coefficient (Deff) of AMPAR diffusion. A recursive algorithm for avoiding false self-blocking during diffusion simulation is also proposed. The findings suggest that high density of AMPARs in the obstacle-free membrane itself engenders strongly anomalous diffusion and severe decline in Deff. Adding transmembrane obstacles to the membrane accentuates the anomalousity arising from self-crowding due to the reduced free diffusion space. Contrarily, enhanced AMPAR-scaffold binding, either through increase in binding strength or scaffold density or both, ameliorates the anomalousity resulting from self-crowding. However, binding has differential impacts on Deff depending on the receptor density. Increase in binding causes consistent decrease in Deff for low and moderate receptor density. For high density, binding increases Deff as long as it reduces anomalousity associated with intense self-crowding. Given a sufficiently strong binding condition when diffusion acquires normal behavior, further increase in binding causes decrease in Deff. Supporting earlier experimental observations are mentioned and implications of present findings to the experimental observations on AMPAR diffusion are also drawn.
Glutamate-binding transmembrane alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid (AMPA) -type receptors (AMPARs) are the pivotal element of fast synaptic transmission at excitatory synapses in the central nervous system [1,2]. At the site of synaptic contact, these receptors are present at high density within a specialized region on the postsynaptic membrane, termed as postsynaptic density (PSD), which is closely apposed to the presynaptic active zone of glutamate release [3–5]. The remaining extra-synaptic region of the postsynaptic membrane is distinguished with a relatively lower density of AMPARs [4]. This unique spatial arrangement of AMPARs is a natural adaptation to expose a sufficiently large number of the receptors to sufficiently high glutamate concentration in the synaptic cleft [6–8] and, thus, enhance the postsynaptic response, which would otherwise be comparatively weaker under a homogeneously distributed receptor condition [9–11]. Like other mobile transmembrane proteins, AMPARs also exhibit lateral diffusion in the postsynaptic membrane [12,13]. AMPAR mobility is crucial for many essential processes associated with the efficiency of synaptic functioning. Lateral diffusion causes continuous exchange of the receptors between PSD and extra-synaptic region [14,15]. This exchange brings about replacement of desensitized AMPARs in the PSD with active AMPARs of the extrasynaptic region after an event of glutamate release and, hence, assists in maintaining the strength of consecutive postsynaptic responses in the presence of high-frequency presynaptic spike train [16–18]. It also underlies the recruitment of new AMPARs into the PSD which are brought by exocytotic vesicles from the local intracellular reserve-pool and are initially unloaded on the extra-synaptic membrane [19,20]. In a similar manner, the older receptors in the PSD diffuse out to the extra-synaptic region where they are endocytosed [21]. Accordingly, lateral diffusion of AMPARs assists in the appearance of long-term potentiation (LTP) [20] or long-term depression (LTD) [22,23] at excitatory synapses and, hence, assists in the molecular basis of learning. Owing to such a crucial and indispensable role of AMPAR lateral diffusion in shaping the density and spatial localization of AMPARs in the PSD, it has always engaged attention of a wide scientific community. A remarkable thing is the gathering of AMPARs in the PSD despite that PSD spans a much smaller area than the extra-synaptic region and the AMPARs are significantly mobile. It seems that the receptors get trapped in the PSD while diffusion because, in the absence of trapping, diffusion would lead to a homogeneous distribution of AMPARs on the entire postsynaptic membrane. In the backdrop of continuous receptor exchange between PSD and extra-synaptic region, the trapping can be viewed in terms of the considerably longer residence time [24] of an AMPAR in the PSD. Using techniques like fluorescence recovery after photobleaching (FRAP), electrophysiology with mutant variants of AMPAR, and different versions of single particle tracking such as sptPALM, uPAINT and quantum dot (QD) -tagging of receptors, various experimental studies [24–29] in hippocampal slices and live hippocampal neurons in dissociated cultures have provided a wealth of observations on the nature of AMPAR diffusion at excitatory synapses. In conjunction with the theoretical investigations [29–34], these studies have so far clearly shown that the molecular composition of the PSD is indeed responsible for the trapping of diffusing AMPARs. The PSD is rich in large number of transmembrane as well as submembrane proteins [3,35–37], owing to which it possess a very high molecular weight [37,38]. The crowding of inert transmembrane proteins strongly obstructs the AMPAR diffusion within the PSD region through steric repulsion [29,31]. Further, the reversible binding of intracellular domain of AMPARs as well as their associations with transmembrane AMPAR regulatory proteins (TARPS) to the submembranous scaffold proteins, such as PSD-95, also substantially reduces the AMPARs mobility [29,39,40]. An AMPAR is a tetramer and is consist of any combination of the four kinds of subunits GluR1, GluR2, GluR3 and GluR4 [41,42]. Typically, GluA1-GluA2 and GluA2-GluA3 heterotetramers are most abundant in the adult brain [43–45]. These receptors are very bulky [41] and carry along a large steric-exclusion volume. The bulkiness of AMPARs is further increased due to the various auxiliary proteins [46–50] associated with it. In fact, the size of the native complexes of AMPARs isolated through biochemical techniques have been found to be approximately double the original size of the tetramer [51]. Further, AMPARs reside at high density in the PSD and contributes to a substantial fraction of the local macromolecular crowding [3,37]. Therefore, it is reasonable to envisage that these receptors may block the diffusion paths of each other and may lead to a situation of self-obstruction or self-crowding. Moreover, the distribution of AMPARs in the PSD is not strictly homogeneous. Rather, there are smaller subregions or nanodomains within the PSD which have more AMPARs cluttered [52,53] and the self-crowding of these receptors would be more pronounced. Accordingly, besides inert transmembrane protein crowding and binding to scaffold proteins, self-crowding of AMPARs may appear an additional factor behind the reduced or hampered mobility and trapping of these receptors in the PSD. However, in the earlier experimental and theoretical studies, the possible role of self-crowding factor has remained completely unaddressed. The above speculation regarding self-crowding of AMPARs is a seemingly interesting issue and, therefore, is the source of motivation for carrying out the present theoretical investigation. The effect of various crowding factors on the AMPAR diffusion can only be enquired through detailed numerical simulation of independent diffusing receptors. Therefore, the present study involves the Monte Carlo simulation of receptor diffusion using lattice model of diffusion, which has proven to be an effective approach in the earlier theoretical studies [31,54,55]. The main body of the present study is comprised of a purely abstract framework with a lattice used as a generalized spatially-discrete medium of diffusion, regardless of whether the lattice represents the entire PSD or a subregion within the PSD. Moreover, the AMPARs are represented by point diffusion tracers (DT) on the lattice [31,54]. On the basis of the ensemble-averaged mean-squared displacement of tracer diffusion, the nature of diffusion is established in terms of two physical quantities viz. anomalousity and effective diffusion coefficient under the different pertinent conditions of crowding and binding events. Both the quantities serve as the suitable marker of resulting dwell time of the receptors within PSD and, hence, can be effectively used to comprehend receptor trapping [31]. It must be noted that the dynamics of the AMPAR accumulation in the excitatory PSD and exchange with extrasynaptic region is not the immediate interest of the present study. Rather, it focusses on capturing the emergent statistical behaviors of the receptor diffusion in the thermodynamic limit when self-crowding is considered in addition to the other obstacles and binding, which may be later used as the building block to comprehend the dynamics of accumulation. The findings reveal that even in the absence of any steric crowding of other transmembrane and scaffold proteins in the postsynaptic membrane, very high density of AMPARs may itself lead to extraordinarily high anomalousity and reduced diffusion coefficient. Remarkably, anomalousity of receptor diffusion may also exhibit a switch-like behavior with respect to their self-crowding density, similar to the switch-like behavior with respect to increase in steric macromolecular crowding of other PSD proteins observed earlier [31] as well as here. Further, increase in the crowding by other PSD proteins may exacerbate the anomalousity and decline in diffusivity arising from self-crowding. Contrarily, binding appears to mark a reverse effect by decreasing the anomalousity of crowded receptor diffusion. The plausible mechanisms underlying these findings are discussed to details. Moreover, the relevance of the use of point tracers in capturing the picture of self-crowded diffusion of the non-zero lateral-sized AMPARs is also drawn. Eventually, the possible elements of self-crowding lying within the earlier experimental observations on the nature of AMPAR diffusion at excitatory synapses are pointed out and the physiological relevance of the present observations made through the abstract framework is established in regard of real biological scenario.
The lateral diffusion of AMPARs is realized here through the numerical simulation of the diffusion of DTs. For simplicity, the entire macromolecular crowding at the PSD can be broadly classified into two pools [31]: Completely-Reflecting Obstacles (CROs) and Partially-Reflecting-cum-Binding Obstacles (PROs) (Fig 1A). In general, the CROs represent various transmembrane proteins in the PSD [3,35] which interact with a diffusing AMPAR only to reflect it away on collision via. steric repulsion. Since they never significantly bind the AMPARs, they behave as inert obstacles [31,54]. On the other hand, the PROs are often the submembranous scaffold proteins which are intracellularly accumulated close to the PSD [3,35]. These proteins generally offer only a partial obstruction to the diffusing AMPARs through steric repulsion of the intracellular C-terminal domains of the receptors [31]. Moreover, they preferentially bind the receptors almost at their location through reversible non-covalent interactions between specific domains of the receptors and the scaffold proteins. For instance, GluR1 subunit of AMPARs can directly bind to the scaffolding SAP-97 proteins [56] whereas GluR2 subunits can bind to submembranous PICK1 or GRIP [57]. However, AMPARs cannot directly interact with the one of the most abundant PSD-95/SAP-90 scaffold proteins at excitatory synapses due to their incompatible PDZ domains. Rather, the receptors require association with auxiliary transmembrane AMPA receptor regulatory proteins (TARPs), such as Stargazin, to bind with PSD-95 [39]. Accordingly, the probability of reflection (Preflect) for tracer collision with CROs is always 1 [31] whereas it may be variable for collision with PROs, as it depends on the size of the C-terminal domains of AMPAR subunits, the size of the specific scaffold proteins, presence of auxiliary proteins etc. For the present investigation, Preflect for collision with PROs is kept fixed at 0. 5, signifying a 50% chance of reflecting the trajectory of a tracer on physical contact without binding [31]. Furthermore, the detailed multi-step kinetic scheme for the binding of AMPARs to scaffold proteins is still unknown. However, the intensity of binding through hydrogen bond interactions within PDZ domains have been estimated to be in the range of 2–13kBT [31,58–60], which may serve here as a rough estimate for the binding energy of AMPAR-scaffold interactions. Accordingly, three specific situations of homogeneous DT-PRO binding viz. weak, intermediate and strong bindings with binding energy 2,6 and 10 kBT, respectively, are taken into account in the present investigation. Further, a conventional approach is to assume the entire crowding factors to be static at their spatial locations in the PSD while an AMPAR diffuses through the crowd [29,31]. This approach is reasonably correct to a great extent as the mobility of the crowding factors is too low [61,62] in comparison to that of AMPARs and their average life-time in the PSD (in hours) [63] is substantially high relative to the typical measurement time-duration of AMPAR diffusion (in seconds). Accordingly, the obstacles CROs/PROs are considered here immobile and their density preserved throughout the duration of tracer diffusion. A two-dimensional square lattice (Fig 1B) is considered for performing the lateral diffusion of AMPARs in the postsynaptic membrane [31,54,55,64]. The lateral diffusion coefficient of an AMPAR in the extra-synaptic membrane is known to be almost 0. 2 × 10−3μm2. ms−1 [24,31,65]. Since the extrasynaptic membrane offers least macromolecular obstructions relative to the PSD, this estimate is considered here as the natural free diffusion coefficient of an AMPAR in an unobstructed lipid medium of the postsynaptic membrane and is here assigned to the effective diffusion coefficient of DT (Deff). The diffusion is performed at discrete time-steps Δt of fixed size 10−3ms [31]. The mean-squared displacement (MSD), 〈r2〉 (t), of a DT undergoing free normal diffusion in a two-dimensional medium is given by, 〈 r 2 〉 (t) = 4 D e f f t (1) Therefore, using Eq 1, the desired finite diffusion length Δl for the lattice diffusion can be computed to be 8. 9 × 10−4μm [31]. This estimate is assigned to the edge length between any two lattice-sites in the square lattice. In this way, the size of the entire lattice in terms of the number of lattice-sites is kept 1119 × 1119 such that the lattice approximates to an area of 1μm2 [31]. It must be noted that the lattice employed here is a purely abstract framework to procure the salient features of the tracer diffusion under different crowding conditions. Therefore, depending on the requirement, it may be used to address the properties of AMPAR diffusion over the entire PSD as well as within a subregion of the PSD. The two kinds of obstacles, CROs and PROs, are considered as point obstacles over the lattice (Fig 1B). The CROs and PROs are uniformly distributed over the lattice according to their desired area fractions aCRO and aPRO, respectively. The two classes of obstacles are dealt separately so that how these obstacles of different nature may affect tracer diffusion in their specific manners can be clearly examined. Since the main objective of the present study is to investigate the effect of self-crowding of tracers on their lateral diffusion, a standard situation of “no-self-crowding” is also taken into account which serves as a benchmark for comparative analysis of the observations made under varying self-crowding situations. In this standard situation, DTs are uniformly placed on the square lattice where each tracer behaves as an independent diffusing entity. Therefore, while diffusion, two or more tracers can together occupy the same lattice site. No steric exclusion among the tracers is considered. In fact, this constitutes an ensemble of multiple copies of independently diffusing tracers but with different initial positions on the lattice under an identical distribution of CROs or PROs. While initially placing DTs on the lattice, it is taken care that a tracer should not lie at a lattice site already occupied by a CRO whereas it is allowed to lie on the lattice-site occupied by a PRO. Moreover, in the case of PROs, the tracers are allowed to diffuse for 2s after being initially placed to acquire thermal equilibrium. Only after this annealing period, the measurement of tracer diffusion is performed [55]. However, while dealing with the self-crowding conditions, the density of diffusing tracers placed on the square lattice would also matter (Fig 1B). Accordingly, the area-fractions of the lattice occupied by DTs, aDT, are taken in the increasing orders of the magnitudes such that conditions of six different aDT viz. 0. 00001,0. 0001,0. 001,0. 005,0. 01 and 0. 1, are investigated in the present study. Here too, the DTs are uniformly distributed at the desired aDT and considerations regarding their initial placement on the lattice depending on the CROs or PROs are taken care as described for the standard situation. A periodic boundary condition is imposed on the boundary of the lattice mesh [31,54]. Therefore, as a tracer leaves the mesh, it re-enters the mesh from the exact opposite side. Monte-Carlo simulation of tracer diffusion is performed. At each time-step, all the tracers present over the mesh are inspected for diffusion one by one. For each tracer, a random number is generated from the uniform random number distributed over the interval [0,1] to decide the direction of its diffusion. If the random number is < 0. 25, the tracer would move left. If the random number is ≥ 0. 25 but < 0. 5, the tracer would move right. If the random number is ≥ 0. 5 but < 0. 75, the tracer would move up. Finally, if the random number is ≥ 0. 75, the tracer would move down. Based on this outcome, the tracer intends to hop to its nearest-neighbouring lattice site, referred to as the destination site. However, before accomplishing the hopping, the occupancy status of the destination site in regard of CRO or PRO is checked. If the destination site is occupied by a CRO, no hopping is performed and the tracer stays at its original lattice-site. In the case of PRO occupying the destination site, a uniform random number is again generated over the interval [0,1] to check for the partial reflection of the diffusing AMPAR with the Preflect = 0. 5. If the random number is ≥ Preflect, the tracer is allowed to diffuse to the destination site. Once the tracer reaches the PRO-occupied lattice site, it is considered to be bound. It will unbind and diffuse at a further time-step only when another uniform random number generated in a similar manner is greater than or equal to the probability of escape, Pescape, which is defined from the binding energy as [55], Pescape=e (−BindingEnergykBT) (2) Once the tracer unbinds from the PRO, it is allowed to diffuse in either of the directions isotropically. Therefore, it may be noted that rotational diffusion of DT is neglected in the present framework [31]. The above algorithm is identically shared by the standard condition as well as the conditions of self-crowding. However, the latter condition involves some additional restrains for hopping to the destination site. Since steric-exclusion of DTs among themselves is present in the case of self-crowding, the destination site already occupied by another tracer does not allow hopping of the subject tracer while diffusion. Under such a situation, a peculiar phenomenon of false self-blocking of tracers might appear during diffusion simulation. This condition and its implemented remedy are described in the following subsection. Further, as long as another DT is bound to PRO, the partial reflection of the PRO turns into a complete reflection and the destination site would behave as if it is occupied by a CRO. While performing the lattice-diffusion of a population of DTs with steric exclusion for each other, there appears a computational problem regarding the sequence of performing the finite-step hopping of individual tracers at each time-step of the diffusion simulation. At every time-step, all the tracers are genuinely expected to diffuse simultaneously on the lattice in random directions and depending on the availability of unoccupied neighbouring sites. If the tracers do not have steric-exclusion property, more than one receptor can occupy a single lattice site. Under this assumption, the computational sequence of performing the hopping of receptors one by one during a single time-step of simulation does not matter. However, if the tracers sterically repel each other, the computational sequence of performing the hopping of tracers at each time-step may lead to different profiles of diffusion. This issue becomes clearer when the lattice-diffusion of only two DTs, let’s say, DT1 and DT2 with steric repulsion is illustrated (Fig 1C). If the two tracers are sufficiently isolated from each other on the lattice, the sequence of performing the finite-step hopping of the individual tracers during a forward time-step of simulation does not matter, as either sequence, first DT1 and then DT2 or first DT2 and then DT1, leads to the same diffusion profile. Now consider that the two tracers are sitting at the neighbouring sites and the random number generation leads to the expected movements of DT1 towards DT2 and DT2 towards the upper unoccupied neighbouring site. The sequence where first DT1 is considered for hopping will lead to reflection of DT1 back to its position since the DT2 is presently occupying the lattice-site. Next, when DT2 is considered for hopping, it will easily move to the upper lattice site leaving behind an unoccupied lower lattice site. Here, only one receptor DT2 could practically diffuse. In another sequence where first DT2 is considered for hopping and then DT1, DT2 will move to the upper lattice site and DT1 will arrive at the earlier position of DT2. Here, both the tracers could diffuse. According to the theoretically-expected simultaneous diffusion of both the tracers, the diffusion profile deriving from the latter sequence is correct but the former sequence leads to an artefact owing to the computational sequence of performing the hopping of tracers. To solve this problem, an algorithm with two recursive steps of performing a sequential hopping of tracers at each time-step of simulation is devised: This scheme of hopping the tracers completely removes the possibility of false self-blocking of tracers while diffusion on the lattice. The time-duration of the recording of spatial locations of the DTs is 2s [31]. Wherever necessary, the observation has been made for an extended duration of time. Using the DTs’ trajectories, the temporal profile of ensemble-averaged MSD is computed as, 〈 r 2 〉 (t) = 1 N ∑ i = 1 N (x i (t) - x i (0) ) 2 + (y i (t) - y i (0) ) 2 (3) Where, N is the number of DTs on the two dimensional lattice. xi (t) and yi (t) are the x− and y− coordinates of the ith tracer at time t. xi (0) and yi (0) denotes the initial location of the ith tracer at the beginning of diffusion i. e. t = 0. The MSD profiles are further averaged over 250-700 ensembles of lattices for every crowding conditions. In theory, the MSD of two-dimensional diffusion is described in general as 〈 r 2 〉 (t) = 4 D t α (4) Here, α is the anomalous exponent and D is the diffusion constant of the diffusing particle. If α = 1, the diffusion is normal. However, if 0 < α < 1, it characterizes anomalous sub-diffusion. In this regard, computation of the log (〈r2〉 (t) /t) vs log (t) profile, referred in the following text as log-log profile, is very beneficial for procuring many important features of the tracer diffusion. log (〈 r 2 〉 (t) / t) = log (4 D) + (α - 1) log (t) (5) It may be noted that for normal diffusion with α = 1, the log-log profile would appear a flat horizontal line with slope zero. However, for anomalous diffusion, the log-log profile would have a negative slope of magnitude (1 − α). Higher will be the anomalousity of diffusion, sharper will be the decline in log-log profile. Therefore, the log-log profile can easily provide a clear demarcation for the diffusion to be called normal or anomalous and is useful for computing the anomalous exponent of the diffusion as well. As a matter of fact, one may also be interested in the spatiotemporal profile i. e. probability distribution function of tracer diffusion. However, the main interest of this paper is in properties directly pertinent to tracer mobility viz. anomalousity and effective diffusion coefficient. Since a Gaussian or non-Gaussian diffusion can be normal as well as anomalous [66], the mobility factors are ultimately described by the MSD.
In the absence of self-crowding factor, the 〈r2〉 increases linearly with time for the lower aCRO = 0. 00 − 0. 25 (Fig 2A). On the other hand, the conditions of very high aCRO = 0. 45 − 0. 60 can be clearly recognized by the confined tracer diffusion as the associated 〈r2〉 rapidly reaches a plateau (Fig 2A, inset) where no further variation in it occurs with progression of time. This range of aCRO is close to or above the percolation threshold, θP, of a square lattice framework for tracer diffusion, which is known to be approximately 0. 5 for a sufficiently large square lattice [54]. θP of a diffusion lattice signifies the area fraction of the lattice occupied by immobile completely-reflecting obstacles at and beyond which the possibility of an infinite percolation cluster to exist vanishes. In other words, there is no way left for a diffusing tracer to diffuse/percolate to extremely large distances over the lattice as the time progresses and, rather, gets trapped in small domains or confinements. Therefore, the trapping of tracers observed here at this range of aCRO is technically consistent with the concept of θP of a square lattice. However, for the intermediate range of aCRO = 0. 30 − 0. 40, the 〈r2〉 initially increases in a nonlinear manner but later adopts a linear profile (Fig 2A). The effect of varying aCRO on the 〈r2〉 of tracer diffusion becomes more conspicuous by looking at the log-log profiles (Fig 3A). The log-log plots depict almost flat horizontal profile for the lower aCRO and indicates perfectly normal diffusion according to Eq 5. However, increase in aCRO is marked by a brief initial anomalous diffusion of tracers where the log-log profile bears a negative slope (see Eq 5) and a gradual transition to the later normal diffusion. Remarkably, the crossover length, i. e. the 〈r2〉 traversed by the tracer after which the anomalous diffusion turns into normal diffusion, and the associated crossover time of the transition are observed to increase with rise in aCRO (Fig 3A). Only at aCRO closer to or higher than θP, a long-term anomalous diffusion appears where the log-log profile steeply decreases in a linear fashion at longer time. This, in turn, depicts a power-law time-dependence of 〈r2〉 (see Eq 4). Here, the crossover length and time approach infinity. Using the log-log profiles, the anomalousity of tracer diffusion for the different values of aCRO is computed in terms of the anomalous exponent, α, of 〈r2〉 using Eq 5. It can be easily noted that the flat horizontal log-log plots for lower aCRO have slope zero and, thus, α = 1. For very high aCRO characterized with long-term anomalous diffusion of tracers, the slope of the long-time tail of the log-log plots can be easily used to compute α, which turns out to be close to or equal to 0. However, for the intermediate values of aCRO observed with a transition from anomalous to normal diffusion, α is computed from the linear fitting to the initial segment of the log-log plot within crossover length associated with anomalous diffusion. Consequently, across the increasing aCRO, the anomalousity of tracer diffusion almost exhibits a sharp inverted sigmoidal profile (Fig 4A). There occurs a sudden decline in α (increase in anomalousity) close to aCRO = 0. 4, which has been suggested earlier [31] as the switch-like behaviour leading to the trapping of AMPARs. Further, using the log-log profiles, Deff of tracer diffusion is computed under different conditions of aCRO using Eq 5. The Deff shows a consistent decrease, unlike α, with increase in aCRO (Fig 4D). This indicates that although diffusion remains normal for lower aCRO, the receptor diffusivity indeed decreases with rise in the crowding conditions of completely reflecting obstacles. Under the extreme conditions of confined diffusion, the Deff becomes negligible. Altogether, it must be noted that these observations under the standard condition are identical to that reported earlier in a computational study by Santamaria et al. [31]. The temporal profiles of 〈r2〉 and the associated log-log plots for increasing aCRO under the additional consideration of the different self-crowding conditions of tracers are shown in Figs 2B–2G and 3B–3G, respectively. Particularly looking at the log-log profiles, it is clear that the self-crowding conditions with aDT = 0. 00001 and 0. 0001 (Fig 3B & 3C) exhibit almost an identical behaviour as well as identical to that noted in the no-self-crowding condition (Fig 3A). Even the anomalousity profile across increasing aCRO under these self-crowding conditions almost mimic that of the no-self-crowding condition (Fig 4A). Therefore, it appears that these self-crowding conditions are associated with sufficiently low tracer density such that they could not noticeably affect the features of tracer diffusion observed under no-self-crowding condition. However, aDT = 0. 001 demonstrates a significant intensity of long-term anomalous diffusion across the entire range of aCRO (Fig 3D). Although this behaviour does not become clearly visible across the 2s measurement time duration, longer time duration of 4s makes it clearly visible (Fig 5), where the log-log plots for the selected values of aCRO exhibit a long-time sharp decay profile. Accordingly, the profile of α across increasing aCRO is significantly affected and shifted to lower levels in comparison to that for the lower aDT and no-self-crowding condition (Fig 4A). Further, extremely self-crowded conditions with aDT ≥ 0. 005 lead to a very strong long-range anomalous diffusion across all values of aCRO (Fig 3E–3G) and the entire profile of α remains fairly close to zero (Fig 4A). It can be seen that the range of aCRO = 0. 0 − 0. 4 associated with normal receptor diffusion (α = 1) is almost invariant (Fig 4A) for the lower aDT = 0. 00001 and 0. 0001 and the no-self-crowding condition. Moreover, for these aDT and the no-self-crowding condition, the window of aCRO associated with the transition of tracer diffusion from normal to strongly anomalous nature is very narrow. This signifies a sudden rise in anomalousity and a switch-like behaviour for tracer trapping due to high reflecting obstacles’ density. However, for the higher aDT = 0. 001, the range of aCRO over which perfectly normal diffusion may occur completely vanishes (Fig 4A) and the window of transition from partially normal to strong anomalous diffusion is also very gradual. For aDT ≥ 0. 005, the diffusion remains strongly anomalous irrespective of aCRO (Fig 4A). To understand more about how the increase in self-crowding affects the tracer diffusion in the presence of CROs, the α is now plotted across the increasing values of aDT for a given value of aCRO (Fig 4B). It is interesting to note that intense self-crowding itself may bring strongly anomalous diffusion even in the absence of reflecting obstacles, as observed here for aDT ≥ 0. 005. This contrasts a common fundamental assumption in the earlier theoretical studies [31,64] that the AMPAR diffusion should be normal in the synaptic membrane in the absence of any non-binding completely-reflecting obstacles. Rather, the results suggest that it may also depend on the AMPAR density in the obstacle-free medium. At the same time, the present observations also support the above assumption to remain valid given the fact that the density of AMPARs in the extrasynaptic membrane is considerably low [4]. Another important thing to be noted is that the profile of α across increasing aDT exhibits a switch like behaviour when aCRO = 0, akin to that observed above in the case of variation in aCRO for lower aDT and no-self-crowding condition (Fig 4B). This switch-like behaviour appears to intensify with increase in aCRO as the transition becomes sharper. However, for aCRO ≥ 0. 45, the profile is fairly close to or is identically zero across all values of aDT and the switch-like behaviour completely disappears. Therefore, in regard of the anomalousity-driven trapping of AMPARs within PSD, self-crowding of AMPARs possibly appears as a new dimension to the causality, which was earlier thought to be driven only by the local macromolecular crowding other than the AMPARs. The cumulative effect of various densities of reflecting obstacles and diffusing tracers on the anomalousity of tracer diffusion observed here is summarized in the heat map shown in Fig 4C. It can be easily noted that, both high aCRO and/or high aDT can lead to strongly anomalous confined diffusion of the tracers. Further, the effect of self-crowding on tracer diffusion is distinguishable only at lower or moderate concentrations of reflecting obstacles and increase in aCRO catalyzes the anomalousity caused by higher aDT. However, for very high CRO concentration, tracer diffusion remains strongly anomalous for all self-crowding and no-self-crowding conditions, owing to the lack of percolation clusters on the 2D square lattice. To observe the effect of self-crowding on tracer diffusion in terms of the effective diffusion coefficient, Deff is computed from the log-log profiles under the varying conditions of aDT and aCRO. As a matter of fact, for diffusion marked with α equal to or sufficiently close to 1, the computation of Deff is very straightforward (see Eq 5). On the other hand, for receptor diffusion marked with α equal to or sufficiently close to zero, the Deff will be certainly negligible as there occurs no apparent diffusion at a substantial timescale. However, for intermediate values of α, the diffusion is neither perfectly normal nor completely confined and it becomes difficult to conceive a term like a constant diffusion coefficient to describe the 〈r2〉 over the entire duration of time. Under such conditions, the diffusion coefficient becomes time-dependent and is generally described through the two kinds of time-dependent quantities viz. apparent diffusion coefficient and the instantaneous diffusion coefficient [67]. The apparent diffusion coefficient, Dapp, is a time-averaged quantity and signifies the Deff of normal diffusion which could efficiently lead to the identical 〈r2〉 at a given time which one gets through the anomalous diffusion. This is given as, D a p p (t) = D t 1 - α (6) Here, D is the original constant present in the Eq 4. On the other hand, the instantaneous diffusion coefficient, Dinst, represents the instantaneous rate of change of slope of the nonlinear increase in 〈r2〉 at a given time, which is given as, D i n s t (t) = α D t 1 - α (7) As evident, both the quantities decrease with progression of time in anomalous diffusion [67]. For the case here, use of Dapp is more suitable as it provides a sense of effective diffusion coefficient which could be used to describe diffusion conditions characterized with the intermediate values of α between zero and one. However, the choice of Dapp would necessarily depend on the time duration for which the process is observed. In the earlier studies [24,29,31], distribution of diffusion coefficient is also shown and the statistical parameters such as median diffusion coefficient is computed. Yet, there also the distribution is strictly dependent on the time at which the observation is made and the statistical parameters do temporally evolve. Therefore, the Dapp is computed for the time point of 2s, which is the time duration of diffusion measurement performed in the present study, and will be considered here as the Deff of tracer diffusion characterized with intermediate values of α. For aDT = 0. 00001 and 0. 0001, the variation in Deff with increase in aCRO is completely overlapping with that for the no-self-crowding condition (Fig 4D) and, accordingly, the tracers mobility gradually decreases with increase in aCRO. However, the profile for aDT = 0. 001 is shifted to slightly lower values depicting reduced mobility due to increased self-crowding of the receptors. Indeed, in this case too, the mobility appears to decrease with increase in aCRO. For the rest very high values of aDT ≥ 0. 005, the entire profile of Deff is shifted to extraordinarily low levels (Fig 4D) depicting heavily hampered mobility of tracers owing to steric-exclusion and confinement among themselves as well as in the presence of CROs. Altogether, high density of completely reflecting obstacles and/or tracers engenders reduced mobility and confinement in terms of both the anomalousity as well as effective diffusion coefficient of the tracer diffusion. In this regard, the earlier experimental studies involving monitoring of the properties of a diffusing entity in the presence of same entity acting as the crowders also appears to strongly corroborate the above observations resulting from the self-crowding. A recent study by Roosen-Runge et al. [68] on the diffusion of bovine serum albumin in the aqueous solution using neutron backscattering has revealed that increase in the volume fraction occupied by the protein (even upto 30%) causes strong decline in the translational diffusion coefficient and leads to shorter-time self-diffusion, implying anomalous nature in action. Similarly, another experimental study by Ramadurai et al. [69] involving fluorescence correlation spectroscopy of the lateral diffusion of a variety of integral transmembrane proteins of different sizes, such as monomeric LacY to trimeric glutamate transporters, at their different density on artificially reconstituted large lipid vesicles demonstrates that increase in the size and density of the subject protein leads to strong decline in the later diffusion coefficient. Further, it has been shown that there occurs a significant decrease in the anomalous exponent of the diffusion at sufficiently high density of the proteins and is evitable even for monomeric proteins, such as LacS. A very recent study by Houser et al. [70] on the lateral diffusion of a homogeneous population of transferrin membrane proteins using fluorescence correlation spectroscopy has also shown that increase in the membrane coverage by the protein leads to strong decline in the diffusivity and has emphasized on the steric-exclusion underlying the self-crowding of the protein. It must be noted that transferrin occupy much lesser membrane area (∼ 24nm2) in comparison of our subject protein, AMPAR. Therefore, the self-crowding of bulky AMPARs implied here through the tracer diffusion indeed appears to be a significant factor at play in the anomalous diffusion and trapping of these receptors in the PSD, where these receptors are generally present at high density. The SI S1 Video demonstrates the temporal evolution of the position of a diffusing tracer under different self-crowding conditions, but in the absence of any other obstacles, as well as a control condition of free-diffusion. Three levels of uniform binding energies representing weak (2kBT), intermediate (6kBT) and strong (10kBT) binding of tracers to the binding obstacles (PROs) are separately considered. Given a binding energy, four arbitrary densities of PROs, aPRO = 0. 2,0. 4,0. 6 and 0. 8, over the lattice are sampled to broadly capture the different situations of the accumulation of scaffold proteins, ranging from sparse to very dense, underneath the PSD. Subsequently, these combinations are examined for the different conditions of self-crowding of tracers. In this part of the study, reflecting obstacles are completely absent and only the role of binding obstacles in shaping the nature of tracer diffusion is examined. The features of tracer diffusion under no-self-crowding condition is surely monotonous in the presence of binding obstacles. The tracer diffusion is always perfectly normal for all binding energies and values of aPRO, as the log-log profiles (Fig 6) remains fairly horizontal along the entire duration of diffusion monitoring and the α remains strictly close to one (Fig 7). However, for a given binding energy, the log-log profile shifts to lower values with increase in aPRO. Increase in binding energy further lowers the levels of these log-log profiles. This has implications in the decline of tracer mobility in terms of Deff. Accordingly, the Deff exponentially decreases with increase in aPRO for a given binding intensity (Fig 8). Moreover, increase in binding intensity shifts the Deff profile to lower orders of magnitude, depicting further decline in tracer mobility. These observations for no-self-crowding condition are equivalent to that observed in the computational study by Sanatamaria et al. [31]. Further, the absence of anomalousity in tracer diffusion in the presence of a wide range of PRO density and binding energy is also consistent with the previous study of anomalous diffusion in the presence of binding performed by Saxton [55], where it is implied that simple valley models of tracer binding always leads to normal diffusion under thermally-equilibrated initial condition. The self-crowding conditions with aDT = 0. 00001,0. 0001 and 0. 001, are found to exhibit behaviors identical to that under the no-self-crowding condition. For a given binding energy and aPRO, the log-log plots across these self-crowding conditions are strongly overlapping with that of the no-self-crowding condition (Fig 6). Accordingly, diffusion is normal across all the values of aPRO and the levels of binding energies with α close to one (Fig 7). Further, the Deff for these self-crowding conditions demonstrate a consistent decrease in the tracer mobility with increase in binding energy and PRO density (Fig 7). Also, the Deff profiles are sufficiently overlapping for these conditions of self-crowding as well as no-self-crowding. Therefore, it appears that the increase in tracer density to 0. 001 has no distinguishable effect on the tracer diffusion in the presence of binding obstacles. Rather the diffusion is being mainly governed by the PRO density and the binding energy. On the other hand, the self-crowding conditions with aDT ≥ 0. 005 exhibit a peculiar behaviour. For weak binding events, these self-crowding conditions clearly demonstrate a strong long-range anomalous diffusion for lower PRO density, aPRO = 0. 2, (Fig 6A) and the values of α are close to zero (Fig 7A). However, as the PRO density is increased, the anomalousity of diffusion gradually reduces and the log-log profiles tend to approach normal diffusion behaviour. For the case of aDT = 0. 005, diffusion becomes fairly normal at aPRO = 0. 8 (Fig 6D) and α reaches 1 (Fig 7A). Tracer diffusion for aDT = 0. 01 also tends to acquire normal behaviour with rising aPRO, though there remains slight anomalousity even at aPRO = 0. 8. However, for = 0. 1, the diffusion remains strongly anomalous even at = 0. 8 (Figs 6D & 7A). An important thing to observe is that the log-log profile of normal diffusion that the anomalous tracer diffusion for aDT = 0. 005 and 0. 01 gradually approaches (Fig 6D) with increase in aPRO appears to overlap with that obtained for aDT ≤ 0. 001 as well as for the no-self-crowding condition. Remarkably, aDT = 0. 005 and 0. 01 consistently exhibit normal diffusion across all PRO densities for the intermediate and strong binding energies, as their log-log plots (Fig 6E–6L) remain horizontal with α = 1 (Fig 7B & 7C). Moreover, these log-log plots fairly overlap with that of the lower self-crowding conditions under the respective conditions of binding energies and aPRO. However, for the intermediate binding intensity, aDT = 0. 1 exhibits significant anomalous diffusion for lower aPRO = 0. 2 (Figs 6E & 7B). The anomalousity soon vanishes for aPRO ≥ 0. 6 (Fig 6G & 6H) and α reaches 1 (Fig 7B). For strong binding intensity, aDT = 0. 1 exhibits perfectly normal diffusion for all values of aPRO and identical to the lower self-crowding conditions (Figs 6I–6L & 7C). At this point, if we remind the observations regarding tracer diffusion in obstacle-free medium, aDT ≥ 0. 001 demonstrated a marked long-range anomalous diffusion (Fig 3D–3G) with significantly low α (Fig 5A–5C). Together with the observations made here in the presence of binding obstacles, it is strongly evident that increase in binding phenomenon, either through increase in PRO density or/and increase in binding energy, reduces the anomalousity in tracer diffusion arising from higher self-crowding. However, it is also observed that the intensity of amelioration of the anomalousity with increase in binding further depends on the intensity of self-crowding. Very intense self-crowding conditions would require a considerably large increase in binding energy and scaffold density to exhibit perfectly normal diffusion. As shown here, for strong tracer-PRO binding, receptor diffusion is completely governed by binding obstacles’ density, regardless of the self-crowding conditions. Nonetheless, for weak binding intensity, tracers mobility in terms of Deff for aDT = 0. 005 initially increases with increase in aPRO and later decreases along the profiles obtained for lower aDT (Fig 8A). However, for aDT = 0. 01, Deff consistently increases with increase in aPRO. The increase in Deff under these conditions of aDT owes to the concomitant relaxation of anomalousity of tracer diffusion. For aDT = 0. 1, Deff remains significantly close to zero for all values of aPRO (Fig 8A) due to strongly anomalous tracer diffusion. For intermediate binding intensity, the profile of variation in Deff for aDT = 0. 005 and 0. 01 is identical to that of the lower self-crowding as well as no-self-crowding conditions (Fig 8B). However, for aDT = 0. 1, Deff sharply rises with increase in aPRO but soon gets along the decreasing profiles obtained for lower values of aDT. For the strong binding intensity, all conditions of aDT exhibit an identical profile of decrease in Deff with rise in aPRO (Fig 8C). Therefore, increase in binding indeed ameliorates anomalousity of tracer diffusion arising from the self-crowding of the tracers and stronger binding favors normal diffusion even under high tracer density. However, in regard of Deff, increase in binding consistently reduces the mobility for low tracer density. But for high tracer density, increase in binding leads to higher mobility in the situation where concomitant reduction in the diffusion-associated anomalousity is observed. Otherwise, given a sufficiently strong binding condition, any further increase in binding leads to consistent reduction in the tracers mobility.
In the case of obstacle-free diffusion of tracers, increase in the tracer density is responsible for the more frequent self-crowding collisions among the tracers and the resulting obstructions of their diffusion paths. Through the above observations, it is realized that, for the given specifications of the lattice dimension, a rise in tracer density (aDT) to the order of 0. 001 commences the appearance of anomalousity in tracer diffusion and further increase in the tracer density leads to its more noticeable magnitude. The appearance of anomalousity can easily be better portrayed under an assumed condition of extreme self-crowding when aDT is sufficiently close to 1 and almost all lattice sites are occupied with the tracers. At every time-step of simulation, the hopping of a tracer to any random direction would be denied because the neighboring destination sites in almost all directions are occupied with the tracers. This will repeatedly occur across sampling of the entire population of tracers and, as a consequence, the tracers would remain stuck at their positions along a unit advancement in time. This condition would remain unchanged for every further time-steps and the MSD would not increase with time, depicting extraordinarily strict confinement of the tracers. It can now be extrapolated for the lower tracer densities that the confinement would be certainly reduced but the abundance of restricted diffusion would accordingly lead to anomalousity. Need not to say that, for the no-self-crowding condition, diffusion of single tracer on the obstacle-free lattice would always remain perfectly normal. When reflecting obstacles (CROs) are added to the system, the unoccupied fraction of the lattice sites connected to each other through the diffusive edges decreases. In fact, this decrease in percolation paths is significant only when the CRO density reaches the percolation threshold of the square lattice. This is the reason that, for the conditions of single tracer diffusing in the lattice frame with no self-crowding at all or low tracer density with insignificant counts of self-obstructions during diffusion, what only shapes the nature of tracer diffusion is the extent to which CRO density is close to or beyond the percolation threshold. However, when tracer density is sufficient to effectuate a considerable amount of self-crowding against their free diffusion, slight decrease in percolation paths even at much lower CRO density can exacerbate the anomalousity of diffusion arising from the self-crowding. Furthering this description to the conditions of extreme self-crowding at very high tracer density, a situation appears where even in the absence of reflecting obstacles, the anomalousity of tracer diffusion is close to its possible maximum level and adding reflecting obstacles does not manifest into any significant change. Unlike the above cases, stating the exact mechanism involved in the observed effects of increase in binding on the self-crowded diffusion of the tracers is not so straightforward. Therefore, the attempt here would be to carefully and systematically deduce the plausible mechanism, while keeping in mind the specific arrangements utilized in the above simulation experiments and the features associated with them in the background. If aPRO is the fraction of lattice sites occupied by binding obstacles (PROs), (1 − aPRO) is the fraction unoccupied by them. It essentially results into a partition of the lattice medium into two spatial subsets viz. non-binding and binding spatial subsets. The latter is capable of binding a diffusing tracer and freezing it at its location for a random size of waiting time. Notably, the mean waiting time is directly proportional to the intensity of binding such that higher is the binding energy, longer is the mean waiting time. Another important thing to note is the size of binding subset relative to the non-binding subset. Higher is the aPRO, larger is the binding subset. At the beginning of the simulation, when tracers are uniformly distributed over the lattice with the desired area fraction aDT, aDT aPRO would be the fraction of aDT lying on the PROs and, thus, lying in the binding spatial subset whereas aDT (1 − aPRO) will be lying on completely empty nascent lattice sites and, thus, belongs to the non-binding spatial subset. As the lateral diffusion proceeds, there occurs diffusion of tracers within their own spatial subsets as well as diffusion-associated exchange of tracers across the subsets. Due to reduced tracer mobility in the binding subset, there would occur an initial drift of a certain fraction of the tracer population belonging to the non-binding subset towards the binding subset acting as a sink, until a thermal equilibrium is achieved. Higher is the binding energy and larger is the binding subset, the thermal equilibrium would be acquired with a larger fraction drifted. Once such an equilibrium distribution of the tracer population between the two spatial subsets is achieved, contribution of each population to the anomalousity of entire tracer diffusion can be easily compartmentalized and examined. The process of equilibrium distribution engenders two consequences for the tracers belonging to the non-binding spatial subset. First, the resultant density of tracers within the non-binding subset is significantly reduced leading to a reduced self-crowding condition. Second, the binding subset-associated larger population of tracers appear as almost static reflecting obstacles (CROs) to the highly mobile tracers belonging to the non-binding subset. Here comes the role of longer mean waiting time under stronger binding condition which leads to larger decline in the hopping rate of the tracers belonging to the binding spatial subset. Therefore, under strong binding conditions, the entire diffusion system for the tracers associated with non-binding spatial subset turns into the diffusion of tracers at low density but in the presence of less or moderately dense CROs. And, according to the previous experiences with the reflecting obstacles, the contribution to anomalousity from the diffusion of unbound tracers is severely reduced. One can now envisage that decrease in binding will certainly violate this setup by bringing more self-crowding encounters amongst the tracers and their resulting anomalousity would be higher. On the other hand, the diffusion of tracer population belonging to the binding subset within its own spatial subset appears, according to the results, less anomalous under stronger binding conditions. It seems that declined rate of hopping is beneficial in reducing anomalousity by frequently avoiding self-crowding encounters. This is even helpful for the case of encounters with highly mobile tracers belonging to the non-binding spatial subset, which are themselves in lesser density too. However, binding certainly reduces the mobility of the tracers belonging to binding spatial subset. Corollary, lesser and weaker binding would increase the tracers mobility but would concomitantly cause more frequent self-crowding encounters within the binding spatial subset as well as across the non-binding subset and lead to higher anomalousity. This entire description of the possible mechanism concludes at one interesting fact that self-crowding collisions are the main source of anomalousity. Although binding reduces the effective mobility of the tracers, it ameliorates anomalousity by avoiding such collisions. Therefore, the phenomenon of binding plays its role at a trade-off point between the effective mobility of the tracers and the anomalousity of their diffusion. In the present study, the diffusing tracers are point particles diffusing on the lattice framework. Therefore, one may argue over how the self-crowded diffusion of point tracers may capture the crowded diffusion of AMPARs, which are bulky transmembrane structures with non-zero lateral span. The reply to this question is hidden in the description of area-fraction of the lattice occupied by the point tracers and the use of ensemble-averaged MSD. The extracellular domain of an AMPAR is the most bulky structure with lateral dimensions of length 16nm and width 8nm [35,42]. Therefore, its two-dimensional projection on the lipid membrane would occupy a surface area of roughly 128nm2. For the purpose of realizing side-ways collisions during lateral diffusion, the complex details of an AMPAR structure can be essentially reduced to a transmembrane cylindrical structure [29] of radial width 6nm. This lateral radial span characterizes the exclusion area (128nm2) which avoids approach of another receptor closer than this radius and presumably reflects it away in an elastic manner. Certainly, association of the receptor with other auxiliary proteins [49–51] would further stretch the exclusion area, as it becomes more bulky along the lateral dimension. Given the density of the AMPARs and the areal span of the PSD or a subregion within the PSD, one may easily procure the resultant fraction of the PSD area occupied by the total exclusion area of the receptor population. This fraction amounts to the area-fraction of the self-obstructing point tracers, aDT, on the lattice referred here. Nevertheless, this approach gets complete only when ensemble-averaged MSD of the tracers is used to capture the bulk diffusion properties of AMPARs. Had it been time-averaged MSD observation of single tracers, the statistical approximation using aDT would not suffice to fully reproduce the time-averaged MSD behaviour of the non-zero size AMPARs [71]. For instance, the density of AMPARs in the extrasynaptic membrane has been experimentally measured to be 3–5μm−2 [4]. The estimated length scale of the region of extrasynaptic membrane on the spine head is approximately 1μm [72] and an effective surface area close to 1μm2. Therefore, the fraction of extrasynaptic region occupied by the total exclusion area of the receptor population would be 0. 00038 − 0. 00064. Given this fraction as aDT in the present study, it is shown that the tracer diffusion would be perfectly normal in the absence of any obstacle. The same is observed in the particle tracking experiments [24–29] and the receptor diffusion is normal in the extrasynaptic membrane, which contains least transmembrane obstacles [27]. However, within the PSD of typical radial size 100nm [72,73], the AMPAR count may range from 20–100 [74] which is equivalent to receptor density 650–3000μm−2 [74,75]. These estimates lead to an area fraction of 0. 08 − 0. 4 of the PSD to be occupied by the total exclusion area of the receptor population. Given this fraction as aDT, it is shown here that self-crowding of the tracers would immensely contribute to the anomalousity of tracer diffusion and their confinements. Nonetheless, the convergence of lattice model of diffusion under the described conditions of reflecting and binding obstacles to continuous-space diffusion becomes important to be investigated. In regard of the earlier studies on AMPAR diffusion in the absence of self-crowding interactions amongst the receptors, a very recent study by Li et al. [29] has used the approach of continuous-space diffusion on the basis of Monte Carlo simulation of the Langevin dynamics. Using photoactivated localization microscopy (PALM) technique, the spatial distribution of the submembranous PSD-95 binding obstacles was determined and the other reflecting transmembrane obstacles in the simulation space were distributed accordingly. The findings of their study strongly asserts the observations made in an earlier lattice model-based work by Sanatamaria et al. [31] from the same research group. The present study additionally raises a significant factor of self-crowding in shaping the receptors diffusion. Hence, what appears important is to show here how convergent is the lattice-diffusion approach and the proposed recursive algorithm to the self-crowded continuous-space diffusion. Accordingly, the Monte-Carlo simulation of the Langevin dynamics of receptor diffusion is performed with steric-exclusion under the vibrant conditions of self-crowding. In this approach, the AMPARs are modelled as flat circular disks of exclusion radius 6nm. The centre of the disc is moved in random directions over a Δt time-step of simulation as Δ x = 2 D Δ t ξ (t) and Δ y = 2 D Δ t ξ (t). Here, ξ (t) represents white Gaussian noise with mean zero and variance 1. The 2 D Δ t defines the standard deviation of the random-sized steps taken independently in x− and y− directions over single time-steps of the simulation. D is the free diffusion coefficient of AMPARs in the postsynaptic membrane. While diffusion, it is taken care that the centres of any two discs should not be at a relative distance shorter than the double of the radius of the discs to implement steric-exclusion. This minimum distance between the centres of two discs represent collision between the incompressible hard discs. Further, the collisions are considered elastic. The number of these circular disks in a square simulation space of area 1μm2 is computed from the desired area fraction aDT under investigation. The simulations begin with all the disks uniformly distributed across the simulation space with non-overlapping steric conditions. Three kinds of self-crowded conditions with sparse (aDT = 0. 0001), fairly dense (aDT = 0. 01 and 0. 1) and extremely dense (aDT = 0. 6) presence of AMPARs are enquired. These three kinds of self-crowded conditions are chosen in accordance with the order of AMPAR density typically observed in the extra-synaptic membranes and the PSD, as mentioned above. The presence of any other obstacles is not considered. From the continuous-diffusion scheme, the log-log plots of the MSD are obtained by averaging over a sufficiently large size (700) of ensemble of the independent simulations. Based on this, the convergence of the newly-proposed lattice-based recursive algorithm for self-crowded diffusion to the continuous-space diffusion is checked and its efficacy over the other possibility without involving the recursion or repeated check of the labelled “DT-blocked tracers” in the same algorithm is also evaluated. Fig 9 demonstrates the overlaid normalized MSD plots obtained from the continuous-diffusion simulation, recursive lattice-based algorithm and the same algorithm without recursion under the different conditions of self-crowding. Interestingly, it is consistently observed that lattice-diffusion scheme with recursive algorithm is quantitatively sufficiently close to the MSD-profiles of the continuous-space diffusion in comparison to that without involving the recursive algorithm. For very low (aDT = 0. 0001) as well as very high (aDT = 0. 6) self-crowding density, it is quite apparent that the recursive algorithm and its absence are providing fairly identical convergence to the continuous-space model. Such convergence of the two lattice-diffusion schemes under very low density of tracers may be due to the lack of a substantial frequency of self-obstructing events. However, the same observed under very dense crowding of the tracers may arise from the fact that the fraction of false self-blocking events is negligible amidst the very frequent steric-collisions among the receptors, as most of the tracers are unable to diffuse under such strongly crowded condition. It is only for the intermediate densities (aDT = 0. 01 and 0. 1) of the tracers that the distinction between the MSD profiles obtained from the schemes with and without the recursive algorithm is more conspicuous. It describes that majority of the obstructions observed on first-attempt through the computational sequence of hopping appears to be false and, as a result, the observed log-log plot appears steeper (i. e. more anomalous and obstructed) than that obtained from continuous-space schemes and recursive algorithm. However, when such obstructions are checked back in a recursive manner, it leads to their diffusion. Here, the behaviour of receptor diffusion in the absence of recursive algorithm is substantially deviated from that obtained from the continuous-space diffusion. More specifically, the deviation is higher for aDT = 0. 01 in comparison to aDT = 0. 1. Seeing these remarkable consequences, an important question appears: why has the factor of self-crowding of the bulky AMPARs remain unappreciated till now when we already have a sufficiently large body of experimental data on the nature of AMPAR diffusion at excitatory synapses? Possibly, the reason to this ignorance does not entirely or essentially owe to the experimental studies. In fact, the in vivo sophisticated microscopic tracking of endogeneously-expressed AMPARs or less bulky genetically-engineered transmembrane probes at excitatory synapses and the resulting observations regarding their anomalous diffusion within PSD indeed involve all the several factors which are simultaneously present there under the real physiological conditions of the experiments [24–29]. However, the mechanistic deduction of the effects of these pertinent factors to the finer details is beyond the scope of any existing experimental techniques. At this point, the theoretical studies [29–33] using detailed models of the receptor diffusion in the presence of PSD crowd comes forth as the only but efficient option to dig deeper into the mechanisms. Certainly, these studies have so far led us to realize the impacts of obstruction and binding by the local crowd of transmembrane and submembrane scaffold proteins on receptor diffusion in the PSD. On the basis of these factors, the previous experimental data on the tracking of receptor diffusion has also been explained to a great extent. Therefore, the existing ignorance towards the possibility of an additional role of the self-crowding of receptors owes merely to the lack of consideration of the self-crowding in the earlier theoretical approaches. Yet, the possible contributions of self-crowding could be unknowingly by-passed in the earlier theoretical approaches by appropriate parameter estimation within the framework of the previous models and it is a strong possibility that self-crowding remained a hidden variable in the process. For instance, even in the detailed study by Li et al. [29], simulations used to describe the monitored diffusion of genetically-engineered single- or double-pass transmembrane probes using FRAP as well as sptPALM techniques considered a substantial fraction of AMPARs endogenously-expressed in the cultured hippocampal neurons as the part of static crowd only. It must be noted that transmembrane crowds like AMPARs are sufficiently mobile and may impact differently from the other relatively static crowd on the probe diffusion. Nonetheless, this immediately draws attention to the fact that the experimental data on the receptor diffusion should also contain the elements of self-crowding, besides the earlier recognized factors. To throw light on this aspect, some of the previous MSD data on the AMPAR diffusion at excitatory synapses are examined on the basis of the observations acquired in the present study and is discussed in the following subsection. As stated above, the commonly observed density of AMPARs in a typically-sized PSD would lead to an occupied area fraction ranging 0. 08 − 0. 4. The present observations suggest that receptor diffusion would be strongly anomalous at this level of area fraction due to self-crowding, regardless of the other transmembrane obstacles. This leads to a confusing situation where importance of transmembrane obstacles becomes obsolete, whereas the earlier studies have shown that the steric repulsion by the obstacles is an indispensable and critically essential factor behind AMPAR trapping and accumulation within the PSD. This contradiction arises because of the difference between the configuration of the diffusion system employed here and that of the system under natural condition. The diffusion system used here has a periodic boundary condition at its edges, leading to a homogenous condition of obstruction or binding applied on a diffusing entity throughout an infinite two-dimensional space. On the other hand, receptors diffusing within the PSD at excitatory synapses can easily escape the local crowded condition by entering into the extrasynaptic space and, hence, the natural diffusion system is an open system. Therefore, the configuration of the present system mainly captures the diffusive behaviour of a receptor as long as it is diffusing within the PSD region and the density of the receptors is in a perfect or quasi- steady state. This leads to the speculation that self-crowding of AMPARs cannot itself hold the accumulated density of the receptors if the steric repulsion by the other obstacles are completely removed. Rather, steric obstructions by the relatively static density of other transmembrane proteins may provide the initial as well as maintaining driving force by reducing the mobility of the receptors within the PSD and self-crowding may later come into action as the density rises to a certain required level. In fact, this might be possible as increase in reflecting obstacle density leads to consistent decrease in the Deff but a sudden increase in anomalousity occurs only beyond a certain very high CRO density. This speculation would have a remarkable impact on the required concentration of transmembrane obstacles predicted theoretically to effectuate anomalous confined diffusion of the receptors within PSD. Fitting to the data on the diffusion of AMPARs in synaptic and extrasynaptic spaces acquired in the experimental study by Li and Blanpied [76] using single particle tracking and localization microscopy provides α = 0. 22 and 0. 99, respectively (Fig 10A). In the single-particle tracking experiment by Renner et al. [24] using quantum dots, two kinds of trajectories of the AMPARs diffusing in the PSD region were observed (Fig 10B). AMPARs, referred to as trapped, retained for longer durations within the PSD and exhibited strongly anomalous subdiffusion. The other population of AMPARs, referred to as passing, stayed for relatively shorter duration within the PSD but exhibited only a slightly lesser anomalous diffusion in comparison to the trapped receptors. Fitting to the MSD data of trapped and passing receptors provided α = 0. 48 and 0. 5, respectively. In a similar manner, through the fitting to the data on receptor diffusion within synaptic region obtained in the study by Renner et al. [77], the α comes out to be 0. 42 (Fig 10C). If the self-crowding factor is not considered, such high anomalousity of receptor diffusion could be possible only at an obstacle density (aCRO) between 0. 4 and 0. 44. Even, the theoretical study by Santamaria et al. [31] predicts a similar range of obstacle concentration (0. 4 − 0. 46) for achieving such low anomalous exponent of AMPAR diffusion. However, introduction of self-crowding can bring similar high levels of anomalousity even at lower levels of obstacle concentration (see Fig 4). Imagining that expression of transmembrane obstacles at a density lesser than the theoretically-predicted value would cause sharp loss in accumulated AMPAR density seems very strict and unrealistic for the natural scenario. In fact, self-crowding may provide a certain degree of flexibility to this aspect of synaptic homeostasis. This feature can be tested through an experiment where the nature of AMPAR diffusion and receptor accumulation within the PSD is examined under different densities of transmembrane obstacles. If a significantly anomalous receptor diffusion is observed at an obstacle density ammounting to occupied area fraction lesser than the abovementioned, it would be a strong evidence for the speculation drawn here for the self-crowding of the AMPARs. Further, the average density of PSD-95 scaffold proteins in the PSD of an excitatory synapse is known to be 3000μm−2 [78]. The radial size of a PSD-95 protein is estimated to be almost 2. 5nm [35], which results into a lateral span of 19. 64 × 10−6μm2. Assuming a homogenous distribution of the PSD-95, the area fraction of the PSD occupied submembranously by the total PSD-95 proteins would amount to 0. 059. Knowing that AMPARs are present at very high density within the PSD [74,75], the present study suggests that binding in the presence of such low area fraction of PSD-95 would cause an insignificant effect on the anomalousity of receptor diffusion, unless the AMPAR-PSD-95 binding affinity is extremely high throughout the binding sites. Interestingly, this has also been noted in the earlier experimental study by Li and Blanpeid [76], stating that whole-synapse PSD-95 density would have inconsiderable impact on the diffusion of transmembrane proteins. However, the experimental estimation of PSD-95 distribution demonstrates that, rather than homogeneously distributed, these proteins are enriched in smaller subregions or nanodomains within the PSD [52]. Accordingly, their local density and occupied area-fractions within these nanodomains may acquire considerably large magnitudes, such that even moderate binding affinity may appear effective in reducing the anomalousity of receptor diffusion at high receptor density within the nanodomains. Therefore, such PSD-95 distribution appears as a physiological strategy to enhance the effectiveness of binding on the AMPAR diffusion and exchange with the perisynaptic space. Nonetheless, AMPARs have also been observed to accumulate at higher density within the nanodomains [52,79]. The present observations suggest that, given PSD-95 density and AMPAR-binding affinity, the extent to which the anomalousity arising from the self-crowding would decrease further depends on the local receptor density within these domains. As the area fraction occupied by the higher density of AMPARs within the much smaller (∼ 80nm, [79]) PSD-95-rich domains would be very large, receptor diffusion within scaffold-rich domains would retain significant anomalousity, despite the ameliorating influence of binding on the anomalousity. In fact, experimental observations using super-resolution microscopy in the study by Hosy et al. [79] on the nature of AMPAR diffusion within these nanodomains and outside (Fig 10D) indicate that the AMPAR diffusion within these PSD-95-rich nanodomains indeed exhibit strongly anomalous diffusion. Fitting to their data on receptor diffusion within the nanodomains and in peripheral region provides α = 0. 25 and 0. 85, respectively. It must be noted that nanodomains do contain transmembrane obstacles [5], for instance the adhesion proteins LRRTM2 [80]. However, the AMPARs are present at an extraordinarily high density in these nanodomains and self-crowding appears to be a prominent factor underlying the intense anomalousity in receptor diffusion observed earlier [79]. In a recent study by Li et al. [29] using genetically-engineered single-pass transmembrane probes, it has been observed that rapalog-mediated cross-linked binding probes with intracellular PDZ-binding segments exhibit more anomalous and confined diffusing than the single binding probes (Fig 10E). Fitting to the data provides α = 0. 34,0. 28 and 0. 17 for the single probes, binding-non-binding cross-linked probes and binding-binding cross-linked probes, respectively. In a straightforward manner, it appears that enhanced binding with multiple PDZ-domains enhances anomalousity of receptor diffusion. However, it can be also be possible that increase in binding lowers the receptor mobility and causes accumulation of larger number of AMPARs in the local area. This may lead to increased self-crowding in a feedback manner and fuels stronger anomalousity to the receptor diffusion trapping more receptors. Nonetheless, the major practical limitation towards drawing a definitive conclusion is that the earlier experimental studies haven’t taken into account the density of AMPARs at the PSD while tracking their diffusive properties and, in the absence of such mentions, it is difficult to establish the connection between the exact contribution of self-crowding of AMPARs to their overall nature of diffusion within the PSD. Besides this, the present study is also in its preliminary state and contains some important limitations when it comes to closely imitate the true biological condition. Given a population of AMPARs, the receptors may be in different states which may affect the strength of binding to the scaffold proteins, such as association with different kinds of auxiliary transmembrane proteins [39,45] or no association at all [56,57], glutamate-bound desensitized state of the receptor [81], differences in the cytoplasmic domains of receptor subtypes [56,57] etc. The phosphorylation state of the auxiliary transmembrane protein, such as Stargazin [82] or that of the scaffold proteins, such as PSD-95 [83], also significantly affects binding. Together, these factors may lead to heterogeneous distribution of binding strength across the PSD. Further, the spatial distribution of the various scaffold proteins is inhomogeneous and smaller sub-regions within the PSD are found to have higher density of these proteins [52,53]. In a similar manner, various transmembrane proteins such as N-cadherin, have specific spatial distributions [29,35] rather than a perfectly homogeneous distribution in the PSD assumed here. Further, the presence of extracellular matrix in the synaptic cleft has also been observed to affect the lateral diffusion and accumulation of AMPARs in the PSD [84]. However, despite these limitations at the level of finer details, the present investigation indeed serves as an initial step towards gaining insight into the aspect of self-crowding of AMPARs, and similar other mobile bulky transmembrane proteins, and its effect on lateral diffusion in the postsynaptic membrane as well as in the specialized crowded PSD region. These features may serve as the conceptual nut-bolts for understanding the behaviour of more detailed models capturing the true irregular topology of synaptic PSD and the natural spatial distributions of the crowding and binding elements. Further investigations may involve the effects of self-crowding on the dynamics of AMPAR trapping and accumulation under the conditions of house-keeping maintenance of synaptic strength and evoked synaptic plasticity. | The transmembrane AMPA receptors (AMPARs) prominently exhibit lateral diffusion in the postsynaptic membrane at excitatory synapses. Steric obstructions to AMPAR diffusion due to the crowd of other relatively static transmembrane proteins and binding of AMPARs to the submembranous scaffold proteins in the specialized region of postsynaptic density (PSD) are well known to retard receptor diffusion, which causes receptor trapping and accumulation within PSD. However, AMPARs are significantly bulky structures and may also obstruct their own diffusion paths in the presence of their high density. It is shown here that intense self-crowding of AMPARs may lead to highly obstructed and confined receptor diffusion even in the obstacle-free medium, and the presence of other obstacles further aggravates this effect. AMPAR-scaffold binding reduces confined diffusion arising from self-crowding and strong binding engenders normal diffusion even at high receptor density. However, it overall causes reduction in the effective diffusion coefficient of the receptor diffusion. | Abstract
Introduction
Methods
Results
Discussion | medicine and health sciences
classical mechanics
nervous system
electrophysiology
neuroscience
simulation and modeling
membrane proteins
mathematics
statistics (mathematics)
reflection
cellular structures and organelles
research and analysis methods
proteins
mathematical and statistical techniques
statistical methods
transmembrane receptors
chemistry
monte carlo method
cell membranes
physics
biochemistry
signal transduction
mass diffusivity
cell biology
anatomy
synapses
integral membrane proteins
physiology
biology and life sciences
physical sciences
chemical physics
neurophysiology | 2018 | Self-crowding of AMPA receptors in the excitatory postsynaptic density can effectuate anomalous receptor sub-diffusion | 18,800 | 247 |
Quantifying human mobility has significant consequences for studying physical activity, exposure to pathogens, and generating more realistic infectious disease models. Location-aware technologies such as Global Positioning System (GPS) -enabled devices are used increasingly as a gold standard for mobility research. The main goal of this observational study was to compare and contrast the information obtained through GPS and semi-structured interviews (SSI) to assess issues affecting data quality and, ultimately, our ability to measure fine-scale human mobility. A total of 160 individuals, ages 7 to 74, from Iquitos, Peru, were tracked using GPS data-loggers for 14 days and later interviewed using the SSI about places they visited while tracked. A total of 2,047 and 886 places were reported in the SSI and identified by GPS, respectively. Differences in the concordance between methods occurred by location type, distance threshold (within a given radius to be considered a match) selected, GPS data collection frequency (i. e. , 30,90 or 150 seconds) and number of GPS points near the SSI place considered to define a match. Both methods had perfect concordance identifying each participant' s house, followed by 80–100% concordance for identifying schools and lodgings, and 50–80% concordance for residences and commercial and religious locations. As the distance threshold selected increased, the concordance between SSI and raw GPS data increased (beyond 20 meters most locations reached their maximum concordance). Processing raw GPS data using a signal-clustering algorithm decreased overall concordance to 14. 3%. The most common causes of discordance as described by a sub-sample (n = 101) with whom we followed-up were GPS units being accidentally off (30%), forgetting or purposely not taking the units when leaving home (24. 8%), possible barriers to the signal (4. 7%) and leaving units home to recharge (4. 6%). We provide a quantitative assessment of the strengths and weaknesses of both methods for capturing fine-scale human mobility.
Knowledge of daily and routine individual human mobility patterns within urban settings are important for urban planning [1]–[3], developing transportation models [3], promoting healthy lifestyles [4], and understanding infectious disease dynamics [5]–[13]. Measuring mobility at fine spatial and temporal scales through classic data collection methods (e. g. , interviews, diaries, direct observations) presents significant challenges, such as marked heterogeneities in the ability of individuals to recall the locations they visit, changes in people' s lives that affect their daily mobility (e. g. , new partners, change of jobs, school vacation) as well as privacy issues [11], [14]. These challenges can be exacerbated in resource-poor settings [6], [7], [10], [15], [16], such as our study site in Iquitos, Peru, due to the lack of complete and updated address maps (affecting geo-coding of self-reported addresses) and limitations in spatial literacy of interviewed individuals [11]. There is an urgent need to develop and validate easily deployable and culturally-sensitive tools that characterize a person' s routine mobility in order to link such information to health outcomes [6], [10], [13], [17], [18]. This is of particular relevance for understanding infectious disease dynamics, given the dominant role mobility has in driving infectious contacts and thus pathogen transmission, emergence, persistence and propagation [5], [6], [8]–[13], [18]–[22]. The wide availability of emerging location-aware technologies such as Global Positioning System (GPS) -phones or data-loggers provides new opportunities to quantify human mobility at fine spatial and temporal scales. Their use in research projects is feasible: they have decreased in cost and size, the technology has improved (i. e. , GPS chipsets are more efficient in acquiring and fixing a signal as well as in power consumption) and the units are widely accepted by study populations [6], [10], . Over the past ten years, GPS tracking (often coupled with other sensors) has taken a prominent role in physical activity and exposure research [17], [26], [27]. Their implementation, however, in infectious disease research has been limited in part due to the challenges in linking the positional data generated by such sensors with temporally and spatially discrete locations (i. e. , a person' s home) where pathogen exposure occurred, and more importantly, the complexities associated with the analysis of the vast amount of data that these sensors can generate. A recent systematic review [17] shows that most studies using GPS to track physical activity involve few participants (<20), track individuals over short time periods (<12 days) and are focused on specific age groups (children vs. adults) or environmental correlates of activity (e. g. , park vs. school movement) [17], [27]. GPS-based tracking presents enormous opportunities for improving our understanding of individual space-time activities and how they influence health outcomes, which has been done in various studies [6], [7], [10], [11], [15], [16], [28]. GPS technology, however, also has limitations that need to be addressed before considering it a “gold standard” for mobility research [26]. Rates of GPS data loss can reach 92% due to signal drop-outs, dead batteries, participants not wearing the units, signal loss during the initialization period or misuse of the device [17]. In Stothard et al.' s study in Uganda, the authors found that the track logs of the small, wearable GPS units (i-gotU) were accurate compared to a more sophisticated and costly unit (Garmin Oregon 550t) – discordance of <7 m for the 15 households tested – but there was GPS malfunction in units that was believed to be related to “insufficiently robust hardware for field conditions” possibly due to humidity or quality of the software [10]. As part of a larger study investigating risk for dengue (a human disease caused by a mosquito transmitted virus) in Iquitos, Peru, we simultaneously implemented two methods to capture fine-scale human mobility patterns: GPS data-loggers and semi-structured interviews (SSI). Dengue is a mosquito-transmitted viral disease of humans in tropical and subtropical regions of the world that is a rapidly growing public health problem [29], [30]. The main goal of this observational study was to compare and contrast the information obtained through these two methods to assess the issues affecting data quality, and identify strengths and weaknesses of each approach. We used two methods to analyze GPS data, and compared GPS results obtained via both methods with the results from the SSI.
Our study took place in Iquitos, a large and geographically isolated city in the Amazon Basin of northeastern Peru that is accessible only by boat or plane [31], between September 2008 and August 2010. The city of Iquitos has a high population density (∼390,000 inhabitants), and a very informal and dynamic economic structure (33. 4% of those economically active are either unemployed or informally employed) [31]. As observed in other resource-poor cities, Iquitos lacks a unified and updated address system. Car access and public transportation are limited and residents rely on personal motorcycles, ∼20,000 motorized rickshaws [“moto-taxis”], and a few bus lines to move throughout the city. The major industries in the area are small commercial enterprises, fishing, oil, lumber, tourism, and agriculture [7]. Iquitos is the home-base of an extensive, ongoing, long-term project since 1999 led by the University of California at Davis/U. S. Naval Medical Research Unit 6-Iquitos group [5]–[7], [15], [16], [32] studying the environmental, entomologic, epidemiologic and behavioral determinants of dengue virus transmission. Two methods for obtaining fine-scale human mobility data were simultaneously implemented: (1) GPS data-loggers (“i-gotU GT120”, Mobile Action Technology Inc.) and (2) semi-structured interviews (SSIs). Descriptions of GPS features, spatial accuracy, acceptance by participants and device deployment associated to this study were reported previously [11], [12]. The main attributes of selected units were: (1) data storage capacity and battery life capable of recording at least 3 days of data; (2) high spatial accuracy (∼4–10 m); (3) durable, water resistant and tamper-proof; (4) light weight (<50 g); (5) carrying mechanism (lanyard around neck) widely accepted by participants of different ages/sex; (6) little to no maintenance required by study participants; (7) low cost ($49); and (8) password protection and a special socket for data download (to protect participant' s confidentiality). The units are easily worn on a neck strap or in a pocket, and have been used to track routine movement patterns of Iquitos residents over the past three years with a high level of acceptance (98%) [16]. Based on the known limitations of classic interview instruments to capture overt behaviors in space and time [2], [33]–[35], and guided by findings from focus group discussions performed in Iquitos [16], we designed a SSI for capturing positional and temporal information of routine human mobility. Key findings from the focus groups that guided the survey development included [16]: (1) people could clearly identify many of the routine locations they visited, although they sometimes needed certain “triggers” for recall (and these were identified), (2) there were marked differences in reported mobility routines by gender and age groups; and (3) there were clear “common activity spaces” (markets, recreational spots, etc). The developed SSI contained one section listing commonly visited locations, such as markets, health facilities, and schools, and a section that used field-tested triggers to help people recall “individual” locations visited (such as relatives' houses) in the last 14 days. Participants also gave estimates of time spent in each location per week. High resolution satellite (Quickbird, Digitalglobe, CO) and digitized street maps were used during the interview to prompt recall and to mark the position of the places mentioned. Participant recruitment was not random: we used purposive sampling and focused on two Iquitos neighborhoods participating in an ongoing longitudinal study on dengue epidemiology [32], [36], seeking a balanced number of males and females representing age ranges between 7 and 74 (see Table 1). We only excluded those who planned to spend more than a day outside of Iquitos during the following 14 days. Recruitment was performed by trained local technicians who provided a description of the study together with a pamphlet with specific information about the GPS units and the study in general [16]. In the first phase, conducted between September 2008 and March 2009,59 participants were asked to use the GPS units at all times for a period of 14 days and respond to the SSI on day 15 asking for all the places they visited while GPS-tracked during those 14 previous days. The research team scheduled an exchange of the GPS units every three days to download data, verify function, and recharge batteries. At the time of GPS unit exchange, participants were asked about their experiences with the GPS, whether they had used it, if it had been forgotten and, if so, on what days. GPS units were programmed to track a person' s position (latitude, longitude and time stamp) every 150 seconds. The second phase was conducted in July and August of 2010 with 101 participants, who were asked to follow the same procedures as before; use the GPS unit for 14 days and respond to the SSI on day 15. One component was added in this phase: within 3 days of data collection, survey data was entered into a database and GPS-collected data was processed so that information on the locations identified as visited by each method were overlaid in a Geographic Information System (ArcGIS 10, ESRI). With a series of maps noting the position of each place visited by either method, field technicians returned to the participants within 4–5 days to ask them about any discordant information (i. e. , locations on the survey, but not registered on the GPS or vice versa). For Phase 2, the GPS collection frequency was increased to every 15 seconds (45 participants) and 90 seconds (56 participants) to assess the impact of data collection frequencies on GPS-SSI concordance. Whereas with 150 second programming, we could collect and recharge GPS units every 3 days, individuals wearing GPS units programmed at 15 and 90 seconds were provided with a charger and asked to charge the units daily because of the reduction in battery life. Our sample size was sufficient for a descriptive analysis and was limited due to intense participant follow-up for ∼20 days; i. e. , recruiting and consenting, distributing GPS units, exchanging charged GPS units and collecting ones losing power, interviewing participants with SSI at day 14, geocoding locations immediately, inputting all data from GPS and SSI to overlay in a GIS, returning to participants for follow up interview. Considering these complexities, participant recruitment was limited to what was logistically feasible for our field teams. All locations reported on the SSI were identified in the Iquitos GIS and received a unique location code with geographic coordinates that link directly to a SQL database containing participant information. If the location was not already in our system or if there were doubts about the specific location, a research team member went to the described place to assign a geo-code. Based on geo-referenced city-block maps (courtesy of the Peruvian Navy) and field sketch maps, geo-referenced aerial photographs and high resolution satellite imagery (Quickbird, Digitalglobe, CO), a total of 48,365 Iquitos lots were digitized prior to initiation of this study. Given the lack of a formal and consistent address system, we assigned a unique code to each lot. A local GIS specialist on our research team updates the maps on a regular basis, making the Iquitos GIS one of the most complete and up to date geo-spatial databases generated for a resource-poor city of its size. To obtain locations recorded by GPS units, the raw data was processed using an agglomerative algorithm (i-Cluster [15]). In simple terms, when GPS raw data was plotted over a satellite image of the city, we observed “clouds” over specific locations that were frequented by an individual [15]. These “clouds” mark locations that are the product of the frequency of going to that place and the time spent there. This data reduction algorithm works by aggregating consecutive GPS readings that are within a spatial (d) and temporal (t) window, and estimating the total time a participant spent within such a spatio-temporal buffer [15]. The algorithm also allows for identification of locations intermittently visited by applying a threshold time (tintv) in between visits. Based on the inherent spatial error of GPS data (e. g. , 5–10 m) we determined the following configuration: d = 20 m, t = 15 min and tintv = 30 min, for tracking Iquitos participants. The resulting place derived from the i-Cluster algorithm was then manually assigned the nearest location ID in the Iquitos GIS. For the analysis, we directly compared the raw GPS data to the SSI data. Because we know the exact GPS coordinates of every location reported in the SSI data, we could test to see how frequently the GPS unit reported that the individual was in the vicinity of each location. Specifically, for every participant and every location they visited, we calculated the distance from every GPS point registered for that participant to that location. For many locations, we have not just the location, but the footprint of the structure as a polygon within the Iquitos GIS. As such, we could calculate the distance from each GPS point to the boundary of each location (taking GPS points that were within the polygon to have distance zero from the structure). For both locations that we have the footprint of the structure and those that we just have a single GPS location representing the centroid of the building, we consider the location “visited” if there are a sufficient number of raw GPS points within a certain threshold distance of the location. We then vary the number of raw GPS points deemed sufficient (here we used 1,5, and 10 points), as well as the distance threshold selected (defined as the distance allowed for what constitutes a “match” between locations recorded in the SSI compared to a nearby GPS point, in this study, ranging from 0 to 100 meters), to investigate the sensitivity of visitation. We quantified the concordance between SSI and GPS in identifying places visited by participants by comparing the interview locations with (1) i-Cluster-derived locations and (2) raw GPS positions. To compare the interview with the i-Cluster inferred locations we mapped the locations identified by each method in a GIS (ArcMap 9. 3; ESRI). Locations identified both by the GPS and the SSI were considered “concordant” and did not require follow up. All locations that were captured by either GPS or the SSI, but not both, were considered “discordant” and a research assistant was sent back to the participant' s home to ask them about the potential causes of discordance. Before interviewing each participant, the research assistants checked the original SSI to determine how the respondent had described the location (e. g. , “aunt' s house” or “internet cabin”) or the GIS maps to locate a nearby reference point that might help the participant identify each discordant location (e. g. , 2 blocks from market). Research assistants (nurses and biologists) were native Iquitos residents who received specific training on all steps of the interview process to ensure they were aware of sensitive issues they might encounter both when gathering initial SSI information, as well as while following up with discordant locations. Participants were given a 24–48 hour period to decide whether to participate or not in the study. For children, verbal assent of the minor and written consent of the parent or caretaker were required, whereas for adults, a written consent was required. After GPS data collection, a strict protocol for storage (in a secure MySQL database) and management was followed. The procedures for enrollment of participants and GPS data management were approved by the Institutional Review Boards (IRB) of the University of California at Davis (2007. 15244), Emory University (IRB9162) and Tulane University through an inter-institutional IRB agreement with the United States Naval Medical Research Center Unit No. 6 (NAMRU-6). The NAMRU-6 IRB, located in Peru, also reviewed and approved the study (NMRCD 2007. 0007). This IRB functions as a Peruvian IRB and is registered with the Peruvian Regulatory Agency for Clinical Trials with the number RCEI-78.
More than half of the 160 enrolled participants were females (58. 5%) (Table 1). The lower number of males was due to the difficulty in finding them at home during regular interviewing hours. Recruitment was stratified by age; the age range sampled was 7 to 74 years. Recruitment varied across age groups (range of 25–46 per age group), with 7–18 year olds accounting for 28% of the tracked individuals (Table 1). Although not perfectly balanced among sexes and age groups, the recruited population represents a large and diverse demographic sample of the local population. Of the 2,566 locations identified by SSI and/or i-Cluster algorithm, 14. 3% were concurrently identified by both (i. e. , concordant). SSI identified 2. 3 times more locations than the i-Cluster algorithm, with residential (42. 5%), commercial (26. 4%) and educational (10. 8%) spaces accounting for the highest degree of concordance between methods (Table 2). A total of 2,047 places were reported in the SSI as visited by all participants over the 14-day tracking period (of these 2047 places mentioned, 1057 were unique places, see Table 2). Most (96. 7%) places were located within the urban and peri-urban areas of Iquitos (Figure 1A). Participants reported visiting a median (Q1–Q3) of 12 (9–16) places over the 14-day period, with the number of places not differing significantly between sexes (Wilcoxon rank sum test with continuity correction, W = 3140. 5, P = 0. 89). The most commonly reported location types on SSI that were not visualized using the i-Cluster algorithm (considering 1609 locations with land-use information) were commercial locations (34. 2%) followed by residential (22. 1%) and recreational (17. 0%) locations (Table 2). The i-Cluster algorithm identified a total of 886 places as visited by participants while tracked (716 unique locations); 98. 7% of which were found within the urban and peri-urban areas of Iquitos (Figure 1B). A significantly lower median (Q1–Q3) number of places per participant was registered by the i-Cluster algorithm in comparison to the SSI (7,4. 0–10. 0; W = 11990, P<0. 001). Residential spaces represented 58. 6% of the 454 i-Cluster-identified locations with land-use information that were not reported on the SSI, followed by commercial (11. 4%), educational (4. 2%), and recreational (3. 5%) locations (Table 2). Locations with highest percentage of concordance (i. e. , per type of location, the number of concordant sites divided by the total number of sites obtained for that type of location through SSI and/or GPS) were educational settings (24%), followed by residential (19%), other (18%), and religious or market spaces (both at 13%). When the SSI-reported locations were compared to the raw GPS data (Figure 1), differences in the concordance between methods were observed based on the location type, distance threshold selected, GPS collection frequency and number of GPS points considered to define a visit (Figure 2). Both GPS and SSI had perfect concordance in identifying each participant' s home (see Figure 2) at either combination of collection frequency, distance or number of points. There was more concordance for residential sites than non-residential sites at 15 and 90 seconds collection frequency; this difference was minimal at 150 seconds (Figure 2). Not depicted due to the small numbers in each category, there was much variation in concordance when examining by type of location. For example, when examining specific categories such as schools, “other” (ports, storage buildings, empty lots) and lodging places (i. e. , rustic “hostels” for visitors from outside Iquitos, or couples might go for a few hours) there was a concordance of 80–100% between methods, whereas other residential places (i. e. , friends' or relatives' homes), commercial locations (i. e. , shops, markets) and religious buildings (i. e. , churches) showed a concordance of 50–80%. As distance from the SSI reported location increased, the concordance between SSI and raw GPS data increased, independently of the type of location (Figure 2). When at least one raw GPS point was considered (solid lines in Figure 2), concordance between methods was highest at up to 20 meters from each location. Beyond that distance, no dramatic increases in concordance were observed. There was less concordance when we restricted our analysis to 5 GPS points (broken lines) or 10 points (finely broken lines), but the pattern was similar to the line created when 1 point was considered a match. Interestingly, increasing frequency of GPS data collection from 150 to 15 seconds was not associated with a proportional increase in concordance between SSI and GPS (Figure 2). Battery power loss observed at 15 second collection frequency may help explain such results: of the 508 GPS exchanges performed, 56 (11%) of GPS units programmed to collect data every 15 seconds had issues due to battery loss at the time of data download in comparison to 2% (9/379) for GPS units programmed to collect data every 150 sec. At 20 meters from each SSI location, and when 1 GPS point was considered to define a match, overall concordance averaged 72. 6% (SD: 20. 7%) for 15 seconds, 65. 8% (30. 8%) for 90 seconds and 70. 3% (23. 3%) for 150 seconds (Figure 3). When ten points were required to define a match, concordance was reduced to 59. 1% (31. 6%), 54. 3% (31. 0%), and 55. 7% (30. 7%) respectively (Figure 3). Cemeteries, public buildings, recreational areas and health centers were the location types that consistently showed the lowest concordance values (Figure 3). Increasing the data collection frequency from 15 to 150 seconds did not translate into significant variation in the concordance between SSI and GPS across all location types (average [min-max] variation across locations, 2. 3% [0. 7%–9%]) (Figure 3). In comparison to using the raw GPS points (Figure 2), the i-Cluster algorithm evidenced much higher discordance rates for all location types (Table 2). However, this method allowed identifying a total of 519 locations not mentioned in the SSI and not able to be inferred when the raw GPS positions were visualized (Table 2). In Phase 2, with the subset of 101 participants, we further explored the possible causes of discordance between GPS and SSI. Specifically, within 2–3 days of administering the SSI, we used GIS to develop maps identifying “discordant” SSI and i-Cluster locations (Figure 4) (i. e. , locations that were only mentioned in the SSI or only visualized using the GPS data). These maps were used when probing participants about possible causes of discordance. In this phase, regarding locations identified on the SSI, but not detected by GPS (total of 656 locations, Table 3), the most common response to questions about the discordance was an affirmation that these locations had been visited (35. 8%) – they could not explain the discordance. The second most common response was that units had “seemed to be turned off” (30%). Indeed, GPS units initially deployed could accidentally be turned off, so respondents who noticed the lack of a flashing blue light inferred correctly. Once this problem was reported, we programmed GPS units to not allow them to be turned off manually, reducing this problem half-way through this study. Other explanations for the discordance included those who admitted forgetting to take units to some locations (12. 5%; i. e. , rushing out and simply forgetting), not wearing the GPS units to locations that were near their house (3. 4%) or to locations where they might get stolen (3. 5%), and leaving units home to recharge (4. 6%). A small percentage (4. 7%) affirmed having the GPS unit in some locations, but questioned whether the placement of the GPS unit in their purse might have impeded the signal. Regarding locations identified on the GPS unit but not mentioned in the SSI (204 locations, Table 3), the most common response was that they simply forgot to mention it in the SSI (38. 2%), and a few made the additional observation that they had forgotten this location because it was not part of the regular routine (15. 2%). Some locations were not mentioned (until probed directly about them) because they were either transient or en route to another location (22. 1%; i. e. , a path always taken, a bus stop) or because they were outdoors (13. 2%; i. e. , outdoor food kiosk). After further examination of the reasons for discordance between SSI and GPS, we identified 75 locations as being affected by technical failures in generating the maps (the locations were not properly mapped or marked the location next door, 57. 3% and 42. 7% respectively, and hence were incorrectly considered discordant at the time of interview).
GPS technology is increasingly used in behavioral research. Its use has moved beyond feasibility tests [15], [35], [37], [38] to the actual use of GPS-enabled devices (often coupled with other sensors such as accelerometers, air pollution sensors or cameras) in studies quantifying various aspects of human mobility and spatial behavior [7], [10], [11]. As the technology continues to be embraced by researchers across disciplines, it is easy to assume that due to the wealth and resolution of the data it provides, some might consider GPS data to be a “gold standard” for mobility research and a replacement of classic survey instruments [35]. By performing a field validation study tracking 160 individuals, we assessed both the limitations and possibilities of GPS technology for mobility research, and provided evidence of multiple sources of error/uncertainty that can affect quality of data in comparison to survey methods. It is important to mention here that based on our experience, we would expect different results with different GPS units, different SSI and other methods of data analysis. Under perfect conditions of satellite geometry and signal strength, GPS provides very accurate information about the position (latitude, longitude, elevation, time of day) of any stationary object on earth. Wearable GPS devices provide all the essential pieces of information to reconstruct and quantify human movement: positions associated to places visited, time stamp for each potential visit, and routes followed to connect visits. Given technical (e. g. , signal noise, multipath errors, signal obstruction inside buildings, battery life) and human behavioral limitations (e. g. , compliance of use, individuals forgetting to take or charge units), GPS signals are prone to error and estimates of mobility parameters that they generate are considered uncertain. Signal processing algorithms have been developed to reduce such errors and improve interpretation of complex data [39]–[44]. In our study, the application of a signal clustering algorithm (i-Cluster) allowed identifying locations where individuals spent their time, but also added significant uncertainty by flagging locations transiently visited (e. g. , a bus stop; 35. 3%). Such errors were the main contributor to the 85. 7% discordance between methods observed when i-Cluster inferred locations were considered. Because most research describing automated algorithms rely on single (or few) days of data or low sample sizes [39]–[44], the errors found by our study are a likely outcome of the type of error those algorithms may encounter if applied within the same context. Our results can be used as a guide for the development of improved and more accurate methods for GPS location extraction and human movement quantification. An interesting finding was that higher GPS collection frequencies (e. g. , 15 seconds) were not associated with a proportional and significant increase in concordance between methods. Issues of battery life, not securing the “off” option at the start of the study (remedied quickly), and compliance of participants in charging the units compromised the quality of data collected. Similar issues were observed across multiple studies quantifying physical activity [17], [35]. Implementing GPS-enabled smart-phones could have reduced the issue of battery loss, because there is more motivation for individuals to charge the phones overnight and to use them during the day. Because Iquitos is slowly making its transition into smart-phone technologies, different issues were pointed out by a subset of 10 participants when asked about the possibility of using GPS-phones instead of data-loggers: (a) older individuals were intimidated by the technology and by the possibility of having the units stolen (the latter was a concern shared by individuals across all age groups), and (b) school age children mentioned they are not allowed to take phones to elementary or high school or locations where their phones could be taken by older children [7]. When cell-phones can be properly deployed they can provide valuable information. For instance, in Canada a study comparing GPS data collected by cell-phones and self-reported surveys reported (using rudimentary indices of concordance such as convex hulls and kernel density estimations) that 75% of questionnaire-reported activity locations were located within 400 meters of an activity location recorded on the GPS track [26]. In weighting the possibility of adopting novel technologies, consideration of cultural and local concerns will be key for both GPS and SSI instruments [13], [16], [18]. Turning the large amounts of raw GPS positional data into meaningful locations individuals visited is another challenge. Unprocessed raw GPS data can be used to either describe zones or areas in which individuals spend their time or to assess the accuracy of the GPS in identifying precise locations against information provided by another method (i. e. , locations identified by SSI). In our study we implemented a simple algorithm based on an agglomerative clustering method (i-Cluster) to identify locations visited by individuals carrying a GPS unit. Our analysis shows that the algorithm presents low levels of sensitivity and specificity in identifying places reported as visited by participants. This poor performance could be due to: (a) the algorithm' s limited ability to account for changes in accuracy of the GPS signal or to the occurrence of intermittent positions as a consequence of GPS signal loss and (b) the fact that not all reported locations were actually visited by participants while tracked. More complex methods of location extraction that account for signal errors, such as hierarchical dynamic Bayesian network models [39], [41], [44], are being currently developed and are viewed as a promising means of reducing the uncertainty associated with the identification of locations visited by participants [39], [44]. Once those methods are validated, their integration into health research applications will increase our ability to accurately infer the location of potential infectious disease exposure areas. Classic methods (surveys, diaries) have long been considered too limited to quantify behavior due to marked heterogeneities in the ability of individuals to recall the locations they visit, interviewer error, behavior changes and issues associated to privacy [35]. By working with the local community, addressing potential cultural barriers and concerns and adapting the language of interviews, we developed a culturally-sensitive SSI to quantify movement (and potential exposure to dengue). Our comparative analysis shows that, for a 14-day recall period, interviews provide accurate estimates of the locations visited by people (of a total of 892 locations for which we investigated causes of discordance, only 109 [12. 2%] were visited and not reported). The SSI not only identified places, but also characterized the context of visits (i. e. , grandmother' s house), information impossible to obtain directly from GPS. SSI data entry and processing are much more straightforward and faster than of GPS: (a) maps with marked locations were digitized in the Iquitos GIS and each premise reported as visited was assigned a location code and (b) the location code was then linked to the database containing all the SSI information. We concluded that a validated survey instrument that can be adapted to different contexts can be used to understand the role of human mobility in infectious disease dynamics. We encountered several limitations in our study design. Although our sample size was relatively large, the low numbers of participants assigned to each age group precluded statistical tests to look at different causes of discordance. Given that we needed to obtain results quickly to ask participants about possible causes of discordance, we relied on a single GPS data reduction algorithm (i-Cluster). As observed on the survey (Table 3), most of the discordant records occurred due to this algorithm providing false positive or negative results. Since the time this study was performed, new and more sophisticated methods to process GPS data have been developed [39]–[44], as well as more accurate and less error-prone GPS units have likely become available. Future research will involve performing comparative studies to quantify sensitivity/specificity as well as applicability to specific study questions. Also, we considered that our concordance estimates could be, in part, dependent on size and placement of houses in Iquitos. An average household in this city measures 5 m in width, which is within the mean error of a GPS (5–10 m). This could explain the high percent (∼60%) of residences identified by GPS that were not reported on the SSI. Thus, accuracy in identifying locations is not only dependent on the factors explained above, but also on key attributes of the urban landscape (e. g. , household size, prevailing building material, density of high-rise buildings, vegetation cover). We did not test for differences in the SSI results of participants with more contact with our research team (i. e. , those with more frequent GPS exchanges due to differing data collection times) compared to those with minimal contact. We do not expect differences, however, because contact was focused on the GPS exchange and SSI questions about their movement and activities were only asked at the end of the 14 day period. None of the participants, therefore, had an advantage over others regarding the types of questions they would be asked. We also did not estimate nor compare the cost and technical expertise to apply and process by these methods. Both the GPS and SSI capture very complex data. GPS data is in digital form, but needs to be processed. SSI data needs to be verified (i. e. , in our study, someone might go to a location described to geocode the location), entered and mapped. There were costs associated to purchasing GPS units (∼$49/unit), training personnel to set and distribute units, downloading and analyzing the GPS data. Similarly, there were costs associated to developing, refining and improving the SSI, training personnel to apply it, and entering the data in a GIS system. Ultimately, decisions regarding using an SSI or GPS units in a study depend strongly on the study question and the urban context, because both SSI and GPS can provide different but equally valuable information that need to be carefully weighted at the planning stage. For infectious diseases in general, and vector-borne diseases in particular, the need to tie potential exposure to specific locales requires the retrospective investigation of multiple routes of pathogen transmission. Survey instruments like the one we developed in this study not only provide accurate information of places visited, but can also be used to retrospectively infer the likely location where infection occurred [5]. This need to tie exposure to a specific place (s) has limited the use of GPS technology in infectious disease research, but GPS technology could be used in prospective movement studies or in studies obtaining information provided by phone companies. As observed in our study, once locations are identified, the raw GPS positions can be analyzed to quantify temporal patterns of mobility (days and times a person visits such locations, regularity of visits, overlap with other tracked individuals) and to accurately quantify routines and movement of a large segment of a population. This way, key information about mobility and behavior can be inferred and used to parameterize mathematical models that allow better forecasting of disease transmission or design policies targeting activities or segments of the population at greatest risk. No gold standard exists for obtaining and analyzing human mobility data, instead different errors may occur with different methods. Despite the continually improving accuracy available with GPS, barriers persist, including: behavioral aspects (i. e. , people remembering to use the unit), technical aspects (i. e. , accuracy of 5–10 meters in a location with houses averaging 5 meters width), and analytical aspects (i. e. , differences in concordance based on method of analyzing complex data as reported in this article). The SSI is not a gold standard either. Even with the possible drawback of more locations reported than true (i. e. , false positives), compared to GPS units, the SSI provided more true locations, more context about locations, and data were easier to process and analyze. For our study, in which we needed to identify locations retrospectively for possible exposure to dengue virus, the SSI was the only choice because of the logistical and financial difficulty of fitting GPS units on a large sample and, even if that had been possible, being able to quickly identify locations recently visited within a short enough time frame to initiate our possible exposure investigations. For now, SSI remains the most comprehensive method to identify such locations. | Being able to quantify human movement is important for studying activity patterns, exposure to pathogens and developing realistic infectious disease models. We compared fine-scale human mobility data obtained by Global Positioning System (GPS) -enabled devices and semi-structured interviews (SSI) from 160 individuals in Iquitos, Peru, in order to assess the quality of data using these two different approaches and our ability to measure fine-scale human mobility patterns in a resource-poor urban environment. Using various methods to process the GPS data, we found the SSI identified more locations a person had visited than GPS. Though the GPS gave more precise data, there were behavioral, technical, and analytical barriers. The SSI provided richer context and was easier to process, but also had more false positives. SSI was the only option for identifying locations retrospectively. | Abstract
Introduction
Methods
Results
Discussion | public and occupational health
infectious diseases
geography
medicine and health sciences
global health
earth sciences
social sciences | 2014 | Strengths and Weaknesses of Global Positioning System (GPS) Data-Loggers and Semi-structured Interviews for Capturing Fine-scale Human Mobility: Findings from Iquitos, Peru | 9,039 | 188 |
Coding of multiple proteins by overlapping reading frames is not a feature one would associate with eukaryotic genes. Indeed, codependency between codons of overlapping protein-coding regions imposes a unique set of evolutionary constraints, making it a costly arrangement. Yet in cases of tightly coexpressed interacting proteins, dual coding may be advantageous. Here we show that although dual coding is nearly impossible by chance, a number of human transcripts contain overlapping coding regions. Using newly developed statistical techniques, we identified 40 candidate genes with evolutionarily conserved overlapping coding regions. Because our approach is conservative, we expect mammals to possess more dual-coding genes. Our results emphasize that the skepticism surrounding eukaryotic dual coding is unwarranted: rather than being artifacts, overlapping reading frames are often hallmarks of fascinating biology.
Any stretch of DNA contains six reading frames and can potentially code for multiple proteins. Situations when two partially overlapping reading frames code for functional polypeptides (dual coding) are quite common in bacteriophages and viruses (e. g. , φX174, HIV-1, hepatitis C, or influenza A), where constraints on the genome size are strict. On the other hand, dual coding in vast eukaryotic genomes was reported to be scarce and restricted to short regions with secondary reading frames having poor phylogenetic conservation [1]. Yet, three known human genes (GNAS1, XBP1, and INK4a; Figure 1) defy this pattern by having long, well-conserved dual-coding regions (e. g. , dual-coding region in XBP1 is conserved from worms to mammals [2]). In addition, the three cases exemplify some of the most striking biological phenomena and invite us to look at dual coding in greater detail. In GNAS1, a single transcript simultaneously produces the alpha subunit of G-protein from the main reading frame, and a completely different protein, ALEX, using a +1 frame [3]. A transcript of XBP1 can produce only a single protein at a time and uses the endonuclease IRE1 to switch between two overlapping reading frames [4]. INK4a generates two alternative transcripts that use different reading frames of a constitutive exon for translation to tumor suppressor proteins p16INK4a and p14ARF [5]. Although GNAS1, XBP1, and INK4a are drastically different, there are striking parallels in the way they function. Products of the main and alternative reading frames perform related tasks, either by binding and regulating each other (GNAS1 and XBP1), or by complementing each other in performing a common function (INK4a) [6–8]. Dual coding is a costly arrangement because it limits the flexibility of amino acid composition [9]. A silent change in one frame is almost always guaranteed to be amino acid changing in the other. Although counterintuitive, this codependency may in fact lead to an increase of the apparent substitution rate when two frames become locked in an evolutionary race of compensatory changes. A chief example of this is the mammalian GNAS1 locus, where the overlapping reading frames accumulate substitutions so fast that primate and rodent sequences become virtually unalignable [10]. Yet despite this cost, the dual coding in GNAS1, XBP1, and INK4a is preserved throughout mammalian taxa [10,11]. Are overlapping reading frames a new avenue for encoding functionally linked proteins?
Before describing our analyses, we define terms used in this paper. A dual-coding gene contains two frames read in the same direction: canonical (annotated as protein coding in literature and/or databases) and alternative. The alternative reading frame (ARF) is shifted forward one or two nucleotides relative to the canonical frame (+1 and +2 ARFs, respectively). To identify dual-coding genes, we used a comparative genomics strategy, because all presently known alternative reading frames are conserved in multiple species. For example, ARFs in Gnas1, XBP1, and INK4A are conserved in all sequenced mammals [8,10,12]. To reliably find new dual-coding genes, we must determine how likely they are to occur by chance. Simulations designed to answer this question show that dual coding is statistically unlikely, suggesting that if overlapping coding regions are detected in orthologous sequences, they have a high chance of being truly functional. To determine a length threshold for identification of dual-coding regions (what is the longest dual-coding region that can arise by chance?), we conducted the following experiment. First, we generated alignments between 14,159 orthologous canonical reading frames from human and mouse transcripts (sequences, canonical frame boundaries, and orthology assignments were obtained from the Ensembl database at http: //www. ensembl. org). We chose these two species because they have the highest number of annotated transcripts. Next, we “disassembled” all 14,159 human/mouse alignments into codon columns. By randomly picking codon columns from the previous step, we generated 10,000 simulated alignments with 5,000 columns each. Finally, we scanned simulated alignments for the presence of ARFs and built a length distribution (Figure S1). Only 0. 1% of +1 ARFs were ≥500 bp, while none of the +2 ARFs extended beyond this threshold (the longest was 492 bp in the simulation). A possible weakness of this approach is the assumption of codon independence, for it is well-known that protein-coding regions possess Markovian properties [13]. To address this issue, we conducted codon-based phylogenic parametric simulations, which do not break open reading frames (ORFs), and estimated codon frequencies from gene alignments with at least three taxa, which contained conserved, long +1 ARFs. Only 0. 3% of simulated alignments preserved ARFs with 500 or more nucleotides (Figure S2). Thus, both simulations suggest that only a negligible amount of random dual-coding regions will reach 500 bp, and we set this length as the threshold for defining ARFs in orthologous coding regions. Using 500 bp as the lower bound, we identified 149 ARFs that were conserved in human and mouse. An example is shown in Figure 2 (see Figures S3 and S4 for procedure steps and detection of ARFs from multiple alignments). Although all 149 candidate ARFs were conserved in the two species and were longer than the empirically derived threshold, some could still be false positives. For example, the amino acid sequence of the canonical protein may dictate specific codon composition, which in turn may render the nucleotide sequence of the canonical frame such that an ARF can be relatively long simply as an artifact of the codon usage pattern (e. g. , having low complexity regions, or avoiding “problem” codons; see Table S2). To remove potential false positives, we developed the codon column replacement test (CCRT; see Materials and Methods). CCRT estimates how likely a given alignment is to contain an ARF by chance. If an ARF has a CCRT score of ≤5%, it is considered a reliable prediction. From the total of 149 ARFs, 66 satisfied this criterion. To make our final set even more conservative, we considered only those of the 66 ARFs that were conserved in at least one other species (rat and/or dog) in addition to human and mouse. The conservation requirement reduced the final set to 40 ARF-containing transcripts, which we examined in detail (Table 1). Note that our criteria are very conservative because (1) a number of true ARFs may be shorter than 500 bp (261 bp and 210 bp in XBP1 and Ink4A, respectively) and (2) transcript data for dog and rat are incomplete, which may have led to the exclusion of some true ARFs. Genomic location of the ARFs are provided in Table S4 and can be visualized as a custom track at the University of California Santa Cruz Genome Browser [14] (a link is provided at http: //nekrut. bx. psu. edu). Table S3 lists assignment of ARF-containing genes to Gene Ontology categories. Previous studies of ARF-containing genes showed that the region of overlap between canonical and alternative reading frames evolves under unique sets of constraints. If both proteins (encoded by canonical and alternative frames) are functional and maintained by purifying selection, the codependency between codon positions would manifest itself in a nucleotide substitution pattern that is sharply different from the one expected in single coding regions [10,11]. The difference in patterns can be used to test whether the dual-coding genes identified in our study are real. We developed two new approaches for the analysis of nucleotide substitutions—a codon substitution model for overlapping reading frames and a transition/transversion ratio test—to narrow the list of potential dual-coding genes to 15 high-confidence candidates. The codon model estimates five substitution rates for the overlapping reading frames by considering all 64 possible codon contexts for each one-nucleotide codon substitution in a given frame, and weighting each context based on its relative frequency in the extant sequences (see Materials and Methods). One of the rates, βSTOP, which measures the propensity of substitutions in one frame toward introduction of stop codons in the other frame, is especially useful for testing the reliability of ARF predictions. This quantity measures the admissibility of stop codon–inducing contexts in the evolutionary past of the sample and is zero or near zero in functional ARFs. For example, when applied to biochemically characterized ARFs in Gnas1 and XBP1, the hypothesis of βSTOP being exactly zero cannot be rejected (p = 0. 5 from likelihood ratio test). For 34 candidates, the hypothesis βSTOP = 0 could not be rejected. From a series of parametric simulations we estimated that at p = 0. 05, the test fails to reject the null hypothesis for 6% of the datasets that were simulated using a single reading frame model. To confirm our results using an independent nucleotide-based approach (as opposed to the codon-based test described earlier), we applied the transition/transversion (κ) ratio test to make inferences about biological significance of ARFs. The test is based on the following reasoning: in most standard protein-coding regions (with only one reading frame), κ at the third codon position (κ3) is significantly different (higher) than at the first and second codon positions (κ12), so that κ12 < κ3 [15]. This is because most substitutions at the third codon position are synonymous, whereas in the first codon position all but eight substitutions are nonsynonymous, and all substitutions in the second codon position are nonsynonymous. By contrast, in overlapping reading frames, codon positions are codependent. For example, in a +1 ARF, the third codon positions correspond to the first codon positions of the canonical frame. Thus, almost every change in the third codon position of the ARF is guaranteed to change amino acids encoded in the canonical frame. However, if the ARF encodes a truly functional product, purifying selection would resist such changes, and the condition κ12 < κ3 would not hold. This gives us the opportunity to test functionality of ARF in our dataset by contrasting two hypotheses: H0: κ12 = κ3 (ARF does encode functional polypeptide) and HA: κ12 < κ3 (ARF does not encode functional polypeptide). To perform this test, we used a maximum likelihood framework to test κ12 and κ3 for equality [16]. Application of the test to our list of dual-coding genes identified 18 candidates. Intersecting the results of the tests yielded 15 dual-coding genes as high-confidence candidates. The small number of species used in this study (four; a currently unavoidable limitation given the low annotation quality of mammalian genomes) limits the statistical power of our analyses and explains why the other candidates did not pass this test. Similar analyses of Gnas1 and XBP1 genes used eight or more sequences [10,11]. Adding more sequences, which should be possible in the near future, will increase the number of high-confidence candidates. Although experimental confirmation of protein expression and genetic studies will ultimately answer this question, analysis of current literature provided us with clues to potential ARF functions. For example, one of the candidates is adenylate cyclase (ADCY8; Table 1), a membrane-bound enzyme that catalyses the formation of cyclic AMP from ATP [17]. A 534 bp ARF is located in the 5′-end of the ADCY8 transcript. The corresponding region of the canonical peptide has two distinct functions: it interacts with Ca2+/calmodulin and binds to the catalytic subunit of protein phosphatase 2A (PP2a; [18]). Such “multitasking” is one of the features of dual-coding genes, where separate functions are performed by products of canonical frames and ARFs [7,8, 19]. Two nucleotide substitutions affecting the amino acid sequence of ADCY8, W38A, and S66D (produced by mutagenesis) have conspicuous effects on ARF structure and calmodulin binding. W38A creates a stop in the ARF and disrupts calmodulin binding, but has no effect on association with PP2a. On the other hand, S66D does not disrupt ARF and has no effect on either calmodulin or PP2a binding [20]. Because in at least two instances products of ARF bind to the product of the canonical frame (i. e. , Gnas1 [6] and XBP1 [7]), we speculate that the polypeptide encoded by the ARF may mediate the binding of calmodulin by ADCY8. In fact, ADCY8 has a number of unidentified protein interaction partners from yeast two-hybrid screen experiments, one of which may be the ARF-encoded polypeptide [18]. Another gene in our set, Misshapen/Nck-related Kinase (MINK1; see Table 1), is involved in a number of functions related to cell spreading, fiber formation, and cell-matrix adhesion. MINK1 regulates the Jun kinase pathway (JNK) [21], is involved in thymocyte selection, and interacts with a large number of proteins controlling cytoskeletal organization, cell cycle, and apoptosis [22]. The MINK1 protein contains three functional domains (N-terminal kinase, intermediate, and C-terminal germinal center kinase) and exists as five distinct isoforms translated from alternatively spliced transcripts. All five transcripts contain an intact ARF, which covers the entire length of the intermediate domain. Extreme multifunctionality of MINK1 suggests that the ARF-encoded protein may be responsible for some of the functions. In addition, the intermediate region of the protein is the most variable in cross-species comparisons [23]. This provides additional support to the functionality of MINK1′s ARF: regions containing overlapping reading frames encoding functional proteins are likely to evolve faster in comparison with single-coding regions [10,11]. Retinoid X receptor beta (RXRβ; see Table 1) is a member of the retinoid X nuclear receptors that control transcription of multiple genes. In mice, RXRβ binds to the enhancer controlling major histocompatibility class I genes [24]. It is the only gene in our set in which the existence of the ARF was reported in the literature as an alternative N-terminus generated via alternative splicing [25], although this gene failed to pass our transition-to-transversion ratio test. Analysis of transcripts available for this gene shows that this was caused by the skipping of the second coding exon. Because the length of the skipped exon is not in multiples of three, this event switches the reading frame downstream of the splicing point. To recover the phase of the reading frame past the splicing point, the translation must be initiated at the ARF start codon. Because both transcripts (with and without a second exon) have identical 5′ ends, it is likely that the ARF is translated from the full-length transcript. Maintenance of dual-coding regions is evolutionarily costly and their occurrence by chance is statistically improbable. Therefore, an ARF that is conserved in multiple species is highly likely to be functional. Historically, dual-coding regions were largely overlooked as they violated the accepted views of the eukaryotic gene organization. For example, although the fact that XBP1 produces two proteins was known for years, only one of them was considered biologically important. The confirmation for the function of the second protein came only recently, when three groups described its roles [7,19,26]. Dual coding is also difficult to confirm experimentally and computationally. For example, one cannot use expressed sequence tags (ESTs) to confirm expression of ARFs because in the cases described here, the same transcript expresses both proteins via the use of alternative translation starts. Using initiation codon context or protein structure predictions are not guaranteed to confirm or refute ARF functionality either: the most impressive example of dual coding, Gnas1, has poorly defined Kozak motifs [27] and produces proline-rich polypeptides without clearly defined secondary structure elements [3]. However, analyses of confirmed dual-coding regions allowed us to highlight unique properties and to use them in a genome-wide scan that identified 40 candidates. Is this too much or too little? We emphasize that our criteria were set to be very strict to eliminate the noise. Therefore, the seemingly small number of candidates is likely just a subset of a larger “ARFome. ” First, some ARFs are shorter than the stringent length threshold of 500 bp that we have set to eliminate most false positives. For example, the length of the dual-coding region in human XBP1 is 261 bp [28], and is 210 bp in human INK4a [5]. Second, because only four species were included in the analyses of nucleotide substitutions, some dual-coding regions failed codon-based and transition/transversion ratio tests due to the lack of statistical power. As the annotation quality of other mammalian genomes increases, it will be possible to add more sequences into our analyses. Third, we required ARFs to be conserved in multiple species. A recent study has demonstrated that many dual-coding regions are specific to a narrow phylogenetic group (i. e. , primates [1]) and would not be detected by the current implementation of our method. None of the 40 genes identified in our study overlaps with Liang and Landweber' s dataset [1], as these authors primarily focused on short dual-coding regions arising from alternative splicing events. Finally, our approach assumes that the two proteins encoded by the dual-coding region evolve under a purifying selection regime as in all presently known mammalian dual-coding genes. This assumption was shown not to hold for some dual-coding regions of bacterial genomes [29]. Thus, 40 candidates is likely an underestimate. Improving annotation of additional mammalian species will allow us to conduct lower-stringency scans to define the size of the ARFome. Our study provides a robust statistical framework for detection and computational validation of dual-coding regions. This methodology will work equally well in genome-wide screens (this study) and in situations in which an ARF in a single gene needs to be evaluated. Take another look at your gene; you might find an unexpectedly simple explanation, a second protein from the alternative reading frame, for experimental results that are otherwise difficult to interpret.
CCRT estimates how likely an alignment is to contain an ARF by chance. The algorithm works as follows. Consider an alignment of human and mouse protein-coding regions similar to that shown in Figure 2. It contains two reading frames: canonical (ORF, white) and alternative (ARF, black). The objective of CCRT is to test whether the ARF is or is not the artifact of nucleotide composition imposed by the ORF. CCRT takes two inputs: the alignment we just discussed and a codon column frequency table. The codon column frequency table is similar to a codon usage table but instead of codons, it contains alignments of codons from at least two species (in our case, human and mouse). The codon column frequency table is generated by first aligning all possible orthologous protein-coding regions between two (or more) species, splitting these alignments into individual codon alignments, and counting the frequency of each codon alignment. For this study, the table was constructed by aligning ~9,000 orthologous protein-coding regions from human and mouse (alignments can be downloaded from http: //nekrut. bx. psu. edu). Given an alignment and the codon frequency table, CCRT generates multiple simulated alignments (in this study we used 10,000 replicates) by replacing the original codon columns of the alignment with ones drawn from the codon column frequency table so that the amino acid translation is preserved in the ORF. The probability of drawing a codon alignment from the codon column frequency table is proportional to its frequency. The ORF translations of all simulated translations are identical to the ORF translation of the original alignment, but are guaranteed to be different at the nucleotide level. Finally, each simulated alignment is translated in the ARF, and the number of alignments with the full-length ARF is recorded. This number serves as the empirical p-value. A low p-value (<5%) indicates that a small fraction of simulated alignments contain ARFs, and therefore the ARF is not an artifact of nucleotide composition imposed by ORFs and can be considered a true ARF. Consider an alignment of N codon sequences on S codons, which encodes two overlapping reading frames. We present the case in which the frames are shifted by one nucleotide relative to one another, but other cases can be handled by straightforward modifications. We refer to the two reading frames as F0 (frame 0) and frame F1 (frame +1). We also make use of the following notation: πabij denotes the frequency of dinucleotide ij in a and b codon positions (relative to F0) and πck denotes the frequency of nucleotide k in the c-th codon position. These quantities are estimated by observed counts from a given alignment. First, we define the model for codon evolution in F0. We discriminate four types of codon substitutions: SS (synonymous in both frames), SN (synonymous in F0 and nonsynonymous in F1), NS (nonsynonymous in F0 and synonymous in F1), and NN (nonsynonymous in both frames). We model the process of character substitution using a Markov process operating on codons and defined by the instantaneous rate matrix Q. Following the common practice of allowing nonzero rates for single instantaneous nucleotide substitutions only, we assign substitution rates α to all one-nucleotide SS substitutions, β01 to SN substitutions, β10 to NS substitutions, and β11 to NN substitutions. In addition, we introduce another rate—βSTOP—for all those substitutions that introduce a stop codon in one of the two frames. Because the evolution at a given position in F0 depends on the flanking nucleotides (two upstream and one downstream), we condition the substitutions at a codon in F0 on the values of the relevant nucleotides, compute transition probabilities for each of the 64 possibilities, and weight over the frequency distributions π12 and π3. Formally, the instantaneous rate of substituting a nonstop codon x = x1x2x3 with a nonstop codon y = y1y2y3 in F0 conditioned on the values of the two upstream nucleotides u1u2 and the downstream nucleotide d1: Conditioning on u1, u2, d1 is necessary to determine whether a substitution in F0 results in a synonymous or a nonsynonymous change in F1. Rnm denotes the rate of substitution for nucleotides n and m relative to that of A → G. We set Rnm = Rmn to ensure time reversibility. One can check that for any triplet u1, u2, d1, the equilibrium distribution of the Markov process defined by this rate matrix is Second, we describe an analogous rate matrix for F1. This rate matrix is conditioned on one upstream nucleotide u1 and two downstream nucleotides d1, d2. Transition matrices T (t) for the processes are matrix exponentials of Qt, for the appropriate rate matrix Q. For computational tractability, we assume that the evolution at codon c can be adequately described by computing the expectation over flanking upstream and downstream nucleotides. Specifically, if is the phylogenetic likelihood at codon c in frame F0, conditioned on the flanking nucleotides, then the unconditional likelihood can be computed as Analogous calculation can be performed for frame F1. Finally, we define the joint likelihood of the entire dataset (omitting the first and the last codons in F0) as Parameter estimates such as branch lengths and substitution rates can be obtained by maximizing the likelihood as a function of model parameters with standard numerical optimization techniques. Due to the structure of the genetic code, most of the possible single-nucleotide substitutions lead to nonsynonymous changes in at least one of the reading frames (Table S1). To evaluate the evolutionary regime in a multiple reading frame alignment, we test the null hypothesis to evaluate whether the introduction of premature stop codons is disallowed. The test defined a one-sided constraint on a single parameter, and the significance can be evaluated using the likelihood ratio test with the approximate distribution of the test statistic. | A textbook human gene encodes a protein using a single reading frame. Alternative splicing brings some variation to that picture, but the notion of a single reading frame remains. Although this is true for most of our genes, there are exceptions. Like viral counterparts, some eukaryotic genes produce structurally unrelated proteins from overlapping reading frames. The examples are spectacular (G-protein alpha subunit [Gnas1] or INK4a tumor suppressor), but scarce. The scarcity is anthropogenic in origin: we simply do not believe that dual-coding genes can occur in eukaryotes. To challenge this assumption, we performed the first genome-wide scan for mammalian genes containing alternative reading frames located out of frame relative to the annotated protein-coding region. Using a newly developed statistical framework, we identified 40 such genes. Because our approach is very conservative, this number is likely a significant underestimate, and future studies will identify more alternative reading frame–containing genes with fascinating biology. | Abstract
Introduction
Results/Discussion
Materials and Methods | evolutionary biology
homo sapiens
genetics and genomics
computational biology | 2007 | A First Look at ARFome: Dual-Coding Genes in Mammalian Genomes | 6,131 | 226 |
Cytotoxic T lymphocytes (CTLs) are the major killer of virus-infected cells. Granzyme B (GrB) from CTLs induces apoptosis in target cells by cleavage and activation of substrates like caspase-3 and Bid. However, while undergoing apoptosis, cells are still capable of producing infectious viruses unless a mechanism exists to specifically inhibit viral production. Using proteomic approaches, we identified a novel GrB target that plays a major role in protein synthesis: eukaryotic initiation factor 4 gamma 3 (eIF4G3). We hypothesized a novel role for GrB in translation of viral proteins by targeting eIF4G3, and showed that GrB cleaves eIF4G3 specifically at the IESD1408S sequence. Both GrB and human CTL treatment resulted in degradation of eIF4G3 and reduced rates of translation. When Jurkat cells infected with vaccinia virus were treated with GrB, there was a halt in viral protein synthesis and a decrease in production of infectious new virions. The GrB-induced inhibition of viral translation was independent of the activation of caspases, as inhibition of protein synthesis still occurred with addition of the pan-caspase inhibitor zVAD-fmk. This demonstrated for the first time that GrB prevents the production of infectious vaccinia virus by targeting the host translational machinery.
One major strategy of the host to survive the attack of viruses is to induce apoptosis in infected host cells. Cytotoxic T-lymphocytes (CTLs) play an important role in the apoptosis pathway, which activates a family of cytosolic proteins called caspases in target cells. When caspases are activated, they execute the key reactions that drive target cells to their demise. Activation of initiator caspases such as caspase-8 and 10 results in direct activation of the apoptosis executioner caspases like caspase-3 [1]. Caspase-8 and 10 also signal through the mitochondrial pathway by activating a protein called “BH3 interacting domain death agonist” (Bid) [2], [3], resulting in the release of cytochrome c (cyt c). Soluble cyt c also mediates the activation of the executioner caspases [1]. Thus, there is cooperation between the mitochondrial pathway and the caspase system. Active caspase-3 cleaves ICAD (inhibitor of caspase-activated deoxyribonuclease), with subsequent release of CAD and DNA degradation. Other substrates of executioner caspases include cytoskeletal and nuclear skeletal components like fodrin and lamin A, which result in cell shrinkage [1]. The mechanism by which CTLs activate the caspase cascade system has been an active area of research. We now know that electron dense granules found in CTLs carry cytolytic factors that trigger apoptosis in target cells. Granules polarize toward the immune synapse as the membranes of the CTL and target cell make contact. Cytolytic factors in the granules are then delivered to the target cell to induce cell death. Two of the first proteins to be isolated from these granules were perforin [4], [5] and granzyme B (GrB) [6]. Although purified perforin readily lyses cell membranes, perforin alone is not able to initiate DNA fragmentation in the same way as treatment with CTLs [7]. Combined treatment with perforin and GrB reproduces the effects of CTL treatment [8], by inducing both membrane damage and DNA fragmentation. GrB is a serine protease with an unusual substrate specificity, cleaving proteins at aspartic acid residues [9]. GrB is initially synthesized as an inactive zymogen that is activated by the removal of two amino acids at the amino terminus [10]. In the current model of CTL-mediated killing, perforin plays a role in granting GrB access to the cytosol of target cells [11]. Through proteolytic cleavage, GrB activates cytosolic substrates such as caspases [12], [13] and Bid, independent of caspase-8 [14]. Proteolytic activation of Bid results in heterodimerization with Bax (B-cell CLL/lymphoma 2 (Bcl2) associated X protein) and the subsequent recruitment of the Bid/Bax complex to the mitochondria. The Bid/Bax complex promotes mitochondrial membrane depolarization [15] and the release of cyt c and SMAC (second mitochondria-derived activator of caspase). SMAC binds and blocks the actions of caspase inhibitors, namely the inhibitor of apoptosis proteins (IAPs) [16]. Thus, GrB is a powerful pro-apoptotic factor that activates executioner caspases directly and through the mitochondrial pathway. A cell infected with virus becomes a target for destruction by CTLs via the GrB pathway. However, while under attack, the production of infectious virus can still occur and be released from the dying cell. Thus, viral infection of the host would ensue regardless of cell death. It would make sense if CTLs had a strategy to stop the virus in its tracks. In addition, to evade host defenses, viruses have evolved mechanisms to inhibit the caspase-cascade system. For example, vaccinia virus (VV) expresses the cytokine response modifier A (CrmA), which is a serine protease inhibitor or serpin. CrmA is a strong inhibitor of caspase-1,3, 6 and 8 [1], thereby acting to delay the process of apoptosis. Thus, it is possible for viruses such as VV to use dying cells for viral replication. The question arises as to whether CTLs have strategies to deal with this ongoing production of infectious virus. Recent literature reveals new roles of human granzymes that extend beyond apoptosis. Two of these emerging roles involve activation of the inflammatory system and direct antiviral functions that are independent of cell death. The ability of granzymes to induce and process pro-inflammatory cytokines demonstrated that granzymes activate the inflammatory system [17], [18]. Granzymes also have the potential to directly impair viral replication with an increasing list of host and viral substrates involved in the replication process of viruses [18]. GrH can cleave and inactivate host and viral substrates to prevent viral replication. Romero et al. demonstrated that GrH-mediated cleavage of the human protein La decreased hepatitis C virus IRES-mediated translation [19]. One of La' s roles is to stimulate IRES-mediated translation, a mechanism used by many viruses to take control of the host translational machinery. In a different study, Andrade et al. identified adenovirus protein 100K as a novel GrH substrate [20]. 100K is important in the assembly of the virus capsid and also functions as a GrB inhibitor. Thus, the authors uncovered a role in which GrH inactivated adenovirus replication in a pathway that could concomitantly de-repress GrB activity. It seems clear that granzymes have evolved unique antiviral strategies that are independent of their well-characterized cytolytic roles. In this study, we propose a novel mechanism for GrB to directly inhibit the production of VV proteins and thus, inhibit viral replication. In search for novel GrB substrates, we found eukaryotic initiation factor 4 gamma 3 (eIF4G3), a protein critical for initiation of protein translation. In humans, translation is largely regulated by eukaryotic translation initiation factors (eIF). Detailed reviews of the events involved in this process have been published earlier [21], [22]. Briefly, a newly transcribed host mRNA is transported to the cytosol in its linear conformation. Prior to the initiation of translation, a multi-subunit complex called eIF4F needs to be formed. Subunits within the complex have important roles. First, to flank the mRNA, the eIF4E subunit binds to the 5′ end of the mRNA, while the poly (A) tail binding protein (PABP) binds to the 3′ end. Following this, the large scaffold subunit eIF4G3 circularizes the mRNA molecule by binding to both PABP and eIF4E. The eIF4G3 protein is a central player in the initiation of translation since it also contains binding domains for mRNA, eIF4A, and eIF3 (itself a multi-subunit factor that recruits the ribosome). It is not surprising that many viruses interfere with the function of the eIF4F complex to gain control of translation. Some viruses cleave eIF4G3 to halt host protein translation while other viruses use eIF4G3 to make viral proteins. For example, rhinoviruses and enteroviruses shut off host protein synthesis by cleavage of eIF4G3, resulting in inhibition of cap-dependent translation [23]. Cleavage of eIF4G3 is carried out by the viral-encoded protease 2Apro. In contrast, most viruses including poxviruses, adenoviruses, papillomaviruses, polyomaviruses, herpesviruses, and asfarviruses do not cleave eIF4G3, but instead, take full advantage of the eIF4F complex [24]. VV, a double-stranded DNA virus of the poxvirus family, cannot replicate in the absence of eIF4G3 [23], [25], [26]. VV translates its RNA in a cap-dependent manner, increases eIF4F complex levels and enhances PABP binding [24], [27]. Thus, we used VV as a model system to study the role of eIF4G3 degradation by GrB in blocking viral infection. We hypothesized that in addition to pro-apoptotic roles, GrB also targets the translational machinery by cleaving eIF4G3. In the absence of a functional eIF4G3, translation initiation would be arrested. In this study, we demonstrate that eIF4G3 cleavage by GrB is the critical event that prevents VV protein synthesis and reduces subsequent production of infectious VV particles.
Jurkat cells were obtained from American Type Culture Collection (ATCC). Jurkat cells were passaged approximately every 2 days when cell density reached 1×106 cells/ml to bring cell density back to 1×105 cells/ml. Cells were grown in RPMI supplemented with 10% fetal bovine serum (FBS), 0. 06 mg/ml penicillin G, and 0. 01 mg/ml streptomycin sulfate. Cell number was then adjusted prior to each experiment as needed for each method below. Human CTL line was generated as previously described [28] and maintained in RHFM media containing 90 U of interleukin 2 (Chiron) per ml. Treatment of Jurkat cells with CTL generated an allo-response. African “Buffalo Green” Monkey Kidney (BGMK) cells and Vaccinia Virus (VV) were obtained from ATCC and cultured as previously described [29]. Replication-deficient adenovirus (AD) (type 5) was a generous gift from Jack Gauldie, McMaster University. The GrB inhibitor L-038587-00Y001 (GrB inh) was a kind gift from Dr. Nancy A. Thornberry (Willoughby et al. , 2002). The inhibitor was incubated with GrB for 30 min at a concentration of 10 µM at 37°C prior to treating cells with GrB/AD. A non-replicating strain of AD was used as previously described [30] to internalize GrB into cells. AD was used at 1 plaque forming unit (PFU) /cell in all experiments involving purified GrB-mediated killing of cells. GrB activity was measured in a colorimetric assay [31]. Substrate degradation of Ac-IEPD-pNA (Kaiya Biomedical, Seattle WA) was monitored at A405. GrB-mediated hydrolysis of Ac-IEPD-pNA was inhibited in the presence of purified serpina3n (0. 6 to 0. 0001 mg/mL) or GrB inh. In these experiments, 2. 5×106 Jurkat cells were infected with VV at a multiplicity of infection (MOI) of 10. Unless otherwise specified, GrB concentration was 1 µg/ml (∼30 nM). AD concentration was always the same at 10 PFU/cell. All treatments that contained GrB also contained AD and zVAD-fmk (50 µM, 30 min prior to GrB/AD/VV). Following infection with VV (6 hr), cells were treated with GrB/AD for 1 hr, washed 3 times in PBS and then lysed. To estimate the number of VV present in the cell lysates, plaque assays were performed. BGMK were grown in monolayers and infected with the cell lysates as previously described [29]. Jurkat cells were co-treated for 30 min at 37°C with GrB at 1 µg/ml and AD at 10 PFU/cell. Following cell fractionation, cytosolic fractions were analyzed by 2-Dimensional electrophoresis and silver staining. This technique was performed as previously described with some modifications [32]. Cells were centrifuged for 1 min at 18,000 g and the pellet was resuspended in total lysis buffer and vortexed for 1 min. Proteins were purified and desalted using the ReadyPrep 2-D Cleanup kit (Bio-Rad, ON, Canada) and resuspended in 120 µl of rehydration buffer as provided by Bio-Rad. Samples containing 120 µg of protein were electrofocused on Immobiline DryStrip pH 3–17,7 cm IPG strips (GE Healthcare, NJ, USA) using the IPGphro isoelectric focuser (Pharmacia Biotech). Strips were equilibrated in ReadyPrep Equilibration buffers I and II (Bio-Rad) for 30 min in each buffer. Second dimension was run in Ready Gels 8–16% Tris-HCl, IPG COMB (Bio-Rad). Following electrophoresis, gels were fixed and silver stained using the silver staining plus kit (Bio-Rad). Mass spectrometry analysis was conducted as previously described [33]. Jurkat cell extracts were also tested by treating with inactive and active GrB, and ran in 1-Dimensional gels. Gel slices were analyzed by mass spectrometry at the Manitoba Centre for Proteomics and Systems Biology. GrB and CTL-mediated cytolytic activity was measured using methods that have been previously described [14], . 3×105 cells/ml of Jurkat cells were labeled with 40 µCi 3H-thyamidine for 20 hr prior to the experiment. Cells were treated with GrB at 1 µg/ml and AD at 10 PFU/cell. Human CTL-mediated killing was conducted as above with the following modifications. Jurkat cells were set at 1×106 cell/ml from 16 ml of total cell suspension. Cells were treated with a human CTL line [28] at an effector to target ratio of 5 to 1. Jurkat cells were killed by either GrB treatment or CD8-mediated killing, then kept in the incubator for the duration of the experiment. At each time point, triplicates of 200 µl from each tube were collected. A total of 20 time points were collected, once every 15 min from 0 min to 300 min. Immediately after collection, 200 µl/tube of lysis buffer (1% Triton x100 in TE buffer) was added and vortexed for 1 min. Cells were then centrifuged for 10 min at 18,000 g in a Microfuge 22R (Beckman Coulter) at 4°C. From the supernatant, 200 µl was collected and mixed with 5 ml of scintillation fluid. Each individual tube was counted for 2 min. Total 3H (100% Detectable DNA) was measured by collecting five-50 µl (quintuplicates) samples from each tube above. To each of these five tubes, 400 µl of total lysis buffer (2% SDS) was added and vortexed for 1 min. 5 ml of scintillation fluid was added and samples were counted for 2 min each. Cytolytic activity was measured by calculating the percentage of specific [3H]-thymidine release as follows: 100 x [ (sample cmp – spontaneous cmp) / (total cpm – spontaneous cpm) ]. The pan-caspase inhibitor Z-Val-Ala-Asp (OMe) -CH2F (zVAD-fmk) was used to assess the role of caspases (Kamiya Biomedical Company, WA, USA). In these experiments, cells were incubated with 50 µM zVAD-fmk for 30 min prior to GrB treatment. Inversion of phosphatidylserine on apoptotic plasma membrane was determined by FITC-annexin V binding and flow cytometry (Kamiya) [35]. Jurkat cells were pulse-labeled with 33 µCi [35S]methionine (PerkinElmer) for 1 hour at a density of 3×105 cells/ml. Following treatment, cells were centrifuged for 1 min at 18,000 g. Proteins were recovered through acetone purification using the ReadyPrep 2-D Cleanup kit. To each protein pellet, we added 400 µl of total lysis buffer (2% SDS) and vortexed for 1 min. 5 ml of scintillation fluid was added and samples were counted for 1 min each. Protein concentration was determined using the BCA assay (Pierce). The amount of total translated protein was estimated through the incorporation of MetS35 per mg of total protein (% Translated Protein). The highest count per minute (CPM) per mg of protein for each experiment was considered 100% translated protein. % Translated Protein was calculated from this value and plotted over time. 10% acrylamide gels were used. Electrophoresis was run in a mini protean 3 cell (Bio-Rad) connected to a Power Pac 1000 (Bio-Rad). Quantitative PCR (qPCR): This assay was performed as previously described [36]. β-actin primers were: forward 5′-CGA GAC CAC CTT CAA CTC CAT C-3′ and reverse 5′-GCG GTG GAT GGA GGG-3′. The TNT quick-coupled transcription/translation kit was used (Promega, WI, USA). This kit is a rabbit transcription/translation system. The translational machinery is thus Rabbit and the transcriptional machinery is added as a supplement to the rabbit reticulocyte lysate and is of bacteriophage origin. Master mix aliquots were thawed by hand warming and placed on ice. Individual reactions were prepared by combining 40 µl of master mix with 0. 02 µg/µl of template DNA, 0. 04 mM [35S] Methionine (MetS35) and water up to a total of 50 µl. Tubes were incubated at 30°C for 90 min. To perform IVT, the above method was modified by substituting the template DNA with purified 10 µg/µl mRNA, in the presence of 0. 02 U/µl of DNase. To control for translation, the reaction was arrested by adding 0. 01 U/µl of RNase. Radiolabeled proteins were used for in vitro proteolysis studies, or further purified by affinity chromatography. Visualization was performed by SDS-PAGE followed by either silver staining or autoradiography of the dried gels. IVTT or IVT reaction products were put on ice immediately after synthesis. GrB was diluted in PBS to yield a working concentration of 30 nM unless otherwise specified. Proteolytic degradation by GrB was carried out in a water bath for 30 min at 37°C and terminated by adding 6 µl of sample solubilizing buffer (SSB) (0. 5 M Tris-HCl, 16% glycerol, 3. 2% SDS, 0. 8% β-mercaptoethanol, 10 µg/ml bromophenol blue). The resulting proteolytic products were visualized by silver staining and autoradiography (Biomax MR film, Kodak, NY, USA) following SDS-PAGE. Point mutations were carried out using the Quikchange site-directed mutagenesis kit (Stratagene, CA, USA). The vector pcDNA3 containing eIF4G3 was provided by the lab of Dr. Nahum Sonenberg (McGill University, Canada). Oligonucleotide primers (Forward 5′-GTT GGA CTT CAT AGA GTC TGC CAG TCC CTG TTC CTC TGA AG-3′ and Reverse 5′-CTT CAG AGG AAC AGG GAC TGG CAG ACT CTA TGA AGT CCA AC-3′) were designed to point-mutate aspartate1408 (D1408) to alanine. Mutated plasmids were transformed into competent DH5α cells (Invitrogen, ON, Canada). Mutations were confirmed by sequencing the inserts. Successful colonies were purified using the EndoFree kit (Qiagen, ON, Canada). This technique was carried out as previously described with the following modifications [37]. Jurkat cells were mock infected or infected with VV at a MOI of 10 for 0 to 18 hr. 1 hr prior to termination of infection, Jurkat cells were treated with GrB at 1 µg/ml and AD at 10 PFU/cell. Cells were then starved for 30 min and pulse-labeled with 33 µCi [35S]methionine and [35S]cysteine (PerkinElmer) for 1 hour. Samples were analyzed by SDS-PAGE followed by autoradiography. Human Jurkat cells were transfected with pcDNA3-EIF4G3 and pcDNA3-EIF4G3 asp/ala mutant using the Amaxa® cell line Nucleofector® (Lonza, Cologne Germany). Briefly, 1. 5×106 cells were transfected with 2 µg Pvu I linearized plasmids using the Amaxa® Kit V and the program X-05. The transfected cells were then transferred to 1 ml pre-warmed RHFM media in a 12 well plate. After a 24 hr incubation period in a humidified 37°C/5% CO2 incubator stably transfected cells were selected using 1 mg/ml G418 (Gibco BRL Life Technologies inc.) and cloned by the limiting dilution method. Four mutant clones and 12 wild type clones were selected and maintained in 0. 8 mg/ml G418. Two clones from this group (mutant #4 and wild type #12) were chosen for further analysis. To confirm the integration of the plasmid DNA into the host chromosomes, genomic DNA was extracted from the two clones and a PCR fragment was generated using the 5′ eIF4G3 primer and the 3′ bovine growth hormone primer. The resulting PCR fragment was purified and sequenced. This technique was conducted using the ProFound Mammalian HA-Tag IP/Co-IP kit (Pierce, IL, USA). HA-Tagged eIF4G3 and the mutagenized clone eIF4G3Δ were generated through IVTT in the presence of MetS35. Following a total of 12 independent IVTT reactions for each clone, volumes were combined and loaded into the column. A total of 18 µl of anti-HA agarose slurry (30 µg anti-HA antibody) was then added to the sample and incubated for 16 hr at 4°C. Columns were then washed 3 times and the HA-tagged proteins were eluted into a 50 µl final volume. Protein concentrations were assessed by BCA assay (Pierce). Purification efficiency and protein identity was assessed respectively by autoradiography and Western blot analysis. Samples were subjected to 10% SDS-PAGE and proteins were identified using 1 µg/ml polyclonal anti-eIF4G3 antibody (Abcam, MA, USA). Blots or films were scanned at 500 dots per inch in the Agfg Arcus II scanner using Adobe Photoshop. The scanned images were saved as grey. tif files. Densitometric analysis was performed with the SigmaGel software. For each reading, a measurement was taken of the background and subtracted from the reading generated from the bands. For each film or blot, measurements were performed 3 times and the average was calculated. A number in arbitrary units was used to calculate the percent of expression. Results are means ± S. E. of at least three independent experiments with the exception of mass spectrometry analysis. Data were analyzed using one-way analysis of variance (ANOVA), and when significant differences were found, the multiple comparison Tukey-Kramer test was used (GraphPad InStat). Values of P<0. 05 were considered statistically significant. mRNA eIF4G3: NM_001198801 Protein eIF4G3: NP_001185730 mRNA eIF4G1: NM_182917 Protein eIF4G1: NP_886553 mRNA Granzyme B: NM_004131 Protein GrB: NP_004122 mRNA Bid: NM_197966 Protein Bid: NP_932070 mRNA Granzyme A: NM_006144 Protein Granzyme A: NP_006135 mRNA Granzyme K: NM_002104 Protein Granzyme K: NP_002095 mRNA Caspase-3: NM_004346 Protein Caspase-3: NP_004337
To screen for novel GrB substrates, Jurkat cells were treated with GrB and then monitored for the resulting proteomic change. Proteomic analysis was conducted through 1D- and 2D-electrophoresis analysis (data not shown) and mass spectrometry. eIF4G3 was identified as a candidate using this approach as a spot that consistently disappear following GrB treatment. Through sequence analysis of eIF4G3, we found a sequence very similar to those cleaved by GrB in the known substrates Bid and caspase-3 (Table S1). Western blot analysis of eIF4G3 expression in Jurkat cells post GrB/AD treatment showed a significant reduction of detectable eIF4G3 by 15 min post treatment (Figure 1). This reduction of detectable eIF4G3 coincided with the appearance of a cleaved product. The observed degradation of eIF4G3 was not mediated by caspases since the caspase inhibitor zVAD-fmk did not prevent degradation of eIF4G3 (Figure 1A). EIF4G3 degradation was significant by 20 min post GrB/Ad treatment in the absence of caspase activity (Figure 1B left panel). To confirm that GrB activity was present, we monitored activation of Bid and Casp-3, two GrB substrates, during our time course analysis of eIF4G3 degradation. We found that Bid was activated at 60 min, as demonstrated by the presence of p15 and p13 fragments (Figure 1A). Active Casp-3 was also detected at 60 min post GrB/AD treatment (Figure 1A). Conversely, the lack of proteolytic cleavage of Casp-3 in the presence of zVAD-fmk indicated that zVAD-fmk inhibited the upstream caspases responsible for cleavage of Casp-3. The location of the putative GrB cleavage site (P1 residue) of eIF4G3 is at aspartate 1408 (D1408); cleavage at this site would result in 2 fragments of approximately 180 kDa and 20 kDa in size. To test whether eIF4G3 is a substrate for GrB, the factor was synthesized using an in vitro transcription/translation system (IVTT) in the presence of radioactively labeled methionine (MetS35). The radioactively labeled eIF4G3 was then treated with GrB in vitro and analyzed by electrophoresis and subsequent autoradiography of the gels (Figure 2). Treatment of labeled eIF4G3 with 30 nM GrB resulted in the generation of a proteolytic fragment of eIF4G3, corresponding to the expected large fragment (Figure 2A). The small fragment expected from the cleavage of eIF4G3 was not detected in autoradiograms; this fragment was likely below the detectable limit of autoradiography as it contains only 3 radioactively labeled MetS35 residues per polypeptide, as opposed to 30 MetS35 residues present in the large proteolytic fragment. These results suggest that GrB cleaves eIF4G3 at D1408. Factor eIF4G3 is similar to eIF4G1 and may share overlapping roles in translation. These two proteins share 46% sequence homology, but eIF4G1 lacks the D1408 residue. To test if eIF4G1 is also cleaved by GrB, eIF4G1 was treated with GrB under the same experimental conditions. Treatment of eIF4G1 with 30 nM GrB did not result in proteolytic cleavage of this isoform (Figure 2B). D1408 is predicted to be the essential residue for cleavage by GrB. To confirm this, we mutagenized D1408 to alanine and tested the effect of this mutation on the proteolytic cleavage of eIF4G3 by GrB. This mutant (eIF4G3Δ) was resistant to 30 nM GrB treatment (Figure 2C). At a higher concentration of GrB (300 nM), eIF4G3Δ was proteolytically cleaved by GrB. However, this reaction generated different degradation products suggesting cleavage at alternative aspartate residues (Figure 2C). Our results showing that eIF4G3 is a substrate of GrB suggests that GrB can interfere with the translation machinery of target cells. We used the IVTT system to test whether GrB can inhibit protein synthesis. The IVTT system uses a master mix that contains transcription components from bacteriophage and translational components from rabbit reticulocyte lysate. We used a luciferase (Luc) DNA template for transcription and subsequent translation in the presence of MetS35. Luc protein synthesis was monitored by SDS-PAGE followed by autoradiography. Next, the in vitro synthesis of Luc was conducted in the presence or absence of GrB. GrB was added to the IVTT master mix prior to adding Luc DNA. GrB treatment reduced synthesis of Luc protein in a concentration-dependent manner (Figure 3A). At a concentration of 30 nM of GrB, Luc protein synthesis was no longer detectable with this technique. Adding GrB to the reaction following in vitro synthesis of Luc did not degrade Luc protein (Figure 3B), indicating that the changes observed in Figure 3A were a result of changes in expression of Luc and not GrB-dependent degradation of Luc. In the IVTT system, GrB could be targeting proteins involved in transcription, translation, or both. In order to determine if GrB affects mammalian translation independent of transcription, we used purified Luc mRNA and performed in vitro synthesis of Luc protein from mRNA. This approach converted the IVTT system into an in vitro translation (IVT) system. To ensure that the in vitro synthesis of Luc protein was restricted to translation, DNase was included in all IVT reactions to eliminate any possible DNA contaminants. Subsequently, when we added RNase, the synthesis of Luc protein was completely prevented, demonstrating that detectable Luc is a result of mRNA translation under our IVT system (Figure S1). In the presence of 30 nM GrB, IVT was prevented (Figure 3C). These results demonstrated that GrB was blocking in vitro synthesis of Luc by interfering at the translational level. To test whether other proteases had the same effects on translation, we tested GrB, GrA, GrK, Casp-3, Casp-7 and Casp-8 at concentrations of 3 and 30 nM in the same system (Figure 3D). GrB was the only enzyme able to completely prevent IVT of Luc under these conditions. Luminescence from these reactions indicated that 30 nM GrB, Casp-3 and Casp-7 resulted in a significant decrease of Luc activity, whereas GrA, GrK and Casp-8 showed no significant alteration of Luc activity (Figure 3E). To determine whether GrB cleavage of eIF4G3 is responsible for the GrB-mediated translational inhibition of Luc, we tested whether supplementing the IVT reaction with the GrB-resistant mutant eIF4G3Δ would regain translation of Luc. Mutant eIF4G3Δ and wild-type eIF4G3 were HA-tagged and purified using affinity chromatography (Fig S2). Purified GrB-sensitive eIF4G3 and GrB-resistant eIF4G3Δ were then added to the IVT of Luc, in the presence of GrB. Analysis of Luc protein synthesis and luminescence showed that recovery of translation was observed with eIF4G3Δ (Figure 4A), but not the wild-type protein (Figure 4B). Adding 1 and 2 µg/ml of eIF4G3Δ-HA to the GrB-treated IVT led to 22% and 23% recovery of Luc luminescence, respectively, when compared to the luminescence in untreated Luc IVT (Figure 4C). The ability of GrB-resistant eIF4G3Δ to rescue GrB-damaged translation machinery indicates that cleavage of eIF4G3 by GrB is a key event in translational inhibition. Other substrates may be involved but eIF4G3 is important. To exclude the possibility that eIF4G3Δ rescues translation by inhibiting the activity of GrB, we performed a GrB enzymatic assay in the presence of 0. 1 – 2 µg/ml of eIF4G3Δ. Our positive controls for the assay included GrB inh and serpina 3n, both of which significantly reduced GrB activity (Figure 4D). However, eIF4G3Δ did not affect GrB activity at all concentrations tested (Figure 4D). It is well established that GrB-induced DNA fragmentation, a hallmark of apoptosis, in Jurkat cells begins at around 30–60 min and reaches a plateau at around 120 min [38]. We confirmed this observation by measuring %[3H]-thymidine release in Jurkat cells in response to GrB (Figure 5A). Next, a time course analysis of GrB-induced translational inhibition in relation to the timing of DNA fragmentation was performed. The incorporation of MetS35 into newly synthesized proteins at 15 min intervals for 300 min was measured. A significant reduction in MetS35-labeled proteins was detected at 45 min post-GrB treatment, whereas significant DNA fragmentation was detected at 60 min post-GrB treatment. The lowered translational rate did not recover after 45 min (Figure 5A). These results show that translational arrest precedes the onset of apoptosis as measured by DNA fragmentation, which was consistent in all 4 independent experiments. GrB activates several caspases including caspase-3,7, 8,9 and 10. To determine if the observed reduction in the translational rate was mediated by the activation of caspases, we used the pan-caspase inhibitor zVAD-fmk. As expected, treatment of Jurkat cells with zVAD-fmk prevented GrB-mediated DNA fragmentation for the entire 300 min (Figure 5B). The addition of zVAD-fmk delayed the onset of translational inhibition by GrB to 75 min (Figure 5B) in comparison to 45 min (Figure 5A) in the absence of zVAD-fmk. However, translational inhibition by GrB was maintained even in the presence of zVAD-fmk as evidenced by the plateau in translation after 75 min (Figure 5B). These findings suggest that GrB-induced arrest in translation can occur independently of caspase activity, but that there may be cooperation between caspase-dependent and caspase-independent pathways. These results agree with our findings that GrB-mediated degradation of eIF4G3 is delayed in the presence of zVAD-fmk (Figure 1A). To confirm our findings using a more physiological system, we tested translational changes using cell-mediated killing in an allo-response [39]. CTL granules contain a complex proteome that includes an abundance of GrB. Jurkat cells were treated for 180 min with a human CTL line and assessed for MetS35 incorporation at 15 min intervals. As expected, CTL-induced apoptosis in Jurkat cells was prevented in the presence of zVAD-fmk (Figure 5C). This effect was also observed in Annexin V binding experiments (Figure S4B). Independent of zVAD-fmk treatment, CTLs induced a significant reduction of translation by 45 min in Jurkat cells (Figure 5C). These results indicated that CTL-mediated reduction of translation in target cells takes place independently of caspases. In the presence of the specific GrB inhibitor (L-038587-00Y001), inhibition of CTL-mediated translational was significantly decreased by 30 min post CTL treatment (Figure 5D). To confirm that GrB does not affect transcription, we performed qPCR of β-actin in Jurkat cells in the presence and absence of GrB/AD. Addition of GrB had no significant effect on mRNA transcription of β-actin in Jurkat cells; the well-known transcriptional inhibitor actinomycin D significantly reduced transcription in this system (Figure 5D). To measure the effect of CTLs on Jurkat eIF4G3 levels, eIF4G3 protein levels were measured by Western blot. When CTLs were added to Jurkat cells, a time-dependent reduction in eIF4G3 levels was observed (Figure S3A). After 10 min treatment, the eIF4G3 signal was not significantly different from CTL alone (Figure S3B). This suggests that CTL treatment results in degradation of eIF4G3 in Jurkat cells. Furthermore, pre-treating Jurkat cells with zVAD-fmk did not have any significant effect on detected levels of eIF4G3, indicating that eIF4G3 degradation is caspase-independent (Figure S3B). Thus far, the experiments looked at global translational inhibition by GrB. However, the major point of our model is that viral protein synthesis will be blocked early to halt the generation of infectious virus. To test this hypothesis, we infected Jurkat cells with VV and tested viral protein synthesis in the presence of GrB. During an infection, VV will hijack the host translational machinery for viral production. Thus, a successful VV infection is confirmed when only VV proteins are detected. We monitored protein expression by metabolic labeling. By 12 hr of infection, there was a near absence of host proteins and detection of late VV proteins - precursor 4a (P4a), processed structural proteins 4a and 4b and the early protein E (Figure 6A upper panel). To test whether late VV protein expression is affected by GrB treatment, we infected Jurkat cells for 12 hr, treated with GrB for 1 hr, and then performed metabolic labeling analysis. Mock infected cells and VV infected cells were run to compare host versus VV protein expression (Figure 6B upper panel). GrB treatment reduced VV protein expression, which was not affected by the caspase inhibitor zVAD-fmk (Figure 6B). Inactivation of GrB with a specific inhibitor (L-038587-00Y001) or heat inactivation prior to treatment reversed VV protein expression, while AD treatment alone did not affect VV protein expression (Figure 6B). To test whether VV protein expression was a direct result of GrB-dependent degradation of eIF4G3, we transfected Jurkat cells and generated stable clones of wild-type eIF4G3 (JurkateIF4G3-WT) and mutagenized GrB-resistant eIF4G3Δ (JurkateIF4G3Δ). GrB/AD treatment following 12 hr of VV infection prevented the formation of detectable VV proteins in Jurkat cells and in the JurkateIF4G3-WT clone. However, expression of the GrB-resistant eIF4G3Δ reversed the GrB-mediated protection, resulting in formation of late VV proteins (Figure 6C). These results provide strong mechanistic evidence supporting a model where GrB-mediated degradation of eIF4G3 prevents VV protein expression. In the absence of GrB, eIF4G3 protein expression did not change significantly from 0 to 18 hr post infection (Figure 6A lower panel), while GrB-dependent downregulation of viral protein expression correlated with a decrease in eIF4G3 levels (Figure 6B–C lower panel). This confirmed that cleavage of eIF4G3 coincides with VV protein expression. To study the effect of GrB on the formation of VV particles, we used a plaque assay to measure VV titres from infected Jurkat cells. In our optimization time course experiments, we found that significant VV production was detected starting at the 6 hr time point (data not shown). Thus, Jurkat cells were infected with VV for 6 hr, then treated for 1 hr with increasing concentrations of GrB in the presence of AD (10 PFU/cell) and zVAD-fmk. Following treatment, cells were lysed and the cytosolic fraction was used to infect BGMK cells in a plaque assay. GrB significantly reduced the production of VV particles from Jurkat cells at 2 and 4 µg/ml (Figure 6D). The metabolic labeling results shown in Figure 6A–C only detected VV proteins synthesized after GrB treatment and subsequent addition of [35S]methionine and [35S]cysteine. Thus, only newly synthesized (from metabolic labeling and onwards) VV proteins, not total VV proteins, are detected in Figure 6A–C. The plaque assay used in Figure 6D–E is more sensitive and relevant to our model than the metabolic labeling of VV proteins. However, this assay cannot distinguish between VV proteins present before and after GrB treatment. This is evident in our time course analysis of VV protein synthesis, where we detect VV proteins at 6 hr post-infection (Figure 6A). By the time GrB is added at 6 hr post-infection as shown in Figure 6D–E, these VV proteins have already been made. Although GrB treatment quickly halts VV protein synthesis, the already existing VV proteins are not degraded by the addition of GrB and are still capable of generating new VV particles. This fundamental difference between the two assays led us to identify that 2–4 µg/ml of GrB was needed to induce the expected GrB-mediated reduction in VV titers (Figure 6D). Expression of GrB-resistant eIF4G3Δ resulted in significant recovery of GrB-dependent translational halt of host protein synthesis by 2 hr (Figure S4C). Similar results were obtained in CTL-dependent translational halt (data not shown). The critical experiment in this study was to test the production of VV from JurkateIF4G3Δ cells. We predicted that GrB would be unable to degrade eIF4G3Δ in these cells, thus allowing VV replication in the presence of GrB. Expression of GrB-resistant eIF4G3Δ reversed the effects of GrB on VV infection, as the production of VV particles in JurkateIF4G3Δ was not affected by GrB treatment (Figure 6E). These experiments indicate that eIF4G3 is a critical GrB substrate that is targeted to halt production of infectious VV.
When a cell is infected with a virus, it becomes a factory for production of new infectious virus. Immune destruction of the infected cell ultimately puts a halt to this, but during the killing process, large amounts of virus could still be produced. It seems logical to propose that a successful strategy to eradicate virus infection should also bring a dramatic halt to virus production. The function of GrB to proteolytically activate pro-apoptotic factors like caspase-3 and Bid has been extensively investigated [40]. In this study, we describe a novel role of GrB to block protein synthesis in a VV-infected cell independently of GrB' s role in triggering apoptosis. This strategy transforms the host into a deficient system for viral replication, thus preventing viral spread. Importantly, we definitively showed that eIF4G3 is the critical substrate of GrB affected in protein translation. Proteomic approaches led us to eIF4G3 as a putative GrB substrate. Within the C-terminal domain of eIF4G3, there is a sequence similar to that present in Bid and caspase-3, where GrB cleaves these two substrates (Table S1). The only differing amino acid in eIF4G3 is the P2 residue (S1407); however, this serine is found at the same P2 location in other reported GrB substrates like centromeric protein b (CENP-B) (Table S1). eIF4G3 was cleaved in vitro by GrB and in Jurkat cells treated with either GrB or human CTLs. When we mutagenized the critical aspartate residue in eIF4G3 to alanine, the resulting mutant (eIF4G3Δ) was GrB-resistant. This evidence demonstrated that D1408 is the location at which GrB cleaves eIF4G3. In order to test whether GrB inhibited protein synthesis, we used an IVTT system programed by a luciferase gene. When the IVTT master mix was pretreated with 30 nM GrB, we were unable to detect protein synthesis or Luc activity. Since this could occur as a result of disruption of transcription or translation, the IVTT system was modified to specifically look at translation. This was achieved by first performing a regular IVTT of Luc. Then, Luc-mRNA was purified and used as a starting material in a new synthesis reaction. An important experimental parameter was to include DNase in the subsequent IVT reactions, to prevent transcription from any possible DNA that could be present in the master mix as provided by the manufacturer. This approach was used to determine that GrB was targeting and disrupting translation. Following treatment with GrB, the disabled protein synthesis system was restored by supplementing the system with eIF4G3Δ, but not with eIF4G3, demonstrating that cleavage of eIF4G3 by GrB is a critical event in translation inhibition. Recovery of translation was only partial, thus it is possible that GrB treatment directly or indirectly targets additional proteins involved in translation. The capacity of GrB to inhibit in vitro protein synthesis of Luc was confirmed in Jurkat cells. Within 45 min post-GrB treatment, translation was completely stopped with no significant recovery for up to 12 hr (Figure 5 and Figure S4C). When we treated VV-infected Jurkat cells with GrB, there was a significant reduction in the synthesis of viral proteins and new infectious VV particles. Stable transfection of Jurkat cells with the GrB-resistant eIF4G3Δ generated clones that were significantly more susceptible to VV infection in the presence of GrB. This experimental approach demonstrated that GrB blocks cap-dependent translation by specifically targeting eIF4G3 and cleaving this protein significantly reduces VV production. In eukaryotes, translational regulation is a critical method to control gene expression. With the exception of some genes, the majority of eukaryotic mRNAs are translated through cap-dependent translation. This process can be separated into the stages of initiation, elongation, termination and ribosomal recycling. Translational initiation is the stage where the majority of regulation takes place. Global regulation of translation initiation involves eIF inactivation [21]. This approach is used by cells during starvation or periods of stress and is employed during apoptosis to allow expression of key regulators. Although translation inhibition has been reported during apoptosis, we have shown for the first time that GrB alone can initiate translational shutoff and in doing so, prevent VV replication. In this study, we used Jurkat cells. These are commonly used as targets in CTL and granzyme assays since they die fairly quickly as revealed in a 4 hr chromium assay. This is in contrast to other target cells, particularly primary cells in which the time to induce cell death by GrB is much longer. We have estimated that one VV particle infecting a single cell can generate one hundred new viral particles in 24 hr. Thus, over this time frame, the inhibition of virus production will have significant effects. In addition, many viruses produce proteins that can inhibit apoptosis, thereby prolonging the generation of infectious virus. Although not the focus of this work, the inhibition of protein synthesis may also have implications for the control of apoptosis during viral infection. The roles of all three eIF4G isoforms are unclear. It is accepted that isoform eIF4G2 functions to inhibit translation, while isoforms eIF4G1 and eIF4G3 are involved in initiation of translation. Although eIF4G1 and eIF4G3 contain similar domains, they only share 46% of their sequence. Interestingly, when cap-dependent translation is inhibited by viruses, degradation of eIF4G3 is the rate limiting step and not eIF4G1 [23], highlighting the importance of eIF4G3 in the regulation of translation. Using a positional proteomic approach, it was reported recently that GrB can cleave eIF4G1 at an unusual recognition sequence - CGPD665F [41]. While it is possible that GrB can cleave both eIF4G1 and eIF4G3, we found that GrB was unable to cleave eIF4G1 in our IVTT proteolytic assays. Cleavage of eIF4G3 at D1408 results in a protein with a truncated C-terminal domain. We postulate that an undescribed function of this domain exists. The isoform eIF4G2 acts as general repressor of translation by inhibiting the functions of eIF4G1 and eIF4G3. Over-expression of eIF4G2 can inhibit both cap-dependent and IRES-mediated protein synthesis [42], [43]. It is possible that the truncated portion of eIF4G3 inhibits the function of eIF4G1 similar to eIF4G2, thus having a dominant negative effect on translation. In concordance with previous studies, we propose that the large proteolytic fragment of eIF4G3 might have translational inhibitory properties distinct from degradation of eIF4G3. Alternatively, cleavage may result in a conformational change of eIF4G3 folding that is responsible for the loss of its function. Our finding that GrB inactivates eIF4G3 through proteolytic cleavage with resulting decrease in protein translation agrees with the current model of eIF4G3 function. During infection by VV, GrB could also inhibit translation of specific viral proteins that limit host responses. For example, by blocking synthesis of CrmA, a serpin from VV that inhibits caspase activation, GrB would allow caspases to perform their apoptotic functions. Synthesis of eIF4G3 and eIF4G1 can occur through the use of internal IRES elements, independent of cap-mediated translation [44], [45]. In concordance with these results, we found that eIF4G3 downregulation was not detected during infection with VV, although total host protein synthesis was significantly reduced (Figure 6A). These results also agree with existing evidence that VV preserves the eIF4F complex [24] and that VV is unable to replicate in the absence of eIF4G3 (a subunit of the eIF4F complex) [23], [25]. A number of studies have reported caspase-3-dependent degradation of eIF4G1 in cells undergoing apoptosis [46], [47], [48]. In the current study, GrB did not degrade eIF4G1 in vitro; however, GrB could degrade both eIF4G3 and eIF4G1 in cells through direct degradation of eIF4G3 and indirectly through activation of caspase-3. While it has been shown that caspase-3 cleaves eIF4G3 [49], we found that inhibition of caspases by zVAD-fmk did not prevent GrB-dependent degradation of eIF4G3 or arrest of translation. However, zVAD-fmk changed the kinetics of eIF4G3 degradation (Figure 1A–B) and host translational arrest induced by GrB from 45 min with GrB alone to 75 min in the presence of zVAD-fmk. This suggests that there is cooperation between caspase-dependent and caspase-independent pathways mediated by GrB in the arrest of protein synthesis. We propose a model where GrB inhibits translation by 1) proteolytic cleavage of eIF4G3 and 2) indirect degradation of eIF4G1 through activation of caspase-3. Since the discovery of GrB, a significant amount of work has been carried out to decipher the role of GrB in CTL-mediated killing. We now recognize that GrB, directly and through the mitochondrial pathway, activates key substrates like caspase-3 and Bid to induce apoptosis [40]. In this study, we present findings that suggest a dramatic change of paradigm with regards to the roles of GrB. The ability of GrB to halt the host translational machinery is a novel strategy to prevent viral replication in a host cell targeted by CTLs. This mechanism of GrB has not been described for other granzymes within the granzyme family of proteases; specifically, GrA and GrK were unable to inhibit translation. Although GrH has been shown in literature to inhibit viral translation by targeting the RNA-binding protein La, they did not demonstrate an effect of GrH on host translational machinery [19]. Whether the effect of GrB on translation is due to proteolytic inactivation of eIF4G3 or that cleavage of eIF4G3 results in products with new suppressive functions on translation remains to be determined. Our findings unveil a new antiviral axis of GrB function to fight viral infection by targeting translation. | Lymphocytes, a type of white blood cell, are the major killer of virus-infected cells. Lymphocytes secrete proteins like granzyme B that are responsible for the destruction of the virus-infected host cell. However, killing an infected cell through this pathway may take several hours, thus allowing viral replication to occur while the cell is in the process of dying. In this study, we identified a new role of granzyme B in preventing viral replication during the killing process. We found that granzyme B disables the ability of the host cell to make new proteins, including viral proteins of infected cells. Thus, granzyme B is able to halt the production of new viruses by inhibiting protein production. | Abstract
Introduction
Materials and Methods
Results
Discussion | medicine
immune cells
viral transmission and infection
immunology
microbiology
gene expression
t cells
biology
jurkat cells
molecular biology
host cells
cell biology
clinical immunology
virology
protein translation
genetics
cellular types
molecular cell biology
genetics and genomics | 2011 | Granzyme B Inhibits Vaccinia Virus Production through Proteolytic Cleavage of Eukaryotic Initiation Factor 4 Gamma 3 | 13,274 | 168 |
Contact patterns strongly influence the dynamics of disease transmission in both human and non-human animal populations. Domestic dogs Canis familiaris are a social species and are a reservoir for several zoonotic infections, yet few studies have empirically determined contact patterns within dog populations. Using high-resolution proximity logging technology, we characterised the contact networks of free-ranging domestic dogs from two settlements (n = 108 dogs, covering >80% of the population in each settlement) in rural Chad. We used these data to simulate the transmission of an infection comparable to rabies and investigated the effects of including observed contact heterogeneities on epidemic outcomes. We found that dog contact networks displayed considerable heterogeneity, particularly in the duration of contacts and that the network had communities that were highly correlated with household membership. Simulations using observed contact networks had smaller epidemic sizes than those that assumed random mixing, demonstrating the unsuitability of homogenous mixing models in predicting epidemic outcomes. When contact heterogeneities were included in simulations, the network position of the individual initially infected had an important effect on epidemic outcomes. The risk of an epidemic occurring was best predicted by the initially infected individual’s ranked degree, while epidemic size was best predicted by the individual’s ranked eigenvector centrality. For dogs in one settlement, we found that ranked eigenvector centrality was correlated with range size. Our results demonstrate that observed heterogeneities in contacts are important for the prediction of epidemiological outcomes in free-ranging domestic dogs. We show that individuals presenting a higher risk for disease transmission can be identified by their network position and provide evidence that observable traits hold potential for informing targeted disease management strategies.
Heterogeneity in contact rates is influential in the epidemiology of both human and non-human animal diseases. In principle, variation in the contact rates among individuals affects their risk of acquiring and/or transmitting infections [1,2]. Relationships between host social behaviour and the distribution of infections have been demonstrated in several wild animal host-pathogen systems, from tuberculosis in badgers Meles meles [3] and meerkats Suricata suricatta [4] to nematodes in Japanese macaques Macaca fuscata [5]. One driver of these relationships is the variation in contacts between individuals, which can influence the flow of infection through populations [6,7]. Therefore, in order to successfully manage some diseases, it is important to understand the dynamics of host contacts that facilitate the transmission of infection [8]. The number of infectious disease emergence events in humans has been increasing over time, and the majority of these are zoonotic in origin [9]. This may, in part, be associated with the domestication of animals, as evidence suggests that the number of shared pathogens (between humans and non-human animals) increases with the time since a species was domesticated [10]. This is because domestication increases the exposure of people and animals to a greater range of pathogens, and increases the risk of humans acquiring zoonotic infections [11]. If domestic animals are free-ranging, they are also more likely to interact with wild animals, further facilitating the transmission of disease between people and wildlife [12]. Dogs Canis familiaris are among the earliest domesticated animals and they share 16% of their known pathogens with humans [10] and 47% with wild mammals [13]. Amongst these pathogens is rabies, a viral zoonosis that poses a significant public health risk, responsible for approximately 59,000 human deaths annually [14] and primarily transmitted to humans through the saliva of an infected dog when they are bitten [15,16]. Mathematical models can be applied to inform management efforts by predicting epidemics and, for rabies, these models are relatively well developed [17]. However, one of the challenges identified in controlling rabies is a lack of information on dog ecology [18] and variation in contact rates has been identified as being especially influential for epidemic outcomes in a number of modelling studies [19,20]. This is unsurprising given that dogs are social animals that exhibit pronounced between-individual variation in their behaviour [21]. Collecting high resolution data on the contact rates between individuals is a major challenge, particularly for free-ranging animals. This lack of empirical data has meant that stochastic models have relied on assumptions that contact rates are density dependent or have included a frequency dependent function in the form of spatial and/or social scaling parameters to generate variation in the probability of contacts [20,22]. Although these assumptions are biologically sound, they fail to capture social phenomena that could influence disease transmission, such as assortative mixing [23] and clustering [24]. Including observed contact data in stochastic models of communicable diseases could help better predict epidemics at a local scale and help identify novel management techniques [25]. To date, there has been only one study published that integrated observed contact rates of free-ranging dogs into a model for the transmission of rabies [26], in which the contact network of dogs was characterised over 3. 5 days in an urban environment. They found that urban dogs formed communities that were defined by roads, which acted as a barrier to movement. When simulating outbreaks of rabies, the authors observed that major epidemics were avoided when 70% of the population were vaccinated and targeted management based on network measures increased the effectiveness of vaccination. However, it is unclear if this would also apply to rural dog populations, where the landscape and dog-human relationships are likely to be different to that in an urban environment [27], where unowned dogs are rare, roads are few and where hunting, subsistence farming and fishing are more prevalent. In this study, we used automated proximity loggers to generate high-resolution contact networks of free-ranging dogs in an area of rural sub-Saharan Africa, where dogs are susceptible to a number of zoonotic infections. We use these data to model the transmission of an infection that is epidemiologically similar to rabies. We test the effect of including observed heterogeneities in contacts between free-ranging dogs on predictions for epidemic size. Using a network model we simulate epidemics through random networks, the observed network characterised as binomial (present/absent) interactions and the observed network when weighted by the duration of interactions. The observed binomial network introduces non-random structures while maintaining uniformity and the observed weighted network adds non-random and non-uniform mixing. In addition, we investigate the effect of seeding different individuals with the infection. If contact heterogeneity influences epidemics it may be possible to predict epidemic outcomes using the network position and/or associated traits of the seeded individual.
This study was approved by the University of Exeter College of Life and Environmental Sciences (Penryn Campus) Ethics Committee (Reference 2016/1488). Dogs were studied between June 24th and July 12th 2016 in two settlements, each comprising two neighbouring villages, located along the Chari River in the Guelendeng district of the Mayo-Kebbi East region of Chad. The settlement Kakale is located to the south-east of Guelendeng town and includes the villages Kakale-Mberi (10°53' 0. 79" N, 15°38' 8. 45" E) and Awine (10°48' 6. 34" N, 15°37' 56. 61" E). Kakale-Mberi is a linear settlement along a main (dirt) road that runs parallel to the Chari River. Awine is a dispersed settlement that is seasonally occupied by the people of Kakale-Mberi, who move there to cultivate crops. The settlement Magrao is comprised of the villages Magrao and Sawata (centred on 10°59' 44. 31" N, 15°29' 29. 27" E), located to the north of Guelendeng. Magrao is a dispersed village lying between the Chari River and the main road from Guelendeng to the capital, N’Djamena. Sawata is a smaller village that is surrounded by Magrao but is distinguished by different ethnicity and a higher prevalence of pastoralism. All dogs had clear ownership and were associated with a specific household. They were all sexually intact. With the consent of owners, dogs were collared with standard nylon dog collars (Ancol Heritage). Puppies (less than 6 months of age) were not collared. Collars were fitted with two devices; (1) an i-GotU GT-600 GPS unit (Mobile Action Technology Inc. , Taiwan) and (2) a wearable proximity sensor based on a design developed by OpenBeacon project (http: //www. openbeacon. org/) and the SocioPatterns collaboration consortium (http: //www. sociopatterns. org/). The GPS units were configured with a fix interval of 10 minutes and a sleep mode to extend battery life. The proximity sensors exchange one ultra-low power radio packet per second in a peer-to-peer fashion and, have been successfully deployed in several studies on humans [28,29]. The exchange of radio-packets is used as a proxy for the spatial proximity of individuals wearing the sensors [30,31]. Close proximity is measured by the attenuation, defined as the difference between the received and transmitted power. The attenuation threshold used in this deployment was selected to detect close-contact events (within 1–1. 5 m), during which a communicable disease infection might be directly transmitted, either by airborne transmission or by direct physical contact. Additional data collected on the individual dogs included sex and body condition score (BCS; [32]). Due to low frequencies of some scores, we categorised them into poor (BCS ≤ 2) and moderate (BCS ≥ 3). Interviews using a standardised questionnaire were carried out at households to record the number of dogs owned and the dogs’ ages, as recalled by the owner. A single observer estimated BCS and another conducted all household interviews. Dogs aged 12 months or less were classified as juveniles, dogs aged between 13 and 24 months were classed as sub adults and dogs older than 24 months were regarded as adults [33]. Since all households known to have dogs in the settlement were visited, the dog population size (excluding puppies) was calculated for each settlement by summing the reported number of owned dogs from each household. The proximity data were extracted from devices and cleaned by identifying corrupted sensors (where no data were available) or anomalous signals (such as continuous bursts of data). The GPS data were cleaned by removing erroneous fixes with speeds greater than 20 km/hr between locations. For both GPS and proximity data we discarded records collected on the first and last day of collar deployment in each village; providing time for the dogs to habituate to collars at the start and to account for the collection of the collars at the end of the field study. Data analysis was conducted in R v3. 3. 3 [34] and Python v2. 7. The R packages ‘sp’ v1. 2–3 and ‘rgdal’ v1. 2–5 were used to project the GPS data into the relevant coordinate reference system for Chad (EPSG: 32634). The package ‘adehabitatHR’ v0. 4. 14 was used to calculate the dog’s total range (99% minimum convex polygons) and core range (60% kernel density estimate). Networks were treated as undirected symmetric networks. Since dogs were not collared for the same number of days, the weights for the weighted networks were converted to the average number of seconds the dogs were in contact per day monitored. This was done by dividing the total duration in seconds over which a pair was in contact, by the shorter of the two periods in days for which the two dogs were collared. These weights were then log10 transformed. The global and local network metrics were calculated using the R package ‘igraph’ [35]. The network position of individuals was described using metrics most relevant to disease transmission [36], including: degree (the number of unique connections of an individual), strength (the summed strength of all connections for an individual), betweenness (the number of shortest paths between other individuals upon which the focal individual lies), and eigenvector centrality (a measure of second order contacts whereby a higher score is assigned to individuals if they associate with highly connected individuals or many moderately connected individuals). To compute the probability density distribution of contact durations and the complementary cumulative distribution function (CCDF) of edge weights, we used the Python package ‘Powerlaw’ v1. 4. 1. Community membership describes individuals that are closely associated/clustered together and these groups were identified using the edge betweenness and Greedy algorithms in the Python package ‘igraph’ v0. 7. 1. The package ‘Epimodel’ v1. 3. 0 [37] was used to build a Susceptible, Exposed, Infected and Removed (SEIR) network model of infection spread. Simulations were run on the observed binomial network, the observed weighted network and the null model (random networks). Random networks are traditionally used in network analysis to overcome the non-independence nature of contacts, and are typically constrained to biologically plausible scenarios. The null model for this study was that individuals mix randomly and so random networks were generated using the Erdős-Rényi model, conserving the observed number of nodes (individual dogs) and edges (connections). Every individual in the binomial and weighted networks was seeded with the infection and, for each seeded individual, 100 simulations of the model were run. For the null model, the same procedure was conducted, however, each simulation involved a different random network and all seeded individuals experienced the same set of 100 random networks. Simulations were run over 300 time steps (days). The network model assumed that (a) there was no recruitment or loss of individuals to the population (except the eventual removal of those infected), (b) the edges and weights of the network did not rewire over time or in response to infected or removed individuals and (c) individuals do not change their behaviour when infected. For each simulation an initial seed (infectious individual) was selected at time step 1. At time steps 2–300, an edge list of infectious and susceptible individuals was made and transmission events were determined through a random binomial draw using the calculated per link transmission probability (β): β=1- (1-λ) α (1) The probability of infection after being bitten (λ) was taken to be 0. 49 [38]. To our knowledge, no data are available on the act rate (α; number of bites per partnership per day) of rabid dogs and it was therefore taken to be: α=log (1-β) log (1-λ) (2) Where β was calculated by assuming a constant value of the basic reproductive number (R0) and by rearranging its definition in the heterogeneous mean-field approximation [39]: β=R0μ〈k〉〈k2〉-〈k〉 (3) The mean degree 〈k〉 and mean square degree 〈k2〉 were extracted from the observed networks (see Table 1). The infectious period (μ) was randomly drawn from a gamma distribution (shape = 3. 0; scale = 0. 9; see [38 & 40]). Simulations were run for a range of basic reproductive numbers found in the literature for rabies in dogs. The lower R0 was set to 1. 2, the mid value was 1. 8 [38] and the upper R0 was 2. 4 [41]. The transmission probability for different edge weights (βij) was calculated using Eq 4: βij=1- (1-λ) αij (4) αij=αwij1+wij×2 (5) The weighted act rate (αij) was calculated through Eq 5 which is modified from Reynolds et al [7]. Here we assumed that αij was positively associated with the daily average of the total duration that individuals were in contact (wij), and in so doing, we applied a sigmoidal scaling function. This value was then multiplied by two to shift the mean of βij to β. The use of this scaling function is justified where biting is the main method of rabies transmission and only a short contact time is required. Once a transmission event occurred, a random draw from a gamma distribution was used to allocate an incubation period (shape = 1. 1; scale = 20. 1; see [38 & 40]) and infectious period (see above for parameters). During the incubation period individuals were considered to be in the exposed category. Once the incubation period was over, the individual was classed as infected and could transmit the disease until such time as the assigned infectious period was over and the individual, along with its associated edges, was removed from the network. For this study, an epidemic was defined as disease transmission to at least one other individual. Differences in ranked network position (degree, strength, eigenvector centrality and betweenness) between nodal attributes (sex, age, BCS and home ranges) were identified by calculating t-statistics, using either t-tests or linear models. Observed statistics were compared to the distribution of test statistics from null models to identify if they were significantly different to those expected had individuals mixed randomly [42]. Null models consisted of 10,000 random networks generated by randomly shuffling the node attributes while keeping the structure of the observed network the same. Homophily within the attributes age, sex and household was investigated by calculating the assortativity (r) coefficient using the ‘assortnet’ package in R. Again the observed coefficients were compared to the distribution of coefficients from null models. To see if community membership was determined by the dogs’ sex, age or household, we used the Normalized Mutual Information (NMI) score to scale the results between 0 (no mutual information) and 1 (perfect correlation). To investigate if there was a correlation between edge existence/weight and the distance between households, the ‘sna’ package v. 2. 4 in R was used to perform a quadratic assignment procedure (QAP) with 1000 permutations. Generalised additive models (GAMs) were used to identify non-linear relationships between the averaged epidemic outcomes of simulations for seeded individuals and their ranked network position (degree, eigenvector centrality, and betweenness). Models were fit with family set to Gaussian and included a smoothing term (k = 3). Strength was not investigated in these models since no difference in epidemic size between weighted and binomial simulations was observed. Since measures of network position are often correlated, separate models were fitted for each measure of centrality and type of network. Akaike’s Information Criterion (AIC) and adjusted r2 values were extracted and used to identify which centrality measure best explained epidemic outcomes.
In Kakale, collars were successfully deployed for a mean of 8 days (range 2–9 days) on 48 dogs (86% of the population excluding puppies) from 28 different households (Fig 1). The distance between dog owning households ranged from 23–10,002 m. 8561 contact events were recorded between dogs in Kakale and the median contact duration was 20 seconds with a percentile (2. 5% - 97. 5%) range of 20–200 seconds. In Magrao, contact data were collected for a mean of 8 days (7–10 days) for 60 dogs (82% of the population) from 36 households. The distance between households ranged from 35–4758 m. 7361 contact events were recorded between dogs in Magrao and the median contact duration was 20 seconds, with a percentile range of 20–160 seconds. The global structure of both networks revealed high levels of clustering and short average path lengths (Table 1). Furthermore, community analysis using the edge betweenness (EB) and Greedy (G) algorithms showed the dog populations in both settlements exhibited high modularity in the binomial network (Kakale: EB = 0. 48, G = 0. 51; Magrao: EB = 0. 56, G = 0. 57) and the weighted network (Kakale: EB = 0. 57, G = 0. 603; Magrao: EB = 0. 60, G = 0. 617). Magrao was the larger of the two networks and had a wider degree distribution (kmin = 1, kmax = 17) than that of Kakale (kmin = 2, kmax = 14). In both networks the degree distribution was homogenous (Kakale: coefficient of variation (CV) = 0. 49, Magrao: CV = 0. 48) while the distributions for the duration of contacts were highly heterogeneous (Kakale: CV = 1. 88, Magrao: CV = 1. 85), and the probability density distribution declined as contact durations increased (Fig 2). Dogs in Magrao had substantially smaller ranges than dogs from Kakale, and the distribution of ranges was right skewed for both settlements (S1 Fig). Dogs in Kakale that had larger ranges had higher ranked eigenvector centralities and this was significantly different to null models (Table 2). Similarly, the home ranges of dogs in Kakale were positively correlated with their ranked degree, and this correlation was significantly greater than that of null models. In both networks, comparisons to null models revealed no significant association of any ranked network measures (degree, strength, eigenvector centrality or betweenness) with sex, age or body condition. All measures of community membership were strongly correlated with household membership in both the binomial networks (Kakale: NMIEB = 0. 622, NMIG = 0. 625; Magrao: NMIEB = 0. 739, NMIG = 0. 649) and weighted networks (Kakale: NMIEB = 0. 674, NMIG = 0. 70; Magrao: NMIEB = 0. 725, NMIG = 0. 713). Community membership had no significant relationship with either the dog’s sex or age (S1 Table). When compared to null models, dogs in both settlements had a strong preference to associate with individuals from the same household and no assortative mixing patterns were found between dogs of a different/similar age or sex (Table 3). QAP tests found a significant negative correlation for the distance between households and the existence of an edge (Kakale: r = -0. 23, p < 0. 001; Magrao: r = -0. 4, p < 0. 001). A negative correlation was also found for the relationship between household distance and edge weight (S2 Fig Kakale: r = -0. 22, p < 0. 001; Magrao: r = -0. 37, p < 0. 001). For both settlements, larger R0 values resulted in an increased risk of epidemics occurring and larger epidemic sizes (Fig 3, see S3 Fig for the frequency distributions of secondary cases). In simulations when R0 was 1. 8 or 2. 4, mean epidemic size was higher for random networks than that of simulations with observed contacts. Epidemic sizes for simulations using these random networks had a bimodal distribution, whereby epidemics either involved a large number of individuals or very few. In contrast, the distribution of epidemic sizes for observed networks had multiple peaks at intermediate sizes. The distributions of epidemic sizes differed for the two settlements, whereby Kakale had more intermediate peaks. Simulations with the lowest R0 value (1. 2) showed no discernible difference in mean epidemic sizes between the random and observed networks. For the observed networks for both settlements, the seeded individual’s ranked centrality measures (degree, eigenvector centrality and betweenness) were all positively correlated with the proportion of simulations that resulted in an epidemic (S4 Fig). The seeded individual’s ranked degree was the best predictor for the proportion of simulations to result in an epidemic (Table 4), and at larger R0 values the relationship between ranked degree and an epidemic outcome began to plateau for higher ranked individuals (Fig 4). As expected, the seeded individual’s observed centrality measures did not correlate with the proportion of simulations to result in an epidemic in any of the random networks. The seeded individual’s ranked eigenvector centrality and ranked degree were positively correlated with the mean epidemic size in simulations on the binomial and weighted networks for both settlements (S5 Fig). Ranked eigenvector centrality was the best predictor of mean epidemic size (Table 4), and for simulations of Magrao at larger R0 values, mean epidemic size plateaued for individuals with a higher ranked eigenvector centrality (Fig 4). The distributions of eigenvector centralities for dogs in each settlement (S6 Fig), were similar to the distribution of epidemic sizes in respective settlements. No correlation between the seeded individual’s network position and mean epidemic size was found in any of the random networks.
We have gathered high-resolution data on the contacts among free-ranging domestic dogs living in two rural settlements in Chad, an area where rabies infection is endemic and regularly causes human fatalities. Using these data we have demonstrated the importance of including observed contact patterns when simulating the transmission of an infection comparable to rabies. We show that the observed contact rates between dogs are heterogeneous and that interactions were dominated by contacts that were short in duration and between dogs from the same household. In our model, for the transmission of infection, the inclusion of observed contact rates resulted in fewer epidemics occurring compared to when random mixing was assumed and, for all but the lowest R0 values, epidemics were smaller in simulations using the observed networks. We also show that the seeded individual’s first and second order contacts were strong indicators of epidemic outcomes, verifying that individuals differ in the risk they present for the transmission of infections. Furthermore, for dogs in one settlement, second order contacts were correlated with ranging behaviour, suggesting that observable traits exist which could inform targeted management strategies. The transmission probabilities associated with the lowest R0 value rarely resulted in an epidemic and, when one occurred, no more than a few individuals were infected. This meant that there was little difference in the overall mean epidemic size between simulations of random and observed networks. However, heterogeneity in contacts was still important in determining epidemic outcomes whereby the seeded individual’s ranked degree was positively correlated with the proportion of simulations that resulted in an epidemic, and this was echoed in simulations with higher R0 values. This finding demonstrates that, regardless of the transmission probability, dogs that are in contact with more individuals relative to the rest of the population are at higher risk of causing an epidemic should they become infected. In simulations with all but the lowest R0 value, the risk of a large epidemic was higher when infection started in dogs with a higher ranked eigenvector centrality, and this was further emphasised where the distribution of eigenvector centralities paralleled that of epidemic sizes for each settlement. The importance of an individual’s eigenvector centrality in disease dynamics has also been shown in models for the transmission of Mycobacterium bovis in badgers [43] and observed parasite infection in Japanese macaques [5], where this measure was positively correlated with infection status. It appears that eigenvector centrality is a robust predictor of epidemic size and infection status because it describes how an individual is rooted into the network beyond their immediate connections. We show that ranging behaviour was correlated with eigenvector centrality, but this was only true for dogs in Kakale. Both range sizes and eigenvector centralities were higher for dogs in Kakale than those in Magrao. This is likely due to anthropogenic variation in dog behaviour whereby during the study some people in Kakale moved with their dogs back and forth between a permanent residence and a seasonally-occupied homestead, while people in Magrao tended to stay at one. The dogs that accompany their owners in travelling between permanent and seasonal homesteads will have larger ranges and this would influence the dog’s network position by creating new contact opportunities. Nevertheless, the relationships between dog network position and epidemic outcomes were the same in both settlements. We also show that the distribution of dog owning households is important in determining contacts between dogs, with dogs more likely to have been in contact with and having stronger connections with dogs from closer households. However, it is important to note that this distance effect cannot fully explain the structure of the contact networks as many dogs from households in close proximity did not come into contact (S2 Fig). Although the dogs in this study were free-ranging, they were owned and anthropogenic influences on dog contact rates and ranging behaviour should not be overlooked, and understanding these would provide insight into disease management approaches. For both settlements, there was no notable difference in epidemic size between simulations using the observed binomial and weighted networks. This result would suggest that including non-random mixing (whom individuals contact) in disease models is more important than including non-uniform mixing (contact duration/frequency). However, heterogeneities in edge weights are likely to be important and have been shown to further limit epidemic sizes when they are allowed to be dynamic in time [44]. To further understand the effect of non-uniform mixing, future research should try to describe the temporal dynamics of free-ranging dog contacts over a timeframe relevant to the disease in question. Specifically, investigations should look for daily and seasonal differences in network structure and identify whether or not individuals occupy stable network positions. The model of rabies transmission used in this study makes several assumptions that should be considered. First, individuals do not change their behaviour once infected. It is well known that rabies can manifest as either encephalitic (furious) or paralytic (dumb) and evidence suggests that, unless vaccinated, the furious form is more likely to develop in dogs [45]. However, it is not clear what determines the type of rabies an individual develops or if the different forms result in considerable deviations from the individuals’ typical behaviour. Such deviations could result in changes to the contact network with either new connections being formed, the loss of connections or changes in the strength of connections. A second assumption is that when individuals were removed due to death, the network structure did not change. Removing these assumptions would require a rewiring of the network and this process should be biologically informed. Reynolds et al [7] attempted to account for dumb and furious behaviours by assuming different frequencies of each and either changing the transmission probability (higher for furious and lower for dumb) or by altering the individual’s contact behaviour (removing half their connections for dumb or doubling them for furious). They found that both methods produced similar results and the speed of transmission increased when there was a higher frequency of furious individuals and decreased with a higher frequency of dumb individuals. Although this effort to model behavioural change can be insightful, the methods of rewiring are not biologically informed and so should be interpreted carefully as they cover a limited number of possible scenarios in which the network could change. Solutions to such network dynamics are challenging as there is a lack of experimental data on the processes of network rewiring and, without this guidance, the number of potential modes of change is too computationally demanding to include in models. For diseases such as rabies it is unlikely that such data will ever exist given the ethical implications of such experimentation. However, understanding how a network rewires as individual states or community membership change could better allow network models to include such dynamics that are thought to be a major obstacle for controlling rabies [38]. The inflation of predictions for epidemic size in models that do not account for observed contact heterogeneities are of particular concern when public health resources are limited [46]. This is the case for dog-mediated rabies in developing countries, where epidemics are preventable through vaccination but a major challenge is the high incidence of dog infections and human cases, combined with limited public health resources [18]. Currently it is advised that successful vaccination campaigns require 70% coverage of the dog population [47]. However, through targeted management this might be reduced, helping alleviate costs. Further to work on urban dogs [26], our results show that even in a rural context, epidemic risk is not equal among individuals and suggest that, by identifying the network position of individuals and correlates thereof, targeted management could be feasible. We find evidence to suggest that the spatial ranging behaviour of dogs was associated with their network position, though anthropogenic influences clearly have a role in determining free-ranging dog movements and this deserves further investigation. Our research illustrates how a greater understanding of the social contact network of free-ranging dogs can help better inform the management of diseases such as dog-mediated rabies. | For communicable infections, variations in contact rates determine the flow of disease through populations. Therefore, describing contact patterns within populations could help to better predict and prevent disease outbreaks. Free-ranging domestic dogs are susceptible to a number of zoonotic infections yet few studies have investigated their contact patterns. We describe high-resolution contact data for free-ranging dogs in rural Chad and simulate the transmission of an infection comparable to rabies. We show that epidemic outcomes are determined by the seeded individual’s network position, which was also correlated with ranging behaviour. This demonstrates that between-individual variation in the risks of spreading infection may be linked with observable traits that can help inform targeted management strategies. | Abstract
Introduction
Methods
Results
Discussion | animal types
medicine and health sciences
infectious disease epidemiology
binomials
domestic animals
tropical diseases
geographical locations
vertebrates
dogs
animals
mammals
rabies
mathematics
algebra
network analysis
neglected tropical diseases
zoology
polynomials
africa
infectious diseases
computer and information sciences
zoonoses
epidemiology
centrality
people and places
eukaryota
eigenvectors
linear algebra
biology and life sciences
physical sciences
viral diseases
amniotes
chad
organisms | 2019 | High-resolution contact networks of free-ranging domestic dogs Canis familiaris and implications for transmission of infection | 7,444 | 154 |
Antiretroviral therapy (ART) can reduce HIV levels in plasma to undetectable levels, but rather little is known about the effects of ART outside of the peripheral blood regarding persistent virus production in tissue reservoirs. Understanding the dynamics of ART-induced reductions in viral RNA (vRNA) levels throughout the body is important for the development of strategies to eradicate infectious HIV from patients. Essential to a successful eradication therapy is a component capable of killing persisting HIV infected cells during ART. Therefore, we determined the in vivo efficacy of a targeted cytotoxic therapy to kill infected cells that persist despite long-term ART. For this purpose, we first characterized the impact of ART on HIV RNA levels in multiple organs of bone marrow-liver-thymus (BLT) humanized mice and found that antiretroviral drug penetration and activity was sufficient to reduce, but not eliminate, HIV production in each tissue tested. For targeted cytotoxic killing of these persistent vRNA+ cells, we treated BLT mice undergoing ART with an HIV-specific immunotoxin. We found that compared to ART alone, this agent profoundly depleted productively infected cells systemically. These results offer proof-of-concept that targeted cytotoxic therapies can be effective components of HIV eradication strategies.
ART is a lifesaving and effective means to control HIV infection [1]. However, the persistent nature of this infection requires lifelong adherence to daily ART dosing [2]–[4]. This viral persistence, the cumulative costs of ART, adverse events associated with long-term ART and the constant threat of emergence of drug-resistant viral variants have led researchers to pursue HIV eradication therapies that will result in a viral rebound-free interruption of therapy [2]–[4]. Towards this goal, “kick and kill” HIV eradication strategies are being developed [5]. While interventions that can induce expression of latent HIV (e. g. histone deacetylase inhibitors and protein kinase activators) will function as the “kick” [2]–[4], it is important to note that “kill” strategies cannot rely on the induction of virus expression in latently infected cells to result in cell death [6]. Therefore, candidate “kill” agents, such as immunotoxins, are being considered for incorporation into HIV eradication protocols [7], [8]. Immunotoxins are recombinant or biochemically linked bi-functional proteins that combine the effector domain of a protein toxin with the targeting specificity of an antibody or ligand [8]–[13]. Soon after HIV was identified as the causative agent of AIDS, several immunotoxins were described as potential therapeutics for HIV [8], [10]. These interventions had effector domains from plant and bacterial protein toxins and targeting moieties against either the HIV Env glycoprotein (gp120, gp41) or cellular markers including CD4, CD25 or CD45RO [14]–[21]. The immunotoxin we chose to evaluate for in vivo efficacy herein, 3B3-PE38, combines the 3B3 scFv which targets the conserved CD4 binding site of HIV-1 gp120 with the Pseudomonas exotoxin A (PE38) effector domain [22]. The fact that tissue specific effects of ART on HIV persistence are poorly understood in patients [23] meant that there was no baseline for characterizing the systemic effects of an immunotoxin on HIV persistence during ART. Therefore, it was essential that we first fully characterize the systemic impact of ART on HIV persistence cells in vivo. To do so, we sought to analyze HIV persistence during ART in a comprehensive panel of tissues and organs over time – a study that cannot be performed in human subjects. For this reason, we used BLT humanized mice [24], the most advanced, validated, and robust small animal model available for this purpose [25], [26]. The process of bioengineering BLT mice results in systemic dissemination of human hematopoietic cells throughout the animal [24], [27]. This phenotype facilitates the simultaneous analysis of multiple tissues throughout the body. The systemic effects of HIV infection on the BLT mouse human immune system (e. g. , CD4+ T cell depletion and immune activation) recapitulate what is observed in HIV-infected patients [28]–[32]. Once the tissue-specific parameters of HIV persistence during ART were established, we incorporated 3B3-PE38 into the therapeutic regimen. Systemic analyses revealed that 3B3-PE38 treatment during ART reduced the number of HIV RNA producing cells to levels significantly lower than those achieved with ART alone. This observation demonstrates that immunotoxins can play a critical role in successful HIV eradication strategies.
For this study, BLT mice were treated with a triple combination antiretroviral drug regimen that included the nucleotide reverse transcriptase inhibitor tenofovir disoproxil fumarate (TDF), the nucleoside reverse transcriptase inhibitor emtricitabine (FTC) and the integrase inhibitor raltegravir (RAL). This ART regimen was chosen because of its robust pharmacodynamic properties [33], superior efficacy in humans [34] and its efficacy in BLT mice [35]. In human peripheral blood [36], [37] and BLT mice (Fig. 1A), the first few weeks of ART are characterized by a rapid decline in plasma viremia (vRNA) followed by a plateau phase concomitant with a recovery of peripheral blood CD4+ T cells levels (Fig. 1B). We also observed that the blood cell-associated vDNA levels remained stable when compared to plasma vRNA levels during this treatment period (Fig. 1A) (Day −3 vs. Day42, p = 0. 63, signed rank test), as seen in patients on this same ART regimen [38]. Furthermore, the robust suppression of plasma viremia by ART is consistent with the presence of each of the dosed antiretrovirals in the plasma of treated BLT mice: tenofovir (Fig. 1C), emtricitabine (Fig. 1D) and raltegravir (Fig. 1E). Such peripheral blood analyses are readily performed in both humans and BLT mice; however, the ability to simultaneously examine multiple tissues throughout the course of ART is not possible in patients. Therefore, we evaluated antiretroviral drug penetration as well as the impact of ART on vRNA production in multiple organs in BLT mice. We determined drug levels in the thymic organoid, spleen, liver, lung, terminal ileum and rectum of BLT mice. Importantly, each of the three drugs was detected in each matrix analyzed (Fig. 1C–E). To facilitate direct comparisons of drug levels between tissues and plasma, the data are presented as ng/g and ng/ml, respectively [39]. Overall, the tenofovir, FTC and RAL levels in BLT mice were ∼1. 6 µM, ∼0. 4 µM and ∼0. 05 µM, respectively (Fig. 1C–E). Each of these values is higher than the EC50 for that drug in HIV-1 infected peripheral blood mononuclear cells (TDF: 0. 005 µM; FTC: 0. 01 µM; and RAL: 0. 001 µM) [39]–[41]. The extensive antiretroviral drug penetration observed in BLT mouse organs during ART led us to examine the systemic impact of ART on HIV production. We used two different approaches to measure the systemic reduction in vRNA by ART. First, we used in situ hybridization to quantitate the number of individual cells producing vRNA in the human thymic organoid, spleen, lymph nodes, liver, lung, terminal ileum and rectum of HIV infected BLT mice receiving or not receiving ART. We found that antiretroviral penetration and activity in these tissues was sufficient to profoundly reduce the number of cells producing vRNA in all tissues (Fig. 2). However, consistent with the limited human tissue data available [42], vRNA producing cells remained detectable during therapy in all tissues analyzed (Fig. 2). We also used RT-PCR to quantitate cell-associated vRNA levels in the bone marrow, human thymic organoid, spleen, lymph nodes, liver lung, intestines and peripheral blood cells of BLT mice given ART for 0–64 days. We observed a rapid decline in cell-associated vRNA levels that plateaued by Day 28 of ART in all tissues analyzed (Fig. 3A). The longitudinal vRNA data was analyzed using two different regression models (Lowess and cubic). The observed reductions in cell-associated vRNA in each tissue tested indicate that ART penetrates these tissues sufficiently to significantly reduce viremia in a related manner in each tissue as illustrated in the plots depicting the individual Lowess curves graphed together (Fig. 3B). We found that a Lowess curve generated with cell-associated vRNA data for all tissues together (Fig. 3C) is very similar to the reduction in plasma viremia observed specifically in human peripheral blood soon after ART initiation [36]. We also noted significant (p<0. 001) differences in each tissue, including peripheral blood, when we compared cell-associated vRNA levels from mice in an untreated state versus those from treated mice with vRNA levels within the plateau stage (Fig. 4). These similarities in vRNA reduction observed between human peripheral blood and all of the examined tissues from BLT mice during the first months following ART initiation suggest that blood is a reasonable surrogate for the impact of ART throughout the body. Since neither cytopathic effects of HIV expression or antiviral immune responses are sufficient to deplete cells expressing virus [6], these cells represent potential targets for cytotoxic HIV-specific immunotherapies such as the immunotoxin 3B3-PE38. This immunotoxin has been shown to be active against HIV-1-infected primary CD4+ T cells, macrophages and thymocytes [43], [44]. Knowing that the effect of ART on the vRNA production reached a plateau by Day 28 of treatment led us to ask whether targeted killing of the persisting HIV producing cells during this plateau phase would augment the vRNA decline during ART [8]. As with most anti-HIV agents, monotherapy with immunotoxins is not clinically viable [8]; therefore, we quantitated the systemic impact of 3B3-PE38 on HIV persistence during ART (7 IP doses on alternate days beginning on Day 28 of ART). Compared to ART alone, ART plus 3B3-PE38 reduced the cell-associated vRNA by 3. 2 logs (>1,000-fold) in the bone marrow (Fig. 5A). Complementing ART with 3B3-PE38 also led to a reduction in the persisting cell-associated vRNA levels in the human thymic organoid, spleen, lymph nodes, liver lung, intestines and peripheral blood cells (range: 0. 4 to 1. 5 logs) (Fig. 5A). When all tissues were analyzed together, ART plus 3B3-PE38 reduced cell-associated vRNA levels in all organs combined by an additional 0. 8 logs compared to ART alone (p<0. 001) (Fig. 5B). Following the initiation of ART in humans, there is a multi-phasic decay in viremia that reflects the rate of turnover of productively infected cells [36], [45]. Most models of viral dynamics assume that the drugs completely block new infections of susceptible cells, with the observed decay reflecting the decay rate of cells infected prior to the initiation of ART [36], [46]. Because this decay is constantly progressing, it is likely that the number of vRNA producing cells would decline over time during ART in the absence of immunotoxin treatment. However, given that we observed an overall 0. 8 log reduction in cell-associated vRNA levels beyond the effect of ART alone (Fig. 5B), we used in situ hybridization to examine whether there was an associated reduction in the number of cells producing vRNA. We found that complementation of ART with 3B3-PE38 led to a 1. 8 log reduction in the numbers of vRNA producing cells throughout the body (p = 0. 007) (Fig. 5C). This profound reduction in the number of vRNA producing cells provides a mechanistic explanation for the rapid reduction in tissue cell-associated vRNA levels relative to the ART only experimental group.
Despite the lifesaving benefits of ART in HIV patients, relatively little is known regarding the organ specific impact of this therapy on HIV production and persistence [23]. Improving our knowledge of the systemic effects of ART is a critical step in the development of successful HIV eradication therapies. To address this need, we characterized the impact of ART by analyzing both drug penetration and HIV production in different tissues throughout BLT humanized mice. We found that ART penetration into multiple organ systems is sufficient to significantly reduce the number of cells producing HIV, as well as cell-associated vRNA levels, throughout the body. However, HIV persists during ART in every organ analyzed. HIV-infected cells persisting during ART represent suitable targets for cytotoxic HIV-specific immunotherapies such as the 3B3-PE38 immunotoxin. Since such therapies target HIV proteins expressed by infected cells, their efficacy requires active virus transcription. Therefore there is no predicted impact of immunotoxin treatment on the size of the transcriptionally silent latent HIV reservoir. For this reason, the HIV latent reservoir was not quantitated in this study and our conclusions are based on the quantitation of cell-associated vRNA and vRNA-producing cell numbers. Specifically, we demonstrate that 3B3-PE38 kills vRNA producing cells throughout the body such that the reduction of vRNA levels during combined therapy is more rapid than with ART only (Fig. 6). A recent report described the ability of new broadly neutralizing monoclonal antibodies to suppress HIV-1 rebound after termination of ART [47] which led us to consider the possibility that the observed activity of the immunotoxin was due in part to neutralization by the 3B3 scFv moiety. In the Horwitz, et al. study, the amount of IgG (10. 5 mg/injection) administered was over 2000 fold higher than our doses of 3B3-PE38 (0. 005 mg/injection). Thus, the circulating 3B3-PE38 levels in our study could not reach the levels required for neutralization by 3B3 [48]. It is therefore unlikely under our experimental conditions that neutralization by the 3B3 scFv moiety of the immunotoxin accounts for the significant 3B3-PE38-mediated reduction we observed in tissue cell-associated vRNA levels. Comparing and contrasting data from BLT mice with data from patients and NHP is essential to understanding the predictive nature of our model. The most notable difference between our study and those in patients and NHP is the duration of ART. Our study examined the impact of ART in BLT mice over ∼2 months, while patient and macaque studies have examined cell-associated vRNA levels during several years of ART [49], [50]. These differences in ART duration prevent direct comparisons of the impact of long-term ART on persistent vRNA production within these experimental platforms. Despite this constraint, we can compare data across all three experimental platforms for consistency. For example, we found that our data showing the continued presence vRNA in BLT mouse peripheral blood and tissue cells during ART are consistent with the continued presence of vRNA within patient peripheral blood cells, ileal biopsy cells and rectal biopsy cells from patients treated for a median of 12. 5 months with ART [50] and within macaque cells from throughout the body after 26 weeks on ART [49]. In addition to future studies incorporating longer treatment windows in BLT mice, it will be important to determine whether the vRNA plateau phase reached after 28 days is directly comparable to that seen in humans on long term ART. For this purpose, incorporation of an additional drug into the ART regimen could be tested. If the plateau state in BLT mice is comparable to the human situation during long-term ART, then future studies using more sensitive assay for HIV-1 RNA in conjunction with ART intensification should not detect any additional reduction in the steady state level of viremia during this timeframe. The results presented here provide proof of concept for targeted cytotoxic therapy as a successful complement to ART for the depletion of persisting HIV infected cells. In addition to the 3B3-PE38, alternative immunotoxins with different effector domains and targeting moieties are available [8]–[10]. This is important because: (i) treatments with different cytotoxic interventions during ART will likely be required to completely eradicate HIV producing cells due to the immunogenicity of the immunotoxin molecules [8]–[13], [51], [52], the need to target the diverse cell types that are productively infected with HIV [2]–[4], and (iii) the likely selection of viral mutants resistant to any single immunotoxin [53]. Beyond immunotoxins, alternative targeted cytotoxic therapy strategies developed for cancer treatment could also be adapted to generate comprehensive anti-HIV cytotoxic therapy regimens [54], [55]. Possibilities include: aptamer—toxin conjugates; radioimmunotherapies; antibody—cytotoxic drug conjugates; targeted cytolytic viruses; targeted delivery of a cytotoxic peptides; and adoptive immunotherapy with ex vivo expanded natural or genetically modified T cells [54], [55]. Importantly, the model system developed herein to quantitate the effects of 3B3-PE38 can be used to determine the in vivo efficacy of any of these other strategies alone or in combinations. In summary, we determined that the systemic penetration of antiretroviral drugs into different organs throughout the body was sufficient to reduce cell-associated vRNA levels ART alone was unable to completely eliminate vRNA expressing cells in tissues. This observation provided a quantitative framework for the systemic in vivo efficacy evaluation of interventions to destroy HIV producing cells during therapy. Within this framework we were able to demonstrate that persistent HIV producing cells systemically present during ART were significantly depleted by an Env-targeting immunotoxin. Such targeted cytotoxic interventions may prove to be critically important components of an effective HIV eradication strategy.
All animal experiments were conducted following NIH guidelines for housing and care of laboratory animals and in accordance with The University of North Carolina at Chapel Hill (UNC Chapel Hill) regulations after review and approval by the UNC Chapel Hill Institutional Animal Care and Use Committee (permit number 12-171). BLT mice were prepared essentially as previously described [27]. Briefly, thy/liv implanted NOD/SCID-gamma chain null mice (NSG; The Jackson Laboratories, Stock #5557) were transplanted with autologous human fetal liver CD34+ cells (Advanced Bioscience Resources) and monitored for human reconstitution in peripheral blood by flow cytometry [24], [27], [32], [35]. All BLT mice (n = 40) used for these experiments were characterized for human immune system reconstitution prior to HIV infection. Their peripheral blood contained an average of 57% (+/−15 SD) human CD45+ cells of which 58% (+/−22 SD) were human T cells. Of the human T cells, 81% (+/−6 SD) were human CD4+ T cells. Infection of BLT mice with HIV-1JRCSF was monitored in peripheral blood by determining plasma levels of vRNA as described (level of detection = 750 copies per mL of plasma; 40 µl mouse plasma sample size) using one-step reverse transcriptase real-time PCR [ABI custom TaqMan Assays-by-Design] [35], [56]–[58]. Tissues were harvested and cells isolated as we have previously described for RNA isolation or flow cytometric analysis [24], [32]. Flow cytometry data were collected using a BD FACSCanto cytometer and analyzed using BD FACSDiva software (v. 6. 1. 3). For HIV RNA in situ hybridization (ISH), tissues were fixed overnight in 4% paraformaldehyde at 4°C and then transferred to 70% ethanol. Tissues were then embedded in paraffin for sectioning. After deparaffinization with xylene and rehydration through graded ethanols, tissue sections were treated with HCl, triethanolamine, digitonin and 4 µg/mL Proteinase K as previously described [59]. After acetylation with acetic anhydride and dehydration, tissue sections were hybridized at 45°C overnight with a 35S labeled antisense riboprobe and 0. 5 mM aurintricarboxylic acid in the hybridization mix. After extensive washes and ribonuclease treatment, tissue sections were dehydrated, coated in Ilford K5 emulsion diluted with glycerol and ammonium acetate, exposed at 4°C for 7–14 days, and developed and fixed per manufacturer' s instructions. They were stained with Hematoxylin, dehydrated, and mounted with Permount. Non-infected control tissues and sense riboprobe controls were analyzed when appropriate. Photographic images using epifluorescence were taken with a digital camera and the tiffs were analyzed for the area of the sections and the area occupied by silver grains using Photoshop with Fovea Pro. Section weights were estimated from their 5-micron thickness and their area. These methods have been extensively reviewed [59]. ART in infected BLT mice consisted of three antiretrovirals administered daily via intraperitoneal injection: tenofovir disoproxil fumarate (TDF; 208 mg/kg), emtricitabine (FTC; 240 mg/kg body weight) and raltegravir (RAL; 56 mg/kg) [35]. The gp120-targeting immunotoxin 3B3 (Fv) -PE38 [22] (here referred to as 3B3-PE38) was administered intraperitoneally every other day beginning on Day 28 of ART for a total of 7 doses: the first 4 were 1 µg/25 g and the last 3 were 5 µg/25 g. Plasma and snap frozen tissue samples from BLT mice receiving daily ART were analyzed for drug concentrations. Quantification of FTC, tenofovir (TFV) and RAL plasma concentrations was performed by protein precipitation and LC-MS/MS analysis. 10 µl of each stored plasma sample was mixed with 75 µL of methanol containing the isotopically-labeled internal standards (13C TFV and 13C 15N FTC). Following vortexing and centrifugation, the supernatant was removed and evaporated to dryness. The extracts were reconstituted with 1 mM ammonium phosphate prior to LC-MS/MS analysis. All compounds were eluted from a Phenomenex Synergi Polar-RP (4. 6×50 mm, 4 µm particle size) analytical column. An API-5000 triple quadrupole mass spectrometer (AB Sciex, Foster City, CA) was used to detect the analytes. Data were collected using AB Sciex Analyst Chromatography Software (Analyst version 1. 6. 1). The dynamic range of this assay was 1–1000 ng/mL for each compound using a 1/concentration2 weighted linear regression. Calibrators and quality control samples were within 15% of the nominal value for all compounds. Plasma samples with concentrations above the calibration range were diluted into the calibration range with blank plasma prior to reanalysis. The quantitation of FTC, TFV and RAL in tissues started with the homogenization of each tissue in a 70% acetonitrile + 30% 1 mM ammonium phosphate solution. A portion of the resulting homogenate was extracted by protein precipitation with acetonitrile containing isotopically-labeled internal standards (13C TFV, and 13C 15N FTC). Following vortexing and centrifugation, the supernatant was removed and evaporated to dryness. The extracts were reconstituted with 1 mM ammonium phosphate prior to LC-MS/MS analysis. TFV and FTC were eluted from a Waters Atlantis T3 (100×2. 1 mm, 3 µm particle size) analytical column. RAL was eluted from a Phenomenex Synergi Polar-RP (50×4. 6 mm, 4 µm particle size) analytical column. An API- 5000 triple quadrupole mass spectrometer was used to detect all analytes. Data were collected using AB Sciex Analyst Chromatography Software. The dynamic range of this assay was 0. 3–300 ng/mL of homogenate for each compound using a 1/concentration2 weighted linear regression. Based upon the mass of the tissue extracted, the concentrations were converted from ng/mL homogenate to ng/g tissue. Calibrators and quality control samples were within 15% of the nominal value all compounds. All statistical tests performed using an alpha level of 0. 05. Exact log rank tests were utilized to generate p values for Fig. 2 (R v2. 14. 1). Mann-Whitney comparisons were utilized to generate the p values presented in Figs. 4 & 5A (Prism v4). Wilcoxon rank-sum statistics with repeated measures corrections were utilized to generate p values presented in Fig. 5B (R). Exact log rank tests with repeated measures corrections were utilized to generate p values for Fig. 5C (R). For Figs. 3 & 6, the Lowess and cubic regression models were fit using R. The linear heterogeneity models that allowed for the residual variance to change over time were computed in SAS/STAT software. Graphs were generated in Prism v4 with the exception of the fitted Lowess and cubic regression models which were generated in R. | Antiretroviral therapy (ART) improves the quality of life for HIV infected individuals. However, ART is currently a lifelong commitment because HIV persists during treatment despite being suppressed below detection. If therapy is stopped, the HIV reappears. A concerted effort is ongoing to develop new eradication therapies to prevent virus rebound, but there are challenges to be overcome. Our work is a major step forward in this process. We measured persistent HIV throughout the body during ART using bone marrow/liver/thymus (BLT) humanized mice, a model validated to study HIV persistence. HIV infected BLT mice were treated with tenofovir, emtricitabine and raltegravir. Despite documented tissue penetration by these drugs, we found that HIV expression persists in cells isolated from all the tissues analyzed (bone marrow, thymus, spleen, lymph nodes, liver, lung, intestines and peripheral blood cells). We therefore complemented ART with an immunotoxin that specifically kills HIV expressing cells while leaving other cells untouched. Our results demonstrate a dramatic reduction in persistent HIV throughout the body resulting from the killing of virus producing cells. Thus, our study provides new insights into the locations of HIV persistence during ART and a demonstration that persistent HIV can be successfully targeted inside the body. | Abstract
Introduction
Results
Discussion
Materials and Methods | medicine
infectious diseases
preclinical models
hiv diagnosis and management
clinical research design
immunity
hiv
drugs and devices
hiv clinical manifestations
immunology
drug distribution
biology
viral diseases
infectious disease modeling
pharmacokinetics
immunotherapy | 2014 | Targeted Cytotoxic Therapy Kills Persisting HIV Infected Cells During ART | 6,142 | 308 |
Stigma plays in an important role in the lives of persons affected by neglected tropical diseases, and assessment of stigma is important to document this. The aim of this study is to test the cross-cultural validity of the Community Stigma Scale (EMIC-CSS) and the Social Distance Scale (SDS) in the field of leprosy in Cirebon District, Indonesia. Cultural equivalence was tested by assessing the conceptual, item, semantic, operational and measurement equivalence of these instruments. A qualitative exploratory study was conducted to increase our understanding of the concept of stigma in Cirebon District. A process of translation, discussions, trainings and a pilot study followed. A sample of 259 community members was selected through convenience sampling and 67 repeated measures were obtained to assess the psychometric measurement properties. The aspects and items in the SDS and EMIC-CSS seem equally relevant and important in the target culture. The response scales were adapted to ensure that meaning is transferred accurately and no changes to the scale format (e. g. lay out, statements or questions) of both scales were made. A positive correlation was found between the EMIC-CSS and the SDS total scores (r = 0. 41). Cronbach' s alphas of 0. 83 and 0. 87 were found for the EMIC-CSS and SDS. The exploratory factor analysis indicated for both scales an adequate fit as unidimensional scale. A standard error of measurement of 2. 38 was found in the EMIC-CSS and of 1. 78 in the SDS. The test-retest reliability coefficient was respectively, 0. 84 and 0. 75. No floor or ceiling effects were found. According to current international standards, our findings indicate that the EMIC-CSS and the SDS have adequate cultural validity to assess social stigma in leprosy in the Bahasa Indonesia-speaking population of Cirebon District. We believe the scales can be further improved, for instance, by adding, changing and rephrasing certain items. Finally, we provide suggestions for use with other neglected tropical diseases.
The framework for cross-cultural equivalence testing used in this study, draws entirely on the work of Herdman et al [25], [30], Terwee et al [26] and Stevelink & van Brakel [29]. Five equivalences and the universalist approach are important for this study. Herdman et al note that a universalist approach: ‘Conceptual equivalence’ looks at how the concept of social stigma is conceptualized, which domains are important and at the significance accorded to these domains. ‘Item equivalence’ similarly explores how domains are conceptualized and whether items are equally relevant and acceptable in the original and the new culture. ‘Semantic equivalence’ deals with language and how meaning is transferred, for instance, whether the level of language is appropriate. ‘Operational equivalence’ concerned the suitability of the questionnaire format, instructions and mode of administration. Finally, ‘measurement equivalence’ refers to the psychometric properties (internal consistency, construct validity, agreement, reliability, floor and ceiling effects and interpretability) of the scale (for a more detailed description of each equivalence type we refer to Herdman et al [30], [35]). Table 1 describes when each of these equivalences is attained.
The study area of the SARI project is Cirebon District, located on the North Coast of West Java near the provincial border with Central Java. Cirebon District has a multi-cultural population of about 2. 3 million. Different languages are spoken, such as, Bahasa Indonesia (the national language), Sundanese, Javanese and Cirebonese. Annually, about 300 new leprosy cases are detected in the district and, according to key informants, there was a high level of leprosy-related stigma and limited activities to reduce this. The stigma-reduction interventions of the SARI project are implemented in 30 kecamatan (sub-districts). The SARI project team is interdisciplinary, including staff from public health, medicine, disability studies, psychology and development studies from universities in the global North and South. This validation study was executed by one postdoc researcher, three PhD students and ten research assistants from Cirebon or neighbouring districts who spoke the local languages. Four of the research assistants who interviewed community members were disabled or affected by leprosy themselves. The study described in this paper is part of a larger validation study that included persons affected by leprosy and community members from the 30 kecamatan described above. The latter group is the study population for this paper. To achieve adequate power for the various statistical calculations we estimated a sample size of at least 100 community members, with at least 50 repeated assessments to assess reproducibility [26]. The selection was done as follows; first, people affected by leprosy were invited to the puskesmas (Health Care Centre) for an interview. At each puskesmas, three persons affected by leprosy were randomly selected. For each respondent a small paper with a number was created, three papers were drawn, if the respondent came from the same village a new paper was drawn. Their Rukun Tetangga (RT, smallest administrative level in Indonesia approximately 10–20 households) was visited by a small team of research assistants (2–3) of the SARI project. First they visited the head of the RT to introduce themselves and explained the purpose of the project. Using convenience sampling, they then selected three community members from this RT or a neighbouring RT for the interviews. Two key persons, such as, the head of the RT, a teacher, religious leader, women' s leader and one general community member about the same age and sex of the person affected interviewed that morning were selected. Data was collected during three phases: i) first validation study in August 2011, ii) baseline study from September – October 2011 and the iii) second validation study in July 2012. The EMIC-CSS was selected based on its prior cross-cultural and cross-condition use [5], [33]–[35], [38]. The scale has 15 items and covers areas of life that are often affected by stigma, such as concealment, avoidance, perceptions of self-worthy, shame, marriage (prospects) and work. The scale has four response options; yes (2 points), possibly (1), no (0) and do not know (0). Item 15 is scored differently; yes (0 points), possibly (1), no (2) and do not know (0). There was no qualitative component used as part of the scale, as in some previous studies [22], [38]. The SDS was selected because it measures attitudes more directly than the EMIC-CSS and had been used widely in mental health research in different countries [24], [42], [43], [47]. The scale is also short, simple and easy to contextualise, because of the use of vignettes. The SDS interview started with reading out a vignette describing a male named Rahman or female named Rahmi, depending on the sex of the interviewee. The content of the two vignettes is similar. The vignettes were developed by one of the co-authors (WvB) based on vignettes used in the field of mental health used by Angermeyer et al [42], [44], [48]. The scale has 7 items representing different degrees of social distance. The items have four response options; definitely willing (0 points), probably willing (1), probably not willing (2) and definitely not willing (3). Both scales assess aspects of the same construct ‘social stigma’, but take a different approach; the EMIC-CSS asks how leprosy is considered in the community of the interviewee, while the SDS assesses the personal perception of the interviewee. The sum score of the individual items that all have the same weight is used as the overall score and higher scores reflect greater levels of social stigma. The scales are interviewer administrated. Each respondent was first asked to provide demographic information, such as, age, sex, profession and income, next the EMIC-CSS was administrated followed by SDS with vignette. This order was chosen because this sequence allows questions to go from general community perspectives to specific and personal choices and avoids ‘contamination’ of the EMIC-CSS with the vignette. When a respondent did not speak Bahasa Indonesia with sufficient fluency, the questions were translated on the spot into the first language of the respondent often Sundanese or Javanese. To determine the conceptual, item, semantic and operational equivalence different steps were taken. First, an exploratory study took place in which 53 in-depth interviews and 20 focus group discussions (FGDs) were conducted to understand the cultural background and situation in which people lived (see for more details on the methods [1]). Second, the versions of the EMIC-CSS and the SDS that were selected for this study were translated from English to Bahasa Indonesia by someone knowledgeable regarding stigma and later back–translated to English by someone not involved in stigma research. Third, a discussion on the content of the instruments, the vignettes, the phrasing of items, and the response scales took place within the team and with experts knowledgeable on Bahasa Indonesia, leprosy, stigma and quantitative instruments. Fourth, two half-day pre-test sessions were organised with 20 participants (people affected by leprosy and with a disability). The questions of the instruments and the vignette were checked with the participants for coherence, understanding and terminology. Fifth, the research assistants of the SARI project received a full week of training in the use of the scales, with practice sessions in the office. Finally, a two-week pilot study was conducted in the study area, with daily meetings in the office. Once all scales were optimized and the interviewers felt confident, the data collection for testing measurement equivalence started followed by the baseline study. During the validation and baseline study, weekly meetings were held to discuss issues that had arisen during the interviews. To determine the measurement equivalence, the data was entered using Epi Info for Windows, version 3. 5. 3, and analysed using Stata 12. 1 and SPSS 21. Records were deleted from the raw database if the demographic information or a full scale was missing. Outliers were explored with descriptive statistics and box plots. Interviews conducted in a language other than Bahasa Indonesia were left out the analyses. To provide an overview of the socio-demographic characteristics of the sample, basic descriptive statistics were calculated. The respondent was asked for either income per day or income per month; the latter was converted into one variable ‘household income per day’ by dividing the income per month by 30. 5. A mean and SD were used to describe each item of the scales. Psychometric properties were tested using appropriate statistical methods based on predefined quality criteria. The term ‘assessing’ stigma is used throughout this paper, instead of for example ‘measuring’, ‘evaluating’, ‘quantifying’ or ‘rating’ stigma as this reflects best the aim of the applying the scales. The study was approved by the relevant offices; Ethics Committee of Atma Jaya University; Sub-Directorate for Leprosy and Yaws, Ministry of Health, Public Health Office, West Java and District Health Office, Cirebon District. Written consent was obtained from individual study subjects. The study guarantees the confidentiality of the information provided by the participants. No incentives were offered to interviewees other a small token of appreciation such as a drinking mug or t-shirt. The study abided by the CIOMS Guidelines for Research on Human Subjects [50].
Based on the opinion of experts and the responses of participants of the pre-test sessions and pilot study, the domain ‘social distance’ employed in the SDS seems equally relevant and important in the target culture. Because a Likert scale is used (instead of the original Guttman scale), the fact that the type of relationship might represent different degrees of social distance is not relevant. The EMIC-CSS assesses different aspects of a broader phenomenon that can be described as ‘perceived stigma against persons affected by leprosy’. The aspects that can be recognized in the scale applied by van Brakel et al [36] are: i) concealment (2 items), ii) process of discrediting (3 items), iii) shame and embarrassment (1 items), iv) avoidance/taking distance/isolation (2 items), v) problems with getting married or on-going marriage (2 items), vi) problems for family or other people (3 items) and vii) problems with work (2 items). The exploratory study of the SARI project described in Peters et al [1] already indicated that the aspect ‘shame and embarrassment’ and ‘avoidance/taking distance/isolation’ are relevant in the target culture. In the interviews and FGDs we found evidence for the relevance of all the other aspects and there were no indications that led us to change the emphasis placed on the aspects. In the following quote ‘concealment’ comes to the front, and at the same time reveals shame and avoidance: The following three quotes support the relevance of the aspect problems with getting married and on-going marriage: The following quotes confirms the relevance of the aspects problems for family or other people. The first comes from a FGD among mothers of children affected by leprosy, the second from discussion among community leaders: The next quote illustrates the significance of the aspect process of discrediting: One relatively new aspect in the EMIC-CSS is the aspect ‘problems with work’. Several studies [7], [36] have shown that this aspect is relevant. Also the data from the interviews and FGD strongly support this as shown by several quotes each highlighting a different element or perspective of this aspect. The first three quotes are from persons affected by leprosy the latter three from community members: Based on the opinion of experts and the response during the pre-test sessions and the pilot study there was no indication for a need to change any of the items in the SDS or in the EMIC-CSS. The target population speaks different languages, but the scales are translated in Bahasa Indonesia only, because this is the national language and is most commonly spoken by the target population. Some minor changes were made in the first version of the translation to make sure that the words in the scales fitted the day-to-day language used in the people in the rural areas of Cirebon. The response options ‘possibly’ of the EMIC-CSS and ‘probably’ in SDS were difficult to translate into Bahasa Indonesia and therefore changed into ‘maybe’ translated as ‘mungkin’. Sometimes, interviewees requested to fill the forms by themselves, which was often accepted. The interviewer would be there to answer any questions. Therefore, a mixture of interviewer-administrated and self-administered form filling was used, using the same questionnaires for both. No other changes were made to the administration, formats of scales and their scoring. A total of 326 observations were in the initial database. Of these, 29 (8. 9%) were omitted due to missing values and 38 (11. 7%) were omitted due a language used other than Bahasa Indonesia. The remaining 259 community members were included in this validation study. The observations omitted differed from the main sample. The former were less frequently male (58. 8% versus 62. 2%), were older (mean 46. 9 versus 42. 1 years), more frequently married (98. 0% versus 91. 1%) and had fewer years of education (6. 1 versus 9. 1 years). Of the 259 observations, 72 were collected during the first validation study, 142 during the baseline and finally 46 during the second validation study. Their socio-demographic characteristics are described in Table 2. The key persons in this sample were more frequently men (75% compared to 42%), had higher age (mean 44. 3 versus 39. 4 years), were more frequently married (96% versus 85%) and had a higher level of education (mean 10. 2 versus 7. 5 years) than the respondents in the ‘general’ community sample (data not shown). The mean total score of the items of the EMIC-CSS was 15. 38 (SD 6. 46) and ranged from 0 representing the minimum stigma score to 30 representing the maximum total score. These figures for the SDS are, respectively, 9. 05 (SD 4. 01) and 0 to 21. Table 3 and 4 provide the mean score per item. We found a moderately positive correlation between the EMIC-CSS total score and the SDS total score (r = 0. 41). We identified one outlier with contradicting total scores; EMIC-CSS total score of 0 and a SDS total score of 19. This respondent frequently answered ‘do not know’ at the items of the EMIC-CSS. Without this outlier, the correlation increased somewhat (r = 0. 45). This correlation confirmed the a priori hypothesis. Cronbach' s alphas of 0. 83 and 0. 87 were found for the EMIC-CSS and SDS, respectively. Item E15 of the EMIC-CSS has a low item-test correlation (0. 16) and item-rest correlation (0. 04); if left out Cronbach' s alpha of the EMIC-CSS increases slightly to 0. 84. The exploratory factor analysis for both scales showed an adequate fit as a one-dimensional scale, with a first factor explaining 77% of the score variability for EMIC-CSS and 94% for SDS. However, additional factor analysis of the EMIC-CSS also supports two factors as shown in Table 5. The first factor with 9 or 10 items and a second with 4 or 5 items. The two factors were strongly correlated (r = 0. 63), supporting the presence of a single higher-order factor. Item E15 did not fit well in either scale and was therefore omitted. Cronbach' s alphas for the subscales were sufficient and are provided in Table 6. While exploring the data with frequencies and a box plot several outliers where identified and these were checked visually. Three observations seems to be errors and were therefore deleted from the database leaving in 67 repeated observations. Community members were revisited on average after 12 days, but at least after 3 and before 29 days. The mean difference between interviewers is in the EMIC-SDD −0. 52 (SD 3. 37). This led to a SEMagreement of 2. 38, which represents 7. 9% of the score range. The limits of agreement are −7. 12 and 6. 08. The SDCindividual is 6. 60 and SDCgroup is 0. 81. In the SDS, the mean difference between interviewers is −0. 06 (SD 2. 54). The SEMagreement 1. 78, this is 8. 6% of the total score range. The limits of agreement are −5. 04 and 4. 91. The SDCindividual is 4. 94 and SDCgroup is 0. 60. For the EMIC-CSS and the SDS the test-retest reliability was above 0. 70, the ICCagreement is respectively 0. 84 (Confidence Interval (CI) 0. 75–0. 90) and 0. 75 (CI 0. 62–0. 84). No floor or ceiling effects were identified for the EMIC-CSS of SDS. Only 2 (0. 8%) of the respondents scored the lowest possible score of 0 and also 2 (0. 8%) scored the highest possible score of 30 points on the EMIC-CSS. For the SDS, only 4 respondents (1. 5%) had the lowest possible score of 0 and 1 (0. 4) had the highest possible score of 21. The means and SD of the different subgroups of the baseline data (n = 142) show varied results as illustrated in Table 7. The mean total score of EMIC-CSS and SDS is higher in females, but the differences are very small. Among age groups, the EMIC-CSS steadily increases with age, but for SDS it slightly drops at first before increasing again. EMIC-CSS and SDS total scores follow a similar fluctuating pattern across education groups. Finally, EMIC-CSS and SDS total scores are lower fore key persons compared to the ‘general’ community. A summary of the key findings for the two scales can be found in Table 8.
The EMIC-CSS assesses the perceptions towards people affected by leprosy from a general community perspective (E4, E6, E7, E11, E12, E14, E15). It also addresses perceptions towards family members of a person affected by leprosy (E2, E8, E9, E13), towards other persons near a person affected by leprosy (E5) and the disease in general (E3). The EMIC-CSS assesses different aspects related to the social stigma of leprosy. This study showed that all aspects and items assessed in the EMIC-CSS are relevant in the target culture. The question whether all aspects together comprehend the concept of social stigma is more difficult and also a more theoretical/fundamental question. Experiences, such as, mocking and gossiping are real and very important experiences of people affected by leprosy in the target community [1]. These aspects are, for instance, not yet assessed and could be added to improve the content validity of the scale. Suggestions for items are: “Would other people in your community gossip about a person affected by leprosy? ” or “Would other people in your community mock a person affected by leprosy? ” Two relatively new items related to ‘problems with work’ were shown to be highly relevant based on the qualitative and quantitative data. The items have high total means scores. ‘Would having leprosy cause difficulty for a person to find work? ’ has a mean score of 1. 53 (highest) and ‘Would people buy food from a person affected by leprosy? ’ has a mean score of 1. 28 (reverse coded; fourth highest). The psychometric property results, however, show that the last item does not fit in the scale. This item might be an early sign of stigma, it may be scored positive while items are scored negative. We have considered ‘Would people dislike buying food from a person affected by leprosy? , but the translations of this question caused some confusing in Bahasa Indonesia. An option to pilot in the future would be: ‘Would having leprosy cause difficulty for a person to sell food? ’ For now, we recommend researchers who wish to validate or apply the EMIC-CSS in their study to include this item if it is considered culturally relevant. Investigators who are particularly interested in this item may want to develop and test a set of questions that assess the same activity in different ways. In our opinion, question 5 ‘Would knowing that someone has leprosy have an adverse effect on others? ’ is a rather abstract question in particular the ‘others’ part could have multiple interpretations. Although this did not cause challenges in our study, we would recommend caution when piloting this question to make sure the question is phrased in clearly understandable terms. The construct validity of the EMIC-CSS (and automatically also of the SDS) was supported by the moderately positive correlation between the EMIC-CSS total score and the SDS total score. Cronbach' s alpha found in this study (0. 83) is comparable to the values found in other studies [37], [38]. The exploratory factor analysis indicates an adequate fit for a one-dimensional scale, with a first factor explaining 77% of the score variability. We conclude, therefore, that the internal consistency is good. The factor analysis and internal consistency analysis illustrate that two shorter versions of the EMIC-CSS may also be of value; one with 9–10 items that could be conceptualised to assess ‘perceived attitudes towards persons affected by leprosy’ and another with 4–5 items that would assess ‘perceived behaviour towards persons affected by leprosy’. The responsiveness of the EMIC-CSS was sufficient in this study. For evaluation purposes a small measurement error is required, as one wants to be able to distinguish clinically important change from measurement error [51]. The SDC of the EMIC-CSS was small at the group level (0. 81 out of a score range of 30), but larger at the individual level, which means that at individual level, large score differences are required to demonstrate changes, while at group level, small score differences will already be sufficient. With an ICCagreement of 0. 84 the reliability is good. The absence of floor and ceiling effects was akin to other studies [37]. For interpretability, the SDC should ideally be compared with the score difference representing Minimally Important Change (MIC). However, this figure is not yet available for the measures under study. We agree with de Vet et al (22) and prefer not to calculate a MIC based on statistical tests but to use an anchor method. Hence, we underline the importance of future research to investigate the MIC for this scale. The EMIC-CSS as used in the SARI project did not include a qualitative component as described by Weiss [22]. We would like to underscore that we do value qualitative methods very much, especially in the context of the concept of stigma. In the SARI project we use separate methods for qualitative data collection. This study has shown that the EMIC-CSS is adequately culturally valid in the field of leprosy in Cirebon District. The scale could easily be adapted to other NTDs. This has indeed been done already in the case of onchocerciasis and Buruli Ulcer [5], [38]. Certain items can be more or less relevant in different conditions. For instance, the item related to food might be less relevant when there are no feelings of disgust or fear for infection, as in vector-borne diseases. The SDS assesses the perceptions of the interviewee towards people affected by leprosy by asking how they feel regarding different types of social relationships (e. g. neighbours, caretakers, colleagues). The concept of social distance has been studied in several countries around the globe, including Argentina [52], Japan and China [53], Nigeria [54], Germany [42], and Egypt [55]. This confirms its cross-cultural value. The cross-cultural validation of the SDS has to our knowledge not received any attention and, therefore, this is the first study of its kind. The current study has shown that the concept of social distance and the different types of relations used in the items of the scale are relevant and understood in Cirebon District. In retrospect, the item ‘renting a room to an affected person’ is less appropriate in the context of Cirebon District. Difficulties did not emerge with this item during the pilot, validation or baseline study. However, it is not a common practice for community members in the primarily rural study area to rent a room to somebody. Respondents could envision the situation, but were by and large not acquainted with the practice themselves. To make the scale more appropriate for Cirebon District or similar contexts, a replacement item is suggested. A first exploration with the SARI research assistants resulted in the following suggestion: ‘How would you feel visiting the house of someone like Rahman/Rahmi? ’ It important to note that the item renting a room seems appropriate in other contexts in Indonesia. For instance, in Yogjakarta, a university city, many households offer kost (rooms for rent). The internal consistency of the SDS in this study was good. Cronbach' s alpha (0. 87) was equivalent to the alphas found in other studies [24], [42], [43]. Factor analysis suggested one factor, which accounted from 94% of the variance. The SDC of the SDS was small at the group level (0. 61 out of score range of 21), but larger at the individual level (4. 98) resulting in the implications as described for the EMIC-CSS. The ICCagreement was 0. 75, indicating good inter-interviewer agreement. For interpretability a MIC is needed, which is unfortunately not yet available. This study demonstrated that also the SDS is adequately culturally valid in the context of leprosy in Cirebon District. The scale can be easily adapted to other NTDs, by changing the vignette. The items need to be checked for relevance and appropriateness in the new target culture. First, the convenience sampling used, the difference between observations omitted and the main sample and the high proportion of key persons are weaknesses in this study. Ameliorating circumstances are the size and diversity of the sample and the fact that the key persons are likely to know and represent well the views of their community regarding leprosy. The influence of the sampling bias on the results presented in this paper are in our point of view minor. We do not expect any influence on the conceptual, item, semantic and operational equivalences, due to the fact that the exploratory study included ‘general’ community members, the pre-test sessions were with people affected by leprosy and disabled people and the still relatively large group (99) of ‘general’ community members in the sample. We expect an influence on the measurement equivalence, but only on the interpretability as illustrated in the results; a higher total score of the EMIC–CSS and SDS for the ‘general’ community compared to key persons. Second, the effect of the mixture of interviewer-administrated and self-administrated is difficult to ascertain, because not all research assistants did this and those who did, did not do it with all respondents. Given that the same questionnaires were used and that the interviewer was present while the respondent filled in the questionnaire, we do not expect an important influence. Third, the fact that we were able to only validate Bahasa Indonesia versions of the EMIC-CSS and SDS is another weakness. A substantial group of people in Cirebon District do not speak Bahasa Indonesia sufficiently and to assess the level of social stigma in these groups, scales in different languages will still be needed. Some reflections on the process follow. First, as mentioned in the introduction, the validation of scales is a crucial process as unreliable and invalid scales can lead to wrong conclusions. This places a responsibility on researchers who intend to use an instrument in a new cultural setting or with a different target group to test the validity of the instrument using state-of-the-art qualitative and quantitative methods. Tools like the Herdman-Stevelink framework are helpful in conducting such a validation study [29], [30]. Second, a valid scale does not mean a perfect scale that we should set in stone and leave untouched. This paper shows that although the scales are valid, there remain several points for improvements. Several of these suggestions came from observations and reflections after the testing, training and piloting phases. We would like to recommend other researchers also to continue reflecting on valid scales as this might generate valuable insights and lessons in the future. Third, can the construct of stigma be assessed and, if so, how can this best be done? Opinions differ within our own team and these can be linked to our different scientific disciplines and epistemologies. However, we all agree that a combination of qualitative and quantitative methods offer the richest perspective. In relation to stigma-reduction interventions, the data that comes from quantitative assessments have particular value in determining the effectiveness of such interventions in groups and when generalizability is important. Qualitative assessment is particularly valuable when we look at individuals and want to understand the changes (and underlying reasons for these) interventions brought in their lives. Fourth, we consider it important to reflect on the impact assessing stigma has on the interviewee. It goes without saying that an instrument or the way data is collected should not create concern or discomfort regarding people affected by leprosy. This is why, in this type of research, the attitudes, understanding and skills of the interviewers and hence their training is crucial. Also the code of conduct of the team, for instance, on how to deal with questions about leprosy from the interviewee is vital. Phrasing questions in scales more positively, e. g. , ‘Do people in your community support people affected by leprosy? ’ Or assessing different constructs, such as social closeness, inclusion and care would be interesting topics for future research. Fifth, Parker and Aggleton noted that the way one conceptualizes and investigates the construct of stigma influences forms of intervening [13]. To assess social stigma, the concept of stigma needs to be conceptualized, but this simplification of a complex construct should then not dictate other activities in the field of stigma. Simplification for the purpose of quantitatively assessing the effect of interventions is valid. However, when designing or implementing stigma reduction interventions, we believe that we need to step back, appreciate and take into account the complexity of the concept of stigma. According to current international standards, our findings indicate that the EMIC-CSS and the SDS have adequate validity to assess social stigma of leprosy in the Bahasa Indonesia-speaking population in Cirebon District. However, these findings cannot be generalized to other NTDs, countries or even other provinces in Indonesia that are culturally different, such as Papua, Sulawesi, and Nusa Tenggara, where they would need to be re-validated. We believe the scales can be further improved and we have provided several suggestions in the discussion. With some adaptations the scales can be validated for other NTDs. | Persons affected by neglected tropical diseases, such as, Buruli ulcer, lymphatic filariasis, onchocerciasis, leishmaniasis and leprosy, can experience stigma. One important source of stigma are members in the community. Neighbours, religious leaders, and community leaders can exclude, reject, blame or devalue a person affected by one of these diseases. It is important to be able to assess this type of stigma for the prevention and management of these diseases. Assessing stigma is not an easy task. There are several instruments available, but these were developed with different aims or tested in different settings. We can use these instruments, but we need to be sure that they assess what we want them to assess and whether the instrument produces consistent results. In this paper the authors report a study that investigated the validity of two scales that assess stigma in the community towards people affected by leprosy in Indonesia. The names of the scales are Explanatory Model Interview Catalogue Community Stigma Scale (EMIC-CSS) and Social Distance Scale (SDS). The results show the two scales to be adequately valid and reliable in the target culture. There are, however, also several improvements possible and the authors provide suggestions how to incorporate these. In addition, the authors provide recommendations for the use of these scales among people affected by other neglected tropical diseases. | Abstract
Introduction
Methods
Results
Discussion | epidemiological statistics
public and occupational health
bacterial diseases
infectious diseases
medicine and health sciences
disabilities
epidemiology
leprosy
neglected tropical diseases
tropical diseases
epidemiological methods and statistics | 2014 | The Cultural Validation of Two Scales to Assess Social Stigma in Leprosy | 7,583 | 296 |
The intricate interactions between viruses and hosts include exploitation of host cells for viral replication by using many cellular resources, metabolites and energy. Tomato bushy stunt virus (TBSV), similar to other (+) RNA viruses, induces major changes in infected cells that lead to the formation of large replication compartments consisting of aggregated peroxisomal and ER membranes. Yet, it is not known how TBSV obtains the energy to fuel these energy-consuming processes. In the current work, the authors discovered that TBSV co-opts the glycolytic ATP-generating Pgk1 phosphoglycerate kinase to facilitate the assembly of new viral replicase complexes. The recruitment of Pgk1 into the viral replication compartment is through direct interaction with the viral replication proteins. Altogether, we provide evidence that the ATP generated locally within the replication compartment by the co-opted Pgk1 is used to fuel the ATP-requirement of the co-opted heat shock protein 70 (Hsp70) chaperone, which is essential for the assembly of new viral replicase complexes and the activation of functional viral RNA-dependent RNA polymerase. The advantage of direct recruitment of Pgk1 into the virus replication compartment could be that the virus replicase assembly does not need to intensively compete with cellular processes for access to ATP. In addition, local production of ATP within the replication compartment could greatly facilitate the efficiency of Hsp70-driven replicase assembly by providing high ATP concentration within the replication compartment.
Positive-strand (+) RNA viruses build robust viral replication machineries, called replication organelles, with the help of many co-opted cellular factors. In addition, viruses also obtain metabolites and energy from the infected cells. Overall, (+) RNA viruses induce major metabolic and structural changes in the infected cells, which frequently lead to disease states [1,2, 3,4, 5]. One of the best characterized (+) RNA virus is TBSV, which induces large replication compartments consisting of aggregated peroxisomal and ER membranes. The replication compartments contain numerous vesicle-like structures in the limiting membrane of peroxisomes, which harbor the viral replicase [6,7, 8]. TBSV also manipulates the cytoskeleton and endosomal trafficking [9,10]. TBSV co-opts numerous host proteins to support various viral functions, including the heat shock protein 70 (Hsp70), the endosomal sorting complex required for transport (ESCRT) machinery, translation factors and DEAD-box RNA helicases [11,12,13,14,15,16,17]. TBSV also induces enrichment of sterols at the viral replication sites via co-opting oxysterol-binding proteins and membrane contact sites and retargets phosphatidylethanolamine to the replication sites to build suitable membranous subcellular environment for replication [8,9, 18]. TBSV-induced major changes in the cell require energy, but how TBSV can obtain the energy to fuel these energy-demanding processes is not known. Glycolysis is an essential and highly conserved energy-producing pathway in the cytosol. Glucose is converted into pyruvate by a ten-enzyme catalyzed reaction that produces ATP and NADPH. The two ATP-generating enzymes in the glycolytic pathway are Pgk1 phosphoglycerate kinase and Cdc19 pyruvate kinase (PKM2/PKLR in humans). Pgk1 produces ATP from ADP and DPG using substrate-level phosphorylation in the cytosol [19]. Pgk1 is highly conserved and present in every organism. In humans, Pgk1 is also important for tumor growth and DNA replication/repair and mutations in Pgk1 are associated with myopathy, mental disorders and hemolytic anemia [20,21,22]. Pgk1 is also important for the activation of nucleotide-based anti-HIV drugs [23]. Previous genome-wide screens in yeast model host have identified several components of the glycolytic pathway that affected TBSV replication [24,25,26]. Proteomic screens also identified Pgk1 phosphoglycerate kinase, which was pulled-down with the tombusviral p92 RdRp and interacted with p33 replication protein [27,28]. Therefore, we decided to characterize the putative role of Pgk1 in tombusvirus replication in this work. We discovered that TBSV co-opts the glycolytic Pgk1 through direct interaction with the viral replication proteins. Altogether, we provide evidence that the ATP generated locally in the viral replication compartment by the co-opted Pgk1 is used to fuel the ATP requirement of the co-opted cellular Hsp70 molecular chaperones. The functions of co-opted Hsp70 are to facilitate the assembly of new viral replicase complexes and the activation of the viral RNA-dependent RNA polymerase. The local production of ATP by Pgk1 within the replication compartment could greatly facilitate the efficiency of Hsp70-driven replicase assembly by providing high ATP concentration within the replication compartment.
Among the surprises from high-throughput screens with TBSV and yeast model host were the identification of the glycolytic Pgk1 as a putative co-opted host factor in TBSV replication [24,25,26,29]. To demonstrate interaction between Pgk1 and the viral replication proteins, first we performed co-purification experiments from yeast replicating TBSV repRNA. The isolated membrane fraction was solubilized with detergent and the FLAG-tagged p33 and FLAG-p92pol replication proteins were affinity-purified, followed by Western-blot analysis. We found that the His6-tagged Pgk1 expressed from a plasmid was co-purified with the tombusvirus replicase (Fig 1A, lane 2). In contrast, Pgk1 was not co-purified with Flag-GFP control from yeast (S1 Fig). To examine if Pgk1 is a permanent component of the tombusvirus replicase, we halted the formation of new tombusvirus replicase complexes by shutting down p33 and p92pol expression and stopping ribosomal translation by adding cycloheximide [7]. FLAG-affinity-purification of the tombusvirus replicase from the membrane fraction of yeast at various time-points showed the rapid release of Pgk1 from the replicase (Fig 1B, lanes 3–4 versus 2). Thus, Pgk1 seems to be a temporarily co-opted factor in the tombusvirus replicase complex. To obtain additional evidence of subversion of Pgk1 into the tombusvirus replication compartment, we performed confocal laser microscopy experiments in yeast expressing GFP-p33 and BFP-Pgk1. Interestingly, we observed the robust recruitment of the cytosolic Pgk1 into the tombusvirus replication compartment, which was marked by Pex13-RFP peroxisomal marker protein (Fig 2A). Similar experiments with ectopically expressed BFP-NbPgk1 and p92-YFP showed the enrichment of Pgk1 in the tombusvirus replication compartment in Nicotiana benthamiana cells replicating the TBSV repRNA (Fig 2B). In addition, bimolecular fluorescence complementation (BiFC) assay further confirmed the recruitment of the glycolytic NbPgk1 via interaction with p33 into the peroxisomal TBSV replication compartment (Fig 2C). Similarly, we observed that the interaction between TBSV p92pol replication protein and the glycolytic NbPgk1 takes place in the large replication compartment, which was marked by RFP-SKL peroxisomal luminal marker protein (Fig 2C, lower panels). Therefore, we conclude that TBSV recruits the cytosolic Pgk1 through direct interaction with the viral replication proteins into the viral replication compartment in both yeast and plant cells. To test if the recruitment of Pgk1 into the viral replication compartment plays a role in TBSV replication, we have made a haploid yeast strain, in which the wt PGK1 gene was replaced with the HA-tagged PGK1. In addition, expression of HA-PGK1 was placed under the regulation of GALS promoter, thus allowing induction by the addition of galactose and repression by addition of glucose to the culture media [30]. Induction of TBSV repRNA replication in GALS: : PGK1 yeast with suppressed PGK1 expression showed ~4-fold reduction in TBSV repRNA replication (Fig 3A, lanes 7–9 versus 10–12, and S2 Fig). Expression of Pgk1 from a plasmid in GALS: : PGK1 yeast with suppressed Pgk1 expression from chromosome increased TBSV repRNA replication by ~2-fold in glucose-containing media, while did not increase it on the non-fermentable glycerol-media (Fig 3B, lanes 7–10). These data indicate that Pgk1 plays a pro-viral function during TBSV replication in yeast. Over-expression of either yeast Pgk1 (Fig 3C) or the N. benthamiana cytosolic NbPgk1 (Fig 3D) led to ~3-fold and a ~2. 5-fold increase, respectively, in TBSV repRNA accumulation in wt yeast, confirming a pro-viral function for Pgk1 in TBSV replication. Estimation of PGK1 mRNA levels in TBSV-infected N. benthamiana leaves indicated ~2-fold increased level of expression (Fig 4A). Similarly, Western blotting demonstrated a ~50% increase in accumulation of Pgk1 protein in yeast cells replicating TBSV repRNA (Fig 4B), suggesting that a tombusvirus can mildly induce the accumulation of a glycolytic enzyme in cells. To confirm the importance of Pgk1 in TBSV replication in a plant host, we knocked-down via virus-induced gene silencing (VIGS) the expression of the cytosolic Pgk1 in N. benthamiana and the Pgk1-silenced leaves were inoculated with the full-length infectious TBSV. TBSV genomic RNA accumulation was inhibited by ~4-fold in these Pgk1 knock-down plants (Fig 4C). Inoculation of Pgk1-silenced N. benthamiana with the similar peroxisome-replicating cucumber necrosis virus (CNV) (S3 Fig) and the mitochondrial-replicating carnation Italian ringspot virus (CIRV) showed a ~6-fold decrease in the accumulation of genomic RNAs (Fig 4D and 4E). Thus, the cytosolic Pgk1 is a critical pro-viral host factor for various tombusviruses in a plant host. Over-expression of NbPgk1 in N. benthamiana replicating TBSV RNA showed a close to 3-fold enhanced TBSV accumulation (Fig 4F), suggesting that Pgk1 level affects tombusvirus replication. Altogether, these data have confirmed the pro-viral function of Pgk1 in tombusvirus replication in plants. Based on the above results, we assumed that the co-opted Pgk1 likely generates ATP within the viral replication compartment. Therefore, we estimated the local ATP level within the replication compartment by using the p33 replication protein tagged with ATeam, a cellular ATP-sensor module. ATeam module can measure ATP levels via FRET due to the conformational change in the enhanced ATP-binding domain of the bacterial ATP synthase upon binding to ATP without ATP hydrolysis (Fig 5A) [31]. In the ATP-bound form, the ATP-sensor module brings the CFP and YFP fluorescent tags into proximity, increasing FRET, which can be detected by confocal laser microscopy. In the ATP-free stage, the extended conformation of the ATP-sensor module places CFP and YFP at distal position, leading to low FRET signal. The ATeam-tagged p33 localizes to the aggregated peroxisomes that represent the sites of replication (Fig 5B). We found by intracellular expression of p33-ATeam that the local ATP level within the TBSV replication compartment was decreased by ~6-fold in Pgk1 knock-down in N. benthamiana leaves in comparison with control leaves (Fig 5B). Interestingly, the mitochondria-replicating CIRV also accumulated a ~10-fold lower level ATP within the replication compartment in Pgk1 knock-down leaves than in control leaves (Fig 5C). Therefore, we conclude that Pgk1 is recruited to the sites of tombusvirus replication to generate high concentration of ATP within the replication compartment. To decipher the function of the co-opted Pgk1 during TBSV replication, we tested viral RNA levels in yeast replicating TBSV repRNA. Down-regulation of Pgk1 expression reduced both (+) - and (-) RNA levels by ~4-fold (Fig 6A). These data suggest that Pgk1 likely plays an early role, prior to viral RNA synthesis, possibly during replicase assembly steps. To test this scenario, we performed various in vitro assays. First, to further examine if Pgk1 was required for viral RNA replication, we obtained cell-free extracts (CFE) from wt yeast and GALS: : PGK1 yeast with suppressed Pgk1 expression to reconstitute the TBSV replicase in vitro based on purified recombinant TBSV proteins. The yeast CFE-based assay, which supports a single full cycle of viral RNA replication [32], showed ~7-to-10-fold reduced repRNA production when low level of Pgk1 was produced in GALS: : PGK1 yeast (Fig 6B, lanes 3 versus 1 and 4, also S4 Fig). Interestingly, both (+) RNA product (Fig 6B) and dsRNA replication intermediate obtained with CFE from GALS: : PGK1 yeast with depleted Pgk1 in comparison with the wt yeast CFE were decreased by~7-fold (Fig 6C, lanes 7–8 versus 3–4). Therefore, the in vitro replication results support the important role of Pgk1 in viral RNA replication. We also obtained purified replicase preparations from GALS: : PGK1 yeast with suppressed Pgk1 expression versus comparable preparations from wt yeast, which were programmed with viral RNA transcripts [33]. In spite of having comparable amounts of replication proteins, the purified replicase preparations from Pgk1 depleted yeast showed ~3-fold reduced activity in comparison with the similar preparations obtained from wt yeast (Fig 6D). Since the activity of the purified replicase preparations mostly depends on the efficiency of replicase assembly and replication protein activation, we compared the efficiency of p92pol activation by the supernatant fraction of CFEs prepared from Pgk1 depleted or wt yeasts. These assay revealed ~3-fold less RdRp activity of the purified recombinant p92pol mutant by the preparation from Pgk1 depleted than from wt yeasts (compare lanes 1–2 and 5–6 in Fig 6E). The major active component of the supernatant fraction of CFEs is Hsp70 molecular chaperone, which is essential for the activation of p92pol RdRp in vitro [29]. Accordingly, addition of purified Ssa1 (yeast Hsp70) to the WT fraction has increased the efficiency of p92pol RdRp mutant activation by ~5. 5-fold, whereas the comparable assay with Ssa1 involving the preparation from Pgk1 depleted yeast showed only ~2-fold enhancement (compare lanes 3–4 and 7–8 in Fig 6E). Therefore, the in vitro results from multiple assays support the model that Pgk1 has a major role in tombusvirus RNA replication and Pgk1 likely fuels the energy requirements of co-opted Hsp70 molecular chaperones to facilitate the efficient replicase assembly and replication protein activation within the viral replication compartment. TBSV co-opts Vps4p AAA+ ATPase, which is an ATP-dependent host protein, into the viral replicase complex to facilitate replicase assembly [7]. Vps4p is an ESCRT protein involved in membrane deformation, which is necessary for formation of TBSV-induced individual spherules (vesicle-like structures) supporting viral RNA replication [7]. The spherule formation protects the viral dsRNA replication intermediate from the innate RNAi machinery during infection [34]. To test if Pgk1-produced ATP might support Vps4p ATPase function during replicase assembly, we utilized a recently developed intracellular probe based on a reconstituted RNAi system in yeast (Saccharomyces cerevisiae), which lacks the RNAi machinery. The reconstituted RNAi machinery from S. castellii, which consists of the two-component DCR1 and AGO1 genes [35], is a simple, and easily tractable system [34]. We found that the induction of RNAi activity had only slightly more inhibitory effect on TBSV RNA accumulation in Pgk1-depleted yeast than in wt yeast (Fig 6F). Based on these results, the structure of the assembled tombusvirus replicase in Pgk1 depleted yeast might not be different from those assembled in wt yeast. Therefore, likely the co-opted ESCRT machinery and the Vps4p ATPase activity might not be affected when Pgk1 is depleted. Alternatively, the residual Pgk1 still present in GALS: : PGK1 yeast could provide enough ATP for Vps4p ATPase activity during the replicase assembly process.
TBSV replication, similar to other (+) RNA viruses, is a very intensive and robust process that likely depends on consumption of a large amount of ATP. By recruiting the glycolytic Pgk1 into the virus replication compartment, Pgk1 could efficiently supply ATP to facilitate various steps in the replication process, including the assembly of the viral replicase complex. Accordingly, we show that TBSV actively recruits the cytosolic Pgk1 to the sites of viral replication through direct interaction with the viral replication proteins in both yeast and plant cells. It seems that a portion of recruited Pgk1 is getting released from the sites of replication when the formation of new viral replicases is halted via inhibition of translation. This observation suggests that the ATP produced by the co-opted Pgk1 is likely used up locally to fuel early steps in virus replication, such as viral replicase assembly. The proposed role of the co-opted Pgk1 during the early steps in virus replication is further supported by additional observations. For example, down regulation of Pgk1 level in yeast affected both (-) and (+) -strand RNA levels and also decreased the in vitro activity of the purified replicase preparations obtained from yeast. Moreover, the in vitro assembly of functional TBSV replicase in CFEs from yeast with depleted Pgk1 level was inefficient, resulting in reduced production of both dsRNA intermediate and new (+) RNA progeny. All these results point at deficiency in viral replicase assembly when Pgk1 is depleted. We have shown previously that the assembly of the tombusvirus replicase requires the co-opted cellular Hsp70 that uses ATP for its molecular chaperone function [32,36,37]. Based on the above observations, we propose that the ATP generated by the recruited Pgk1 serves the energy need of co-opted Hsp70 to drive efficient replicase complex assembly within the elaborate replication compartment (Fig 6G). In addition, we obtained evidence that the functional activation of the TBSV p92 RdRp, which also requires Hsp70 molecular chaperone, depends on the ATP generated by the recruited Pgk1. The initially inactive p92 RdRp becomes functional during replicase assembly in the presence of p33 replication protein, the viral (+) RNA, ER membrane and the ATP-dependent Hsp70 [27,29]. In a simplified RdRp activation assay, we found that CFEs obtained from yeast with depleted Pgk1 were inefficient in promoting p92 RdRp activity. The addition of purified recombinant Hsp70 to the above assay could stimulate p92 RdRp activity to lesser extent in case of depleted Pgk1 than in the presence of wt yeast CFEs. Therefore, we propose that a major function of the Pgk1-generated ATP is to provide fuel to the co-opted Hsp70 during p92 RdRp activation and the assembly of viral replicase complexes (Fig 6G). The advantage of co-opting Pgk1 to the replicase complex could be that the energy hungry virus replicase assembly process does not need to intensively compete with cellular processes for access to plentiful ATP. Also, local production of ATP within the replication compartment could greatly facilitate the efficiency of Hsp70-driven replicase assembly by providing high ATP concentration within the replication compartment. There is previous evidence that Pgk1 could provide ATP to enhance the chaperone activity of Hsp90 that leads to multistress resistance [21]. In addition to Hsp70 molecular chaperone, TBSV also co-opts additional ATP-dependent host proteins into the viral replicase complex, including Vps4p AAA+ ATPase, involved in replicase assembly [7], and two DEAD-box helicases [38,39]. Using a reconstituted RNAi-based molecular probe, we found only a slightly less RNA protection in Pgk1-depleted yeast than in wt yeast (Fig 6F), which is in contrast with the poor TBSV RNA protection provided by VRCs assembled in the absence of ESCRT components [34]. Therefore, we propose that Vps4p AAA ATPase still functions efficiently enough in VRC assembly in yeast with depleted Pgk1 and ATP levels. On the other hand, the DEAD-box helicases, however, selectively affect (+) -strand synthesis, while Pgk1 affects both (-) and (+) -strand synthesis in CFE-based assay and in yeast to a similar extent. TBSV also affects the actin network, which requires ATP for functions, but it is currently on open question if the co-opted Pgk1 supplies ATP for the subverted actin network. Therefore, based on the collected data, we propose that the ATP-generated by Pgk1 within the replication compartment is primarily used by tombusviruses to provide ATP for the co-opted Hsp70 chaperone to support efficient assembly of the numerous viral replicase complexes. Similar to our current findings, cancer cells also use glycolysis to efficiently generate ATP, even in the presence of oxygen [22]. Up-regulation of glycolytic machinery promotes rapidly proliferating cancer cells survival. Up-regulation of glycolytic pathway also takes place in effector T cells and it is required for immune cells activation [20,40]. Apparently, the relatively inefficient glycolytic pathway can provide enough ATP when up-regulated in these cells. Several viruses are also known to reprogram the glycolytic pathway during infections based on metabolomic profiling [41,42,43], which unraveled enhanced glucose uptake into the infected cells. The hexokinase activity is increased in hepatitis C virus (HCV) or Dengue virus-infected cells [41,44]. ATP was shown to accumulate at the sites of HCV replication, likely to satisfy the energy demand of virus replication [45]. Whereas TBSV exploits glycolytic enzymes, such as PGK1 (this work) and Glyceraldehyde 3-phosphate dehydrogenase (GPDH, called Tdh2p and Tdh3p in yeast) for pro-viral functions [13,15], replication of a plant potexvirus is inhibited by GAPDH due to its RNA-binding function [46]. Altogether, metabolic reprogramming of the glycolytic pathway might be a widespread phenomenon during various viral infections. In addition to its role in ATP production during glycolysis, Pgk1 seems to have additional roles in virus replication. For example, Sendai virus, a negative-strand RNA virus, co-opts Pgk1 to promote viral mRNA synthesis, albeit its enzymatic activity is not required for the stimulation of RNA synthesis [47]. A partial/recessive resistance in Arabidopsis against potyviruses is due to a single amino acid mutation in the conserved N-terminal portion of the chloroplast phosphoglycerate kinase (cPGK2), which is a nuclear gene encoding cPgk2 targeted to the chloroplast [48,49]. In addition, cPgk1 was found to bind to the 3’UTR of the viral RNA and it is involved in the localization of potexvirus RNA to the chloroplasts, which is important for virus accumulation [50,51].
TBSV RNA replication in yeasts and plants was analyzed after total RNA extraction with Northern blot analyses as described previously [52]. Briefly, BY4741 and GalS: : PGK yeast strains were co-transformed with pHisGBK-CUP1- Hisp33/ADH-DI-72 and pGAD-CUP1-His92 [24]. The transformed yeast strains were grown at 23°C in SC-HL− (synthetic complete medium without histidine and leucine) media supplemented with 2% raffinose with or without 2% galactose and BCS for 24 h at 23°C. Then, yeast cultures were re-suspended in SC-HL− medium supplemented with 2% raffinose and with or without 2% galactose and 50 μM CuSO4. Yeasts were grown at 23°C for 16 h before being collected for total RNA extraction. Yeast cells were grown as described above for Northern analysis. Total proteins were isolated by the NaOH method as described previously [53]. The total protein samples were analyzed by SDS-PAGE and Western blotting with anti-His and anti-PGK antibodies, followed by alkaline phosphatase-conjugated anti-mouse secondary antibody (Sigma) as described previously [54]. To examine the subcellular localization of Pgk1 in plants, N. benthamiana leaves were co-infiltrated with Agrobacterium carrying plasmids pGD-p92-YFP and pGD-BFP-PGK (OD600 of 0. 5, each) together with pGD-35S: : p19, pGD-DI-72 and pGD-p33. After 48 h, the agroinfiltrated leaves were subjected to confocal microscopy (Olympus America FV1000) using 405 nm laser for BFP and 488 nm laser for YFP. Images were taken successively and merged using Olympus FLUOVIEW 1. 5 software. To identify interactions between NbPgk1 and TBSV p33 replication proteins, the plasmids pGD-p33-cYFP, pGD-nYFP-PGK and pGD-nYFP-MBP were transformed to Agrobacterium strain C58C1. These Agrobacterium transformants were used to co-agroinfiltrate the leaves of four weeks-old N. benthamiana plants. Transformed leaves were subjected to confocal laser microscopy after 48 h using Olympus FV1000 microscope as described previously [13]. To detect the ATP levels within the tombusvirus replication compartments in plant cells, in the case of TBSV, PGK-silenced plants or control plants were agroinfiltrated with plasmids pGD-p33-ATeamYEMK, pGD-DI-72, pGD-35S: : RFP-SKL, pGD-35S: : p19 and pGD-p92. In the case of CIRV, the leaves were agroinfiltrated with pGD-p36-ATeamYEMK, pGD-35S: : AtTim21-RFP, pGD-35S: : p19, pGD-DI-72 and pGD-p95. The images were taken at 2. 5 or 3. 5 days post-agroinfiltration and analyzed with the method described previously [31]. Confocal FRET images were obtained with an Olympus FV1000 microscope (Olympus America). Cells were excited by a 405 nm laser diode, and CFP and Venus were detected at 480–500 nm and 515–615 nm wavelength ranges, respectively. Each YFP/CFP ratio was calculated by dividing pixel-by-pixel a Venus image with a CFP image using Olympus FLUOVIEW software or ImageJ software. Additional standard experimental procedures are presented in the supporting information S1 Text. | Positive-strand (+) RNA viruses build replication organelles with the help of many co-opted cellular factors and by usurping cellular metabolites and energy, which frequently lead to disease states in plants, animals and humans. The authors discovered that tomato bushy stunt virus (TBSV) co-opts the glycolytic ATP-generating Pgk1 phosphoglycerate kinase to facilitate the intracellular assembly of new viral replicase complexes. The ATP generated by co-opted Pgk1 within the replication compartment is used to fuel the ATP requirement of co-opted cellular heat shock protein 70 chaperone, which is essential for the assembly of new TBSV replicase complexes. Direct recruitment of Pgk1 by TBSV might provide easy access to ATP. Metabolic reprogramming of the glycolytic pathway might be a widespread phenomenon during various viral infections. | Abstract
Introduction
Results
Discussion
Materials and methods | plant anatomy
protein interactions
molecular probe techniques
microbiology
fluorophotometry
northern blot
fungi
plant science
model organisms
experimental organism systems
molecular biology techniques
gel electrophoresis
research and analysis methods
saccharomyces
electrophoretic techniques
spectrum analysis techniques
electrophoretic blotting
fluorescence resonance energy transfer
proteins
viral replication complex
leaves
viral replication
molecular biology
spectrophotometry
yeast
biochemistry
eukaryota
virology
biology and life sciences
yeast and fungal models
saccharomyces cerevisiae
organisms | 2017 | Co-opting ATP-generating glycolytic enzyme PGK1 phosphoglycerate kinase facilitates the assembly of viral replicase complexes | 7,030 | 203 |
High throughput mRNA expression profiling can be used to characterize the response of cell culture models to perturbations such as pharmacologic modulators and genetic perturbations. As profiling campaigns expand in scope, it is important to homogenize, summarize, and analyze the resulting data in a manner that captures significant biological signals in spite of various noise sources such as batch effects and stochastic variation. We used the L1000 platform for large-scale profiling of 978 representative genes across thousands of compound treatments. Here, a method is described that uses deep learning techniques to convert the expression changes of the landmark genes into a perturbation barcode that reveals important features of the underlying data, performing better than the raw data in revealing important biological insights. The barcode captures compound structure and target information, and predicts a compound’s high throughput screening promiscuity, to a higher degree than the original data measurements, indicating that the approach uncovers underlying factors of the expression data that are otherwise entangled or masked by noise. Furthermore, we demonstrate that visualizations derived from the perturbation barcode can be used to more sensitively assign functions to unknown compounds through a guilt-by-association approach, which we use to predict and experimentally validate the activity of compounds on the MAPK pathway. The demonstrated application of deep metric learning to large-scale chemical genetics projects highlights the utility of this and related approaches to the extraction of insights and testable hypotheses from big, sometimes noisy data.
Pharmacology is generally explored in a linear and iterative manner, starting from the observation of an activity, that is then optimized for selectivity and potency via various methods, and subsequently tested in preclinical and randomized controlled clinical studies for safety and efficacy. In general, the activity of compounds on targets other than the intended one (s) is limited, though such ‘off-target’ activities may lead to adverse events. The detection of polypharmacology is desired and could be facilitated by comprehensive profiling of drug candidates on a general phenotyping platform, i. e. a consistent method to broadly assess the phenotypic effects of compound treatment. Such a platform could be used in order to uncover unexpected phenotypic signals beyond the originally identified mechanism-based phenotype. The inference of mode of action or mode of toxicity from gene expression changes has been extensively studied in the past [1–3]. Different methods using only gene expression, or expression in the context of prior biological knowledge, have shown success in identifying the efficacy and toxicity targets of compounds [4,5]. Phenotypic screening is considered by some to be a return to pharmacology’s roots [6], and by others a new discipline, that will need to prove itself over the coming decades [7]. In either case, there is increased interest in applying cell-based pathway or phenotype screens to identify unknown/unexpected targets along with tool compounds that can be used for target validation, and potentially as starting points for drug discovery. One major hurdle in phenotypic screening approach versus a target-based approach lies in the identification of the target (s) of molecules that show an activity in cell-based (or organismal) assays [8]. A general phenotyping platform could be used to infer mode of action of unknown compounds based on induced expression profiles’ similarity to those of annotated compounds. Such data can also in some cases be used to propose new indications for known molecules [1]. Lastly, a general phenotyping platform will allow one to monitor compounds through their maturation and optimization in order to prioritize series based on selectivity and to quickly identify potential polypharmacology and safety warning signals [9]. We suggest that mRNA is a promising analyte for a general phenotyping platform, although the domain of applicability remains to be fully understood. Whereas gene expression changes are often distal to signaling and metabolic pathways that drug discovery aims to modulate, most perturbations of cellular pathways eventually lead to the nucleus [10], and to transcriptional changes that propagate, amplify, or compensate for the immediate effects of a perturbation [11]. mRNA also has the beneficial property that its measurement is fairly easy to generalize, such that any set of target sequences can be measured quantitatively and in parallel [12]. Thus, a potentially broadly useful general phenotyping platform would quantitate mRNA, be medium to high throughput, be affordable to apply to thousands of samples, and produce highly reproducible data. The L1000 platform [13] has the potential to be just such a general phenotyping platform, one that can be used in various stages of drug discovery, including target identification and validation, hit-to-lead, lead optimization, as well as safety assessment and repurposing. 978 genes were selected to be representative of the expression of the remainder of the transcriptome [14], and the platform is used to capture the transcriptional phenotypes using this reduced set of ‘landmark’ genes. The high throughput and relatively low cost of the bead array based implementation permits comprehensive application to large numbers of perturbations, be they different compounds, different cellular contexts, titrations, compound series, etc. However, if such a platform is applied for large sets of perturbations, spanning years of different project stages and various programs, then data analysis, and particularly homogenization, become important. Large scale expression profiling projects such as the Connectivity Map [1] and applications presented herein have to contend with day-to-day variation in cellular responses. Indeed, batch effects were previously considered a nuisance that was dealt with using robust rank-based statistics (connectivity score, [1]), and via use of biologically motivated data summaries such as Gene Set Enrichment Analysis [15,16]. It is not clear that such nonparametric approaches, which depend on prior knowledge (biological pathways or previous expression experiments), yield the highest possible sensitivity and specificity for downstream analyses. Herein we present a novel method of representing the expression profiles of the L1000 platform as a short binary barcode. The approach starts by training a deep model that learns to distinguish replicate from nonreplicate profiles. The internal state of the model thus learned demonstrated properties previously ascribed to deep neural network models, namely a hierarchical representation that captures the regularities and underlying structure of the input data [17,18]. The internal state represents a robust, abstracted representation of the data, one that captures inherent aspects of the biology such as similarity of compound targets and pleiotropy. We go on to apply this framework to prospectively predict the targets of compounds based on the transformation of their induced expression profiles.
We first introduce the data generation process and highlight the need for improved analyses. The disparity between replicate profiles motivates the development of a metric learning approach that improves upon the current state of the art for this type of data analysis. The experimental approach used in this work is depicted in Fig 1A. Cells can be treated, lysed, and measured in medium throughput (384-well) formats. The use of commodity consumables, reagents, and technology permits relatively low cost profiling of hundreds to thousands of compound-treated samples at a time [13]. Reader intensity measurements are normalized using control samples and genes. Next, each perturbation’s expression is compared to an internal (within-batch) negative control. Twelve to sixteen vehicle (DMSO) controls are measured on each 384-well assay plate, and the expression of the remainder of samples is standardized to the average (median), and scale of variation (median absolute deviation), of these vehicle controls to yield robust z-scores, henceforth referred to as z-scores for brevity. Large scale gene expression profiling is known to have significant ‘batch’ effects that confound interpretation [19]. These effects are a result of differences in cell response from day to day. The effects can be controlled for by performing replicate experiments, but the level of replication required to eliminate the batch effects is not practical for high throughput experimentation, which requires knowledge to be extracted from 1–2 replicates. Additional approaches to reduce the influence of experimental batch include normalization of data within a batch such that all perturbations are compared to an internal negative control in the same batch. In spite of measures to control batch effects, it is nevertheless observed that there is variability in cell response that makes it difficult to interpret data obtained on different days [20]. As an example, for the dataset described below, we can assess the similarity of biological replicates’ gene expression profiles using Euclidean distance of the normalized expression changes versus control. Looking at the similarity of each sample profile to every other in the dataset, the median rank of a sample’s biological replicate is approximately the 3rd percentile. While this ranking can be considered an enrichment (versus a null hypothesis of ~50th percentile), the 3rd percentile in our dataset nevertheless implies that >200 treatment profiles are more similar to a given treatment than a sample that was treated identically on a different day. As mentioned above, the z-score data is confounded by batch effects that reduce the apparent similarity of expression responses deriving from identical treatments. In order to improve on the performance of the z-scores, another strategy that was explored is that of gene set enrichment analysis (GSEA, [16]). GSEA is a technique used to extract interpretable information from expression profiles, and is useful for placing the up- and down-regulation of genes into a biological context (e. g. biological pathway, disease state). In addition, GSEA profiles (i. e. enrichment scores of a sample across many different gene sets), can be thought of as a means of averaging, and potentially making more robust, the expression profiles by using biological context as a prior. Thus, it might be expected [21] that GSEA profiles might be more reproducible, and more predictive, than raw expression data. Indeed, when one looks at the correlation of pairs of samples treated with the same compound, one sees that GSEA profiles show higher concordance than z-score profiles day-to-day and cell-to-cell (Fig A in S1 Text). In the previously mentioned measure of average rank of replicates, GSEA yields a median rank of ~1%, with ~70 profiles more similar to a given sample than its biological replicate. Since there is room for improvement over both of the previously described methods, we sought to evaluate alternate strategies to increase the sensitivity and specificity of interpretations and predictions derived from relatively large data sets that we are currently generating on the L1000 platform. In particular, as a first step, we sought to increase the concordance of samples that should exhibit the same phenotypic response, while separating them from those that should be distinct. Using the similarity of biological replicates as a gold standard, the goal was to recast the data in a way that maximizes the similarity of replicates in contrast to non-replicate samples. This can be thought of as a metric learning problem, whereby one learns a new measure of distance or similarity between points, one that distinguished replicates from non-replicates. Initial experimentation with linear and kernel methods [22] failed to yield significant improvements over the raw data, thus deep neural network models were next evaluated. Deep learning is an extension of decades-old artificial neural networks, and has recently shown impressive performance on a number of image and speech recognition tasks [23,24]. As the properties and capabilities of these models have begun to be elucidated, it is becoming more feasible and reproducible to create highly accurate multilayer neural network models. In addition, advances in model design have been complemented by increased computational power and larger datasets that enable the training of comprehensive, expressive models, whose generalization performance is facilitated by regularization [17,18]. In addition to providing a flexible model that is able to fit complicated datasets, deep networks have the additional advantage of learning a hierarchical representation of the data, whereby lower layers of the model learn to represent the fine-grained detail of the input data, and higher layers represent increasing layers of abstraction. Thus, inspecting the higher layers of a model has the potential of revealing the underlying high-level structure of the data, with reduced sensitivity to noise and contextual details. We thus explored a multilayered network with the dual goal of developing a suitable distance metric, and also developing a robust data representation for downstream analyses. Most machine learning methods consider a single input sample at a time, though each sample is generally represented by multiple variables, or features. The goal of this effort was to learn not about individual samples, but about the comparison of pairs of inputs in order to learn a new metric that better captures what is known about the data. Thus, a method was chosen that permits comparative learning in a neural network framework. We implemented a version of a siamese neural network [25,26] to perform metric learning on a dataset of L1000 data (Fig 2). A number of model architecture properties (also known as hyperparameters) need to be set prior to the optimization of model weights of a neural network by gradient descent. These include number of layers, layer sizes, activations of layers, regularization types and weights, and cost type and any parameters. We undertook a greedy manual search of hyperparemters, optimizing the validation set accuracy for predicting replicate vs. nonreplicate pairs. Significantly, other characteristics of the model and its representation, such as those described below and in Table 1 (rows 2–5), were not measured during the model selection process, and only replicate identification accuracy was used to select a final model. Among over 60 models tested, we settled on the architecture described below since it provided the best classification accuracy of replicates/nonreplicates on validation data. The final model that emerged from optimization of the neural network architecture performs the following: a pair of 978-dimensional z-score vectors representing two different expression profiles is given to the network as input, the data is transformed by two layers of noisy sigmoid layers, and then the representations of the members of the pair are compared to each other by calculating a Euclidean distance. A margin loss is calculated based on the distance versus the known target status of the pair (i. e. replicates or not replicates). The cost is used to train the network via backpropagation of the cost gradients [27]. While siamese networks have been previously depicted as a linked pair of networks transforming the pair of inputs for comparison (see [25,26]), in practice we found it simpler to implement the siamese network model as a single network that processes pairs of samples as adjacent inputs (Fig 2). In this architecture, the paired input is provided as two consecutive vectors to one network that processes the elements of the pair through the aforementioned hidden layers and then calculates the cost. The network was trained on 80% of the compounds, using a validation held-out set of 10% to select model architecture and detailed configuration (hyperparameters), and a final test set of 10% was used to estimate generalization performance of the selected model. The trained model was able to correctly categorize 97% of test pairs as replicates/non-replicates (F1 = 0. 87, [28]). As a baseline, a model that randomly samples from the empirical distribution of replicates/non-replicates has an accuracy of 56% (F1 = 0. 33). A nontrivial reference for comparison would be a model that thresholds the gene expression changes for each sample to generate a binary 978-dimensional representation. Depending on whether the threshold is shared among all genes or is gene specific, and on whether the threshold is applied to raw expression changes or absolute changes, a variety of models can be created. The best performance in terms of accuracy on the validation data was observed for a model that uses gene-specific thresholding on the original scale, with thresholds learned from the training data set, this model gave a test set accuracy of 67% (F1 = 0. 66). Another nontrivial model one could apply would use random projections of the data as is done in locality sensitive hashing [29]. Such a model, using 100 random projections, showed an accuracy of 33% (F1 = 0. 50). Optimizing the projections to maximize training set discrimination yields a linear model. A 100-dimensional linear model yielded an accuracy of 71% (F1 = 0. 47). These results show that the learned deep model is superior in classification accuracy to trivial and nontrivial comparator models. Having demonstrated that it is possible to recast the data in a way that captures the similarity of identically treated samples, we set out to determine if the learned model can be used to improve interpretability of large scale expression profiling campaigns. Since the model can separate replicates from nonreplicates, and since the model output is a function of a simple Euclidean distance calculated on the internal representation (hidden layer activations), we reasoned that this internal representation captures the discriminatory power learned by the model. We explored whether this internal representation has additional useful properties, as for example do the internal word representations of skip-gram language models [30] to represent semantic and syntactic relationships. The internal representation was extracted from the learned model by using the activation of 100-dimensional second hidden layer as the learned feature vector of each data point (Fig 2B). This representation is a short, fixed-length, almost binary representation that captures expression profile changes. In order to facilitate computation of similarities and hash-based lookup, the almost-binary representation (~95% of activations >0. 9 or < 0. 1), was thresholded to be exactly binary (i. e. values ≥ 0. 5 were set to 1, rest to 0). This binary representation of cellular gene expression profile changes in response to treatments is referred to as a ‘perturbation barcode. ’ Thus the model yields a new encoding of the original data, a representation that was designed only to increase the similarity of biological replicates relative to non-replicates. Information is necessarily lost in reducing 978-dimensional continuous data to 100-dimensional binary representations. The question that remains is whether biologically interesting aspects of the data are retained in the simplified encoding. Thus, we evaluated whether the model described herein learns useful, generalizable aspects related to compound effects on cells as a byproduct of learning to identify replicates. The perturbation barcodes were assessed for their ability to identify similarities between compounds in a variety of contexts (Table 1). Furthermore, we directly compared the performance of the barcodes to the minimally processed z-score data and the gene set enrichment score profiles. The median rank (out of 7573) of each sample’s biological replicate is shown in Table 1. For each sample, the similarity of all other samples’ profiles is ranked, and the median rank across samples of the biological replicate is reported. For the z-score data this value is 225, indicating that, on average, there are 223 profiles more similar to a given sample (itself ranked #1), than the sample’s biological replicate. The median rank for GSEA score data is 72, and for perturbation barcodes it is 24. The result on the perturbation barcodes demonstrates that the training objective was met: samples that are replicates are more similar under the features derived from the metric learning than they are in the original data space.
We present here a straightforward way to extract features from gene expression data that are in many ways more expressive and robust than the original gene expression changes. While training of a deep network is a somewhat specialized process, once an adequate model is trained, additional samples can efficiently be converted to perturbation barcodes with no (or little), further optimization needed, as evidenced by the simple transfer of the model to external (LINCS) data. Others have proposed novel methods for analyzing L1000 data. In particular, Liu et al [39] presented a pipeline that incorporates raw intensity processing and gene assignment through gene set enrichment and production of features informed by protein interaction data. While this type of biological annotation is potentially useful, our proposed perturbation barcode is self-contained, and does not rely on noisy and incomplete biological databases for utility. In addition, the improvements in performance of the learned features compares favorably with other approaches used in the field, including GSEA. Finally, it was shown that the perturbation barcodes can be used to meaningfully predict pathway modulation activity of compounds prospectively. This ability is of value in uncovering unknown modes of action (either primary or secondary) of compounds of interest, and the fact that it can be read directly from a visualization, shows the potential of the approach to simplify and enhance hypothesis generation from big data. We demonstrate that similarity in the barcode space is indicative of more similarity in compound target and compound structure, activity across biological assays, and predictivity of biological action. It should be pointed out that even with the improved data representation, the level of concordance between the gene expression data and reference data (compound structural features, target annotation), remains imperfect. There are a number of potential explanations for this. Firstly, the gene expression experiments were conducted in one or two cell lines per compound, and for some compounds the (known or unkown) efficacy targets are not expressed, thus limiting the potential correlation to target annotation. Additionally, many compounds were unoptimized screening hits or lightly characterized tools, and thus likely have significant polypharmacology, only some of which is known, and only some of which is consistent within chemotypes due to ‘activity cliffs. ’ Next, target annotation is incomplete, so an apparent false positive association between a compound and an activity may in fact be a genuine connection that has not yet been discovered. Lastly, there is little doubt that obtaining higher data density (more doses, replicates, time points, cell lines, structural neighbors), would complement the sparse dataset explored herein, and in doing so, possibly tie more compounds to benchmark annotations than was possible with the current data. We made the somewhat unexpected observation that models of general compound properties like promiscuity can be better built using the learned features than the data that the features were learned from, indicating that there are positive side effects of the compressive data encoding. We attribute this phenomenon to the denoising property of the learned features, or equivalently, extraction of robust underlying biological factors from the corrupted versions observed in experimentation (due to measurement error, batch effects, and stochastic variation in response). We anticipate that elaborations of approaches such as this one will be fruitful for other applications in biological and chemical domains, as they have been in artificial intelligence. A more ambitious goal of algorithmic design of molecules may be based on the combination of phenotypic information, quantitative structure-activity relationships, and pharmacokinetic/pharmacodynamic models with generative models for chemical structures. In the meantime, the direct application of the described metric learning-based representation technique to other high volume, high dimensional data has the potential to significantly reduce the effects of noise and to improve interpretability and the quality of generated hypotheses.
Doses for treatment were chosen using the following hierarchy of approaches: for compounds with known biological activity and toxicity, the dose with the highest therapeutic window was selected. For compounds with only biological activity, the EC80 was used. For compounds without relevant cell-based assay activity, the highest non-toxic dose was used, up to 20μM. Two cell lines were used: PC3 and ME180. A number of compounds were profiled in both cell lines, the rest were profiled in just one of the two. Most compounds were profiled as biological replicates by performing identical treatments and lysis on separate days, some compounds were only represented by a single instance. Compound treatment time was 6 hours, after which cells were lysed, frozen, and shipped to Genometry for processing. A total of 3699 compounds were screened. Compounds were selected based on being tool bioactive compounds, active compounds for ongoing phenotypic screening programs, and compounds of interest for particular compound optimization programs. Compound profiling was performed in four independent campaigns over the course of 14 months. Initial data processing was conducted by Genometry using a standard pipeline. Briefly, Luminex intensity measurements were assessed for consistency of relative expression of control genes. Samples passing well and plate level thresholds were summarized by conversion of intensity data to calculated log2 Genechip-equivalent intensities and normalized based on control gene intensities. Finally, the data on each plate was standardized based on the median and median absolute deviation of the vehicle control samples to calculate z-scores for each gene. These z-scores were the input for the deep metric learning, and were also analyzed directly as a baseline for data interpretation. A metric learning cost module that takes consecutive pairs of training examples and calculates a distance between them, and thereby a cost, was implemented using the Pylearn2 framework [40]. Internal data representing L1000 data as z-scores of 7573 profiles of 3699 compounds was used. Train/validation/test datasets were made from the initial data by selecting respectively 80,10, and 10% of the initial 3. 7k compounds. Within a dataset, a sample of pairs of profiles representing biological replicates was used for positive examples (n = 40k), and in addition a sample of twice the number of pairs of samples that were not biological replicates was used as negative examples. Hyperparameters (number and type of layers, regularization, dropout), were tuned manually based on replicate/non-replicate prediction accuracy of the model on the validation dataset, and the model depicted in Fig 2 emerged as the best performing. In order to build a model that facilitates application to large scale profiling datasets, it was of interest to bias the model to produce a representation that could be used akin to locality sensitive hashing [41], or semantic hashing [42], whereby large collections of profiles can be queried for similar expression profiles without performing global similarity calculations (which become prohibitive as datasets increase in size). Hashing of L1000 data is the subject of ongoing work, and is not explored further here. Nevertheless, it was found a model designed to facilitate hash-based lookup (i. e. having saturated, nearly binary internal representations), gave a higher validation set accuracy than other models tested, so this is the model that was used for downstream analyses. The model uses an input layer of 978 z-scores, followed by two hidden layers of 400 and 100 units, using a noisy sigmoid nonlinearity: y=σ (Wx+b+N (0,0. 25I) ) (1) σ (z) = (1+e−z) −1 (2) Independent Gaussian noise with mean 0 and variance 0. 25 (N (0,0. 25I) ), is added to the weighted sum of inputs plus bias (Wx + b) before applying sigmoid nonlinearity σ. Noise was added to favor saturated (0/1) activations. Dropout [43] (p = 0. 5) and L1 weight decay (λ = 10−5–10−6) are also used in hidden layers for regularization. Finally, the cost layer uses a rectified loss function with a squared distance margin of 5: c=softplus (1−y (m−d2) ) (3) softplus (x) =ln (1+ex) (4) y={1 if pair are replicates −1 if pair are non-replicates (5) Here c is the cost, m is the margin, y is the training target (-1 for non-replicate, +1 for replicate), d2 is the squared Euclidean distance between hidden vectors. Theano [44] automatic differentiation is used to backpropagate the cost to the model weights (including regularization) via the pylearn2 [40] framework. The model parameters are optimized using the RMSprop [45] algorithm for adaptive minibatch stochastic gradient descent. The model described herein can be trained on the dataset described (120K pairs) in approximately 1 hour on a laptop computer with a 2. 5 GHz CPU. While the approach has not yet been applied to transcirptome-wide profiling, the nature of the stochastic gradient descent learning method allow it to be computationally tractable for such data, the only requirement is a large enough corpus (thousands of samples), of homogeneous full-transcriptome profiles. After training the model, the weights were extracted and used to generate the activation state of the last hidden layer for each input sample. While noise was added during training to regularize the model and encourage saturation, no noise was added during barcode generation in order to yield a deterministic transformation. Perturbation barcodes are generated by thresholding activations of the last hidden layer to create a binary representation. See ‘Availability of supporting material’ for software. For comparison, several baseline models were created to compare replicate recognition accuracy to that of the neural network model. A model that randomly emits 1/-1 with probabilities 0. 33/0. 67 is a naïve model that predicts replicates independent of input with a frequency that matches that of replicate pairs in the training data. A series of models that threshold the gene expression data to generate binary representations (978 dimensional) can also be made. The thesholding was applied to the original scale (marking those with expression greater than the threshold as 1,0 otherwise), or on the absolute value scale (where either upregulation or downregulation greater than the threshold result in 1,0 otherwise). The thresholds were either set to be equal for all genes or were selected separately for each gene. Some hashing approaches use random projections of the data, in order to determine if the learned neural model is superior to this approach, a 100 random projection model was tested for classification of pairs as replicates vs. nonreplicates. Finally, the random projections can be replaced with learned linear projections optimized to minimize classification error, also known as a linear model. The above models with tunable parameters (thresholds, projections), were learned using gradient descent, minimizing margin error on the training set, and tested on the test set. GSEA was conducted using 6034 gene sets derived from experimental gene signatures of up- and down-regulated genes curated by Nextbio (Santa Clara, CA). Enrichment was measured using a Wilcoxon test, and the resulting rank sum test z statistic was used as a score for a given gene set’s enrichment in a sample. Data downloaded from the NCBI GEO repository, accession GSE70138. The file GSE70138_Broad_LINCS_Level4_ZSVCINF_mlr12k_n78980x22268_2015-06-30. gct. gz was downloaded in December 2015. This dataset represents 273 compounds tested in 6-point dose response at two time points in 15 cell lines. The level 4 (z-score vs. vehicle control) data was utilized. The first 978 landmark (i. e. measured) gene measurements of the data matrix were analyzed. In order to train the perturbation barcode model on the LINCS data, a subset of the data was selected: 10 and 1 μM treatments at 24 hours. 80% of the data was used for training, and 20% for validation, and the same hyperparameters were used for the model (see section Deep Metric Learning). After the model was trained, the entire dataset was encoded with the perturbation barcodes learned from the training set. Targets for 149 of the compounds were found to be annotated in Chembl [46] or Metabase [47] databases, only those targets affected with potencies <1μM were considered. For each target, distances between profiles derived from distinct compounds sharing the target were compared to a sample not sharing the target via a t statistic. Compound structures were clustered by hierarchical clustering of ECFP4 fingerprints, and gene expression profiles were clustered by the clara [48] algorithm using cluster numbers (k) determined from optimal cuts of hierarchical clustering trees of samples of the datasets. The clusterings were compared via the Adjusted Rand Index. Crossvalidated support vector regression was performed with the caret package [49], clustering, and correlation analyses were performed in R [50]. For visualization, data, either z-scores or barcodes, were reduced to two dimensions using the t-distributed stochastic neighbor embedding (t-SNE [38]) algorithm (implemented in tsne [51] package for R). Data was visualized by plotting the t-SNE features in Spotfire (Tibco). Compound structures were clustered using in-house fragment-based descriptors [52], and hierarchical clustering with a threshold of Dice similarity>0. 6. Promiscuity was defined based on the frequency with which each compound was considered a ‘hit’ across HTS assays. Project-specific hit calling was used for each assay. Merck’s chemogenomic database, the Chemical Genetic Interaction Enterprise (CHEMGENIE) was used to annotate compounds with their known targets. CHEMGENIE contains harmonized data from external (e. g. , ChEMBL, Metabase, and PDB) sources as well as internal sources (e. g. , project team data, kinase profile panels, counterscreen panels, etc.) where compounds are represented as desalted InChIKeys and targets are represented with their Entrez Gene ID and/or UniProt Accession. All target-based dose response data was added to each compound, and the highest affinity target was selected as a representative target for each compound. HTS fingerprints were constructed as described in the past [37]. Briefly, the z-scores of primary HTS screens at Merck were calculated for all screens where > 1e6 compounds had been screened (344 total screens). The z-score for each screen was then stored as a vector for each compound. Any z-scores > 20 or <-20 was assigned a value of 20 or -20, respectively, so that artifactual values would not impact Pearson correlation calculations as significantly. The Pearson correlation of HTS-FPs was calculated between pairs of compounds by calculated the correlation of z-scores for assays that were in common between them (all other z-scores were ignored). 3471 compounds possessed HTS-FPs and were used in this analysis. t-SNE visualizations of z-scores representing the normalized L1000 data, and also of the 100-dimensional (nonthresholded) perturbation barcodes derived from the metric learning network were generated. The locations of known EGFR/MEK/MAPK inhibitors on these maps were plotted, and points (i. e. representing compound treatments), that were surrounded by these known actives were selected for testing from each of the two maps. Separately, in order to select actives from the full 978-dimensional z-score or 100-dimensional barcode space, the ten nearest neighbors (by Euclidean distance) of each of the seed compounds were identified, and available compounds were tested for activity. Compounds were tested in dose titration assay in 1536 well plate format using a previously optimized reporter gene assay. Cells (CellSensor ME-180 AP-1-bla, Life Technologies) were plated in the manufacturer’s recommended assay medium at 3000 cells/well in 9μl in black/clear tissue culture treated 1536 well plates (Greiner), and allowed to adhere overnight. Compounds were added from DMSO serial dilution plates (50μM maximum assay concentration, 8-point, 3-fold dilutions), using a 50nl pintool (GNF Systems), and the cells incubated 30 min in a tissue culture incubator. The cells were then stimulated with Epidermal Growth Factor (EGF, Life Technologies) at 10ng/ml final concentration by adding 1μl 10x stock, and allowed to respond for 5 hours in a tissue culture incubator. Beta-lactamase detection reagents (ToxBlazer, Life Technologies), were added per manufacturer’s instructions (2μl 6x mix), plates were incubated 2 hours at room temperature, and read using a bottom-reading multimode reader (Pherastar, BMG). Data was normalized to no-stimulation (100% inhibition), and stimulation + DMSO (0% inhibition), controls, and percent inhibition was plotted along with logistic regression curve fits using Spotfire (Tibco). Compounds were classified as active if they exceeded 50% inhibition of reporter activity without toxicity at concentrations below 10μM.
Code for performing deep metric learning and a demonstration of the analysis on the LINCS data are available at https: //github. com/matudor/siamese. | The effects of small molecules or biologics can be measured via their effect on cells’ gene expression profiles. Such experiments have been performed with small, focused sample sets for decades. Technological advances now permit this approach to be used on the scale of tens of thousands of samples per year. As datasets increase in size, their analysis becomes qualitatively more difficult due to experimental and biological noise and the fact that phenotypes are not distinct. We demonstrate that using tools developed for deep learning it is possible to generate ‘barcodes’ for expression experiments that can be used to simply, efficiently, and reproducibly represent the phenotypic effects of cell treatments as a string of 100 ones and zeroes. We find that this barcode does a better job of capturing the underlying biology than the original gene expression levels, and go on to show that it can be used to identify the targets of uncharacterized molecules. | Abstract
Introduction
Results
Discussion
Methods
Availability of supporting materials | chemical compounds
neural networks
neuroscience
optimization
mathematics
mapk signaling cascades
discrete mathematics
combinatorics
computer and information sciences
cluster compounds
gene expression
chemistry
signal transduction
permutation
data visualization
cell biology
phenotypes
genetics
biology and life sciences
physical sciences
cell signaling
signaling cascades | 2017 | Representing high throughput expression profiles via perturbation barcodes reveals compound targets | 8,395 | 194 |
Guided migrations of cells and developing axons along the dorso-ventral (D/V) and antero-posterior (A/P) body axes govern tissue patterning and neuronal connections. In C. elegans, as in vertebrates, D/V and A/P graded distributions of UNC-6/Netrin and Wnts, respectively, provide instructive polarity information to guide cells and axons migrating along these axes. By means of a comprehensive genetic analysis, we found that simultaneous loss of Wnt and Netrin signaling components reveals previously unknown and unexpected redundant roles for Wnt and Netrin signaling pathways in both D/V and A/P guidance of migrating cells and axons in C. elegans, as well as in processes essential for organ function and viability. Thus, in addition to providing polarity information for migration along the axis of their gradation, Wnts and Netrin are each able to guide migrations orthogonal to the axis of their gradation. Netrin signaling not only functions redundantly with some Wnts, but also counterbalances the effects of others to guide A/P migrations, while the involvement of Wnt signaling in D/V guidance identifies Wnt signaling as one of the long sought mechanisms that functions in parallel to Netrin signaling to promote D/V guidance of cells and axons. These findings provide new avenues for deciphering how A/P and D/V guidance signals are integrated within the cell to establish polarity in multiple biological processes, and implicate broader roles for Netrin and Wnt signaling - roles that are currently masked due to prevalent redundancy.
Migrating cells and axons respond to a multitude of extracellular cues encountered along their migratory paths. These include secreted cues such as Netrins, which are known to guide migrating cells and axons along the D/V axis of invertebrates and the vertebrate spinal cord [1]–[3], and Wnts, which mediate guidance along the A/P axis [4]. While considerable advances have been made in identifying guidance cues and their downstream mediators, how information from multiple cues is integrated within the cell to enact normal migration patterns has yet to be fully elucidated. To illuminate how a cell calculates the net response to multiple, sometimes additive, overlapping, or opposing inputs we decided to examine genetic interactions between UNC-6/Netrin and Wnt signaling components in the migration of cells and axons that navigate along the D/V or A/P axes of the body wall. In C. elegans a polarity-determining gradient of UNC-6/Netrin secreted by ventral sources of this guidance cue mediates apparent attraction of some migrating cells and growth cones toward the ventral side by signaling through the transmembrane receptor UNC-40/DCC, and also mediates apparent repulsion of other cells and growth cones away from the ventral side by signaling through the transmembrane receptor UNC-5 alone or together with UNC-40/DCC [1], [5]–[7]. This highly conserved instructive guidance system is critical for nervous system patterning in both vertebrates and invertebrates [2], [3], [5]. Wnts also play key roles in cell migration and axon guidance [4], [8], [9]. The C. elegans genome encodes five Wnt ligands (EGL-20, LIN-44, MOM-2, CWN-1, CWN-2), four frizzled receptors (LIN-17/Frizzled, MOM-5, MIG-1/Frizzled, CFZ-2) and a single RYK/Derailed receptor tyrosine kinase (LIN-18) [10]. Wnts, like UNC-6/Netrin, act as both short-range and long-range repellents or attractants, and can function instructively (i. e. , their graded distribution determines polarity) as well as permissively (i. e. , do not instruct, but are necessary for polarity) [8], [9], [11]–[16]. The Wnt binding protein MIG-14/Wntless is required in Wnt producing cells to facilitate Wnt secretion [17]. Wnt activity is further modulated by a number of inhibitors and activators [18]. One family of inhibitors is the Secreted Frizzled Related Proteins (SFRPs), which are soluble glycoproteins widely involved in embryonic development and homeostasis. SFRPs contain two functional domains: the cysteine rich domain (CRD) related to the extracellular portion of Frizzled Wnt receptors, and the Netrin related motif (NTR) defined by homology with Netrin-1. SFRPs can sequester Wnts thereby preventing Wnt ligand-receptor interactions [18]. Netrins and Wnts in C. elegans are well known for having a graded distribution along the D/V and A/P axes, respectively, and can provide polarity information for guiding migration up or down their respective gradients. Accordingly, unc-6/netrin mutants were originally found to affect D/V but not A/P migrations, whereas wnt mutants were originally found to affect A/P but not D/V migrations [1], [3], [14], [15], [19]. However, there have been hints that these signaling pathways, or components thereof, could have functions that are not restricted to migration along a single axis. For example, UNC-40 is involved in A/P migrations of Q neuroblasts [4], [20] and in A/P motor axon dendrite growth [21]. Moreover, we and others [22], [23] have shown that over-expression of UNC-40/DCC in the mechanosensory neurons causes A/P polarity reversals in ALM and PLM axons akin to the effects of impairing Wnt signaling in these neurons [14], [15], [24]. Intrigued by the possibility of integration between Netrin and Wnt signaling, we examined the effects of simultaneously impairing Netrin and Wnt functions on cells and growth cones that navigate along the A/P, the D/V, or both axes. This revealed previously unrecognized and unexpected, redundant roles for Wnt signaling in D/V guidance, and for UNC-6/Netrin signaling in A/P guidance as well as redundant roles that affect organ function and embryonic viability. We further found that a balance between signaling by UNC-5 and LIN-44/Wnt and between specific Wnts, like EGL-20 and CWN-1, contributes to the regulation of A/P polarity and that in the absence of UNC-6/Netrin function, Wntless and SFRP, and by implication one or more Wnts, are required for a long-sought mechanism that functions in parallel to UNC-6/Netrin signaling to regulate D/V migrations. These findings open new avenues for deciphering how A/P and D/V guidance signals are integrated to establish polarity in multiple biological processes and implicate broader roles for Netrin and Wnt signaling - roles that are hidden due to prevalent redundancy between the functions of these cues.
In C. elegans hermaphrodites, migration of the DTCs of the somatic gonad represent an excellent model system to study how polarity information provided by extracellular cues is utilized to enact normal migration patterns. The two DTCs are born in the ventral mid-body of the animal and migrate post-embryonically in three sequential phases alternating between the A/P and D/V axes of the body wall as they lead the elongation of anterior and posterior mirror image U-shaped hermaphrodite gonad arms (posterior arm shown in Figure 1A). In phase 1 the anterior and posterior DTCs migrate away from one another along the ventral body wall muscles towards the head and tail, respectively. In phase 2 the DTCs reorient 90° and migrate along the D/V axis of the lateral epidermis. In phase 3 the DTCs reorient again 90° and migrate on the dorsal body wall muscles back to the mid-body of the animal [5]. Many of the genes that regulate DTC migrations, such as Netrins, Wnts, integrins and matrix metalloproteases, are highly conserved and function to guide cell and axon migration in vertebrates and invertebrates [25]. UNC-6/Netrin, through its transmembrane receptors UNC-40/DCC and UNC-5, guide the D/V migrations of the DTCs [5]. In unc-5, unc-6 and unc-40 loss of function (lof) mutants, the DTCs execute phases 1 and 3 with normal timing but frequently fail to execute phase 2 migration, which is normally mediated by UNC-40 and UNC-5 eliciting migration away from ventral UNC-6 sources [5]. Phase 2 failures cause ‘ventralized’ gonad arms that lie solely over the ventral muscle bands (Figure 1B). The incomplete penetrance of this defect in null mutants of Netrin signaling components (which is also observed in Netrin-dependent axon guidance) suggests the existence of a previously unknown, long-sought signaling pathway that functions in parallel with Netrin signaling to execute D/V migrations. mig-14 encodes the C. elegans homolog of Wntless, a seven transmembrane domain protein necessary for Wnt secretion [17]. When the function of mig-14 is impaired, phase 2 migration of both DTCs is essentially normal, but phase 3 migrations display a 180° polarity reversal and are frequently mis-oriented away from mid-body rather than towards it (Figure 1C) [26]. We refer to this defect as a phase 3 A/P polarity reversal. unc-6, unc-5 and unc-40 lof mutants rarely display phase 3 polarity reversals; however, these reversals are observed when UNC-5 is over-expressed in the DTCs (N. Levy-Strumpf & J. Culotti, in preparation). This prompted us to examine the outcome of simultaneously impairing the function of MIG-14/Wntless and UNC-6/Netrin signaling components. We therefore generated double mutants carrying different combinations of mig-14/wntless, with unc-5, unc-40 and unc-6 alleles (detailed allelic description is provided in Table S1). The DTC migrations observed in these double mutants exhibited one of four migratory patterns: a normal migratory pattern (Figure 1A), a phase 2 D/V migration failure (Figure 1B; Figure 2C, 2D grey bars), a phase 3 polarity reversal (Figure 1C; Figure 2A, 2B black bars), or a combination of a phase 2 failure followed by a phase 3 A/P polarity reversal resulting in a ‘no turn’ phenotype (Figure 1D; Figure 2 red bars). While neither allele of mig-14/wntless caused significant phase 2 D/V migration failures (<1%, n = 1147), both alleles significantly enhanced phase 2 failures caused by unc-5, unc-6 and unc-40 null mutations. Anterior (Figure 2C; Table S2) and posterior DTCs were markedly affected, with as many as 98% of posterior DTCs exhibiting phase 2 D/V migration failures in the double mutants (Figure 2D; Table S2). A reciprocal situation was found for the regulation of A/P guidance by MIG-14 and Netrin signaling components. While Netrin signaling mutants rarely caused phase 3 polarity reversals, they significantly enhanced the phase 3 A/P polarity reversals of posterior DTCs in mig-14/Wntless mutant animals, with some allelic combinations exhibiting almost complete penetrance of the defect (97%, n = 151) (Figure 2B; Table S2). These results reveal redundant roles for UNC-6/Netrin signaling components and MIG-14/Wntless in determining the A/P polarity of the DTC during their phase 3 migration, and together with the previous results demonstrate that the mechanism that functions in parallel with UNC-6/Netrin signaling to regulate D/V DTC migration depends on MIG-14/Wntless, and by implication, on one or more Wnt signals that function redundantly with UNC-6. Wntless displays greater effect on the posterior DTC as evidenced by the higher frequency of phase 3 A/P polarity reversals of posterior DTCs in mig-14/wntless mutants. To further explore a role for Wnts and Netrins in guiding the anterior DTC, we included an additional Wnt regulator in this analysis. sfrp-1 encodes the C. elegans homolog of SFRPs. sfrp-1 is expressed anteriorly in C. elegans and functions to inhibit anterior Wnts such as CWN-1 and CWN-2 [27]. unc-5 and sfrp-1 (gk554) mutants displayed a low incidence of A/P polarity reversals (2% n = 230 in sfrp-1); nevertheless, simultaneous loss of sfrp-1 and unc-5 caused polarity reversals in 58% (n = 189) of the anterior DTCs (Figure 2A; Table S2) supporting the finding of redundant functions for Wntless and UNC-5 signaling in guiding the phase 3 A/P migration of the DTCs. Similar to mig-14/wntless mutations, the sfrp-1 mutation also enhanced the D/V guidance defects of unc-5 lof mutations (Figure 2C). unc-5 null mutants display a high percentage of posterior DTC phase 2 D/V migration failures resulting in DTCs that remain on the ventral side throughout their migration. This raises the possibility that the mig-14; unc-5 double mutant enhancement of A/P polarity reversals might be an indirect consequence of the ventral positioning of the DTC. To exclude this possibility, we repeated the mig-14; unc-5 double mutant analysis using the weak unc-5 (ev644) allele, which manifests only a low penetrance of phase 2 migration failures [28]. Similar to what we observed with the unc-5 null mutations, the frequency of mig-14/wntless posterior DTC phase 3 A/P polarity reversals was enhanced in the mig-14 (k124); unc-5 (ev644) double from 37% to 83%, [of which 71% occurred on the dorsal side (Figure 2B black bar) ]. These results demonstrate that a simultaneous reduction in Wntless and UNC-6/Netrin signaling components causes an increase in A/P polarity reversals regardless of whether phase 3 occurs on the ventral or dorsal side and regardless of whether phase 3 is preceded by a normal or a failed phase 2 migration. Furthermore, the finding that the mig-14/wntless; unc-5 (ev644) double mutant displays a low frequency of D/V migration defects, but a high frequency of A/P reversals (the same as the unc-5 null), raises the possibility that the functional requirements for UNC-5 in D/V versus A/P guidance are genetically separable. To further examine whether the genetic interactions between Netrin signaling mutants and mig-14/wntless or sfrp-1 reflect Wnt signaling defects, we analyzed genetic interactions between unc-5 and various Wnt- and Wnt receptor-encoding genes. For most of this analysis we used unc-5 (RNAi) to impair unc-5 function. unc-5 (RNAi) causes DTC phase 2 D/V migration failures (visualized as ‘ventralized’ gonad arms) typical of unc-5 lof alleles, which were quantified to provide a measure of efficacy of the RNAi treatment on UNC-5 function (Figure 3B; Table S3). The effect of unc-5 (RNAi) on A/P polarity reversals was comparable to that of unc-5 lof alleles (Figure S3; Table S4). We examined whether mutations in each of the five Wnt receptor-encoding genes (lin-17, lin-18, cfz-2, mig-1 and mom-5) might function redundantly with unc-5 to prevent phase 3 polarity reversals. Except for mom-5 alleles, which cause a high frequency of phase 3 DTC A/P polarity reversals [29] (also the subject of a report by N. Levy-Strumpf & J. Culotti, in preparation), these mutations cause few if any phase 3 reversals. For example, two putative null alleles of lin-17/frizzled, (n3091) and (n671), caused only 1–2% phase 3 DTC A/P polarity reversals, however both alleles were significantly enhanced for these defects by unc-5 (RNAi) or unc-5 mutations (Figure 3A; Table S3; Figure S3; Table S4), whereas lin-18, cfz-2, and mig-1 mutations were not enhanced (Figure 3, Table S3). Thus, out of the four Wnt receptor genes examined here (lin-17, lin-18, cfz-2, and mig-1), only lin-17 was found to function redundantly with unc-5 for phase 3 A/P reversals. Consistent with a role for LIN-17/Frizzled in phase 3 polarity determination, we found that GFP-tagged LIN-17 is expressed in the DTCs throughout development (Figure 4), whereas LIN-18, CFZ-2, and MIG-1 are not reportedly expressed in these cells [15], [30], [31] (see also Figure S2). In examining the role of specific Wnts in DTC migration, we found that single wnt gene mutations caused few or no DTC migration defects (Figure 5, Table S5); however, simultaneous impairment of unc-5 and egl-20/wnt (by RNAi or by mutation) caused synergistic enhancement of phase 3 A/P polarity reversals from 13% (n = 555) in egl-20 (n585) to 66% (n = 353) in unc-5 (RNAi) egl-20 (n585) (Figure 5B; Table S5) animals, to 70% (n = 124) in unc-5 (ev489) egl-20 (n585) animals, and to 55% (n = 210) in unc-5 (e53) egl-20 (n585) animals (Figure S3; Table S4). This demonstrates that unc-5 functions redundantly with egl-20/wnt to direct the posterior DTC back to the mid-body during phase 3 just as unc-5 functions redundantly with lin-17/frizzled, mig-14/wntless and sfrp-1 in this process (Figure 2; Figure 3). Interestingly, the frequency of posterior DTC phase 3 A/P polarity reversals in lin-17; unc-5 (RNAi) and in unc-5 (RNAi) egl-20 (n585) animals was suppressed by lin-44 (n1792) from 23% to 6% and 66% to 38%, respectively (Figure 3A; Table S3; Figure 5B; Table S5). These results demonstrate that enhancement of lin-17/frizzled and egl-20/wnt single mutant A/P polarity defects by impaired unc-5 function requires LIN-44 activity and suggest that a balance between UNC-5 and LIN-44 activities promotes normal DTC phase 3 A/P polarity. unc-5 (RNAi) did not enhance the frequency of cwn-1 (ok546); cwn-2 (ok895) double mutant DTC phase 3 A/P polarity reversals (n>300) (Figure 5; Table S5); however, a redundant role for lin-44/wnt in anterior and posterior DTC migrations was uncovered when cwn-1 and cwn-2 were severely compromised (n>290) (Figure 5). This function of lin-44/wnt is partially dependent on unc-5 as determined by the suppression caused by unc-5 (RNAi) of the lin-44; cwn-1; cwn-2 triple mutant (Figure 5; Table S5). These results demonstrate that enhancement of cwn-1; cwn-2 double mutant phase 3 A/P polarity reversals by a lin-44 lof requires UNC-5 activity and provides further evidence that a balance between UNC-5 and LIN-44 activities promotes normal DTC phase 3 A/P polarity. A balance between various Wnts in determining DTC polarity was also observed. We analyzed single or combination mutants of egl-20, lin-44, cwn-1, and cwn-2; in all cases null or severe lof alleles were used (Table S1). This analysis revealed either greater than additive enhancement of the phase 3 A/P polarity reversals (indicating redundancy) or mutual suppression of the defects (indicating the requirement for a balance between gene functions). For example, we found that cwn-1 functions redundantly with cwn-2, while lin-44 functions redundantly with egl-20 and with cwn-1 or cwn-2 (or both) to regulate DTC phase 3 A/P polarity. Conversely, the egl-20 phase 3 reversals were markedly, but not completely, suppressed by mutations in cwn-1 (Figure 5A black bars; Figure 5B grey bars; Table S5). This suppression implies that a balance between EGL-20/Wnt and CWN-1/Wnt is also required to promote normal phase 3 A/P polarity. Most of the Wnts and their receptors, either alone or in combination, mainly affected posterior DTC migration; the only exceptions were sfrp-1 and the double knockout of cwn-1 and cwn-2, which mainly affected anterior DTC migration (Figure 5A black bars). This corresponds to the expression pattern of CWN-2 and SFRP-1, which are mainly expressed in the anterior, compared to the posterior expression of LIN-44, EGL-20 and CWN-1 [27]. The lin-44; cwn-1; cwn-2 triple mutant reveals a redundant role for lin-44 and the cwn genes in both anterior and posterior DTCs. This is interesting given that lin-44 is the most posteriorly expressed Wnt in L1 larvae [27]. However, this observation is not unprecedented [32]. It is possible that a more anterior source of LIN-44 [27], [30] accounts for this, or that CWN-1, which is expressed more broadly, somehow facilitates the LIN-44 effect on the anterior DTC. To examine whether axon pathfinding is also regulated by redundant functions of Wnt and Netrin signaling, we determined whether unc-5 egl-20, or mig-14/wntless; unc-6/netrin double mutants display any synergistic axon guidance defects. We analyzed two different types of neurons: the CAN neuron, which is bipolar and extends axons along the A/P axis (Figure 6A and 6B), as well as the mechanosensory neurons, which extend axons along the A/P and D/V axes. Single unc-5 (e53), unc-5 (ev489) or egl-20 (n585) mutants rarely displayed CAN axon guidance defects (Figure 6G; Table S6), whereas the unc-5 egl-20 double mutants displayed a synthetic defect resulting in 5–10% (n>100) CAN axon reversals, which were observed predominantly in the posterior axon (Figure 6C, 6D, and 6G). Other defects such as premature axon termination and excessive branching (Figure 6E, 6F) were also observed, resulting in a total of 9–17% defects (Figure 6G) depending on the allele or the incubation temperature [33]. Similarly, unc-6 (ev400) and mig-14 (k124) mutants rarely displayed CAN axon guidance defects, while the penetrance of these defects in mig-14 (k124); unc-6 (ev400) double mutants was 18% (n = 104). These results suggest a role for Netrin signaling in guiding CAN axons - a role that is redundant with Wnt signaling, which has established instructive and permissive functions in A/P guidance of several other axons in C. elegans [14], [15]. These results are consistent with the apparent Wnt-redundant role of UNC-6/Netrin signaling in A/P guidance of DTCs. To explore D/V axon guidance, we analyzed the mechanosensory AVM and PVM neurons, which normally extend axons toward ventral sources of UNC-6 [5] (AVM is shown in Figure 7A). A marked enhancement of AVM and PVM axon guidance defects was observed when both Wnt and Netrin signaling were impaired (Figure 7; Table S7). Although unc-5 and egl-20 mutants each displayed mild D/V guidance defects (e. g. , partially longitudinal axons) and low frequency severe defects (entirely longitudinal axons) mainly at 25°C (Figure 7C; Table S7), the unc-5 egl-20 double mutants displayed greater than additive penetrance reaching approximately 40% defects at 25°C (n = 125) (Figure 7B, 7C; Table S7). UNC-5 was not known to be involved in attraction towards ventral sources of its ligand UNC-6 [34], but rather to elicit migration away from these sources. These results demonstrate that UNC-5 has a role in guidance toward ventral sources of UNC-6/Netrin that extends beyond its conventional instructive role in axon repulsion [5], [6], [35], [36], and that this role is redundant with a role for EGL-20 in D/V guidance. Observations implying a role for UNC-5 in apparent attraction of HSN axons to ventral sources of UNC-6 were recently made independently by another group [37]. A/P polarity reversals were also observed in the touch neurons of unc-5 egl-20 double mutants; these included axon reversals of the AVM, PVM or, in rare instances, the ALM axon (Figure 7C; Table S7). Taken together these observations demonstrate that Wnt and Netrin signaling, as they do in phase 2 and phase 3 DTC migration, also function redundantly in regulating D/V and A/P guidance, respectively, of migrating axons (e. g. , CAN, AVM, PVM and ALM). To determine whether the unconventional role of UNC-5 in AVM and PVM axon guidance is cell autonomous, we expressed unc-5 under the mec-7 mechanosensory neuron-specific promoter and assayed its ability to rescue AVM and PVM axon guidance defects of the unc-5 (e53) egl-20 (n585) double mutant. Combined defects were rescued by roughly half, whereas the severe AVM and PVM guidance defects were rescued by nearly 60% (Figure 7, Table S7), suggesting that at least part (and perhaps all) of UNC-5 function in this context is cell autonomous. In addition to the redundancy observed for Wnts and UNC-6/Netrin in DTC migration and axon guidance, in the process of generating the mig-14/wntless; unc-6/netrin double mutants we observed synergistic effects on vulval morphology and function resulting in a nearly complete egg laying defect (Figure S4A) as well as a higher incidence of protruding vulvae (Figure S4C). A marked synergistic effect was also observed on viability. The mig-14 (ga62); unc-6 (ev400) double mutant could not be generated, while the weaker mig-14 (k124) allele in combination with the unc-6 (ev400) null allele displayed extensive embryonic lethality causing extremely low brood sizes averaging less than 10 worms (Figure S4B). Arrested and malformed embryos were frequently observed in the mig-14 (k124); unc-6 (ev400) gonads (Figure S4D and S4E). Although embryonic lethality could be secondary to the egg laying defect, comparing the brood sizes of unc-6 (ev400) hermaphrodites that failed to lay any eggs to those of mig-14 (k124); unc-6 (ev400) hermaphrodites indicated a greater reduction in viability in the double mutant. These results indicate that Wnts and UNC-6/Netrin function redundantly to orchestrate both vulval function and at least one essential developmental process critical for early development in C. elegans.
We have identified and characterized unconventional roles for Wnts and UNC-6/netrin in guiding cell and axon growth cone migrations in C. elegans. Graded distributions of Netrins and Wnts along the dorso-ventral (D/V) and antero-posterior (A/P) axes of the body wall, respectively, have long been thought to provide polarity information for migrations along these respective axes since UNC-6/Netrin signaling deficits were originally found to primarily affect migrations along the D/V axis and Wnt signaling deficits were found to primarily affect migrations along the A/P axis. Here we have characterized the effects of individual and combined loss of function of Wnt and Netrin signaling components on the migrations of cells (the hermaphrodite DTCs) and axons (of CAN, AVM, PVM and ALM neurons) in vivo. Our analysis indicates that the idea that guidance cues like Netrins and Wnts contribute to guidance only along the axis of their gradation is an oversimplification. We found that compromising Wnt signaling reveals an unexpected, redundant role for Netrin signaling components in orienting DTC migration along the A/P axis and, conversely, compromising Netrin signaling reveals an unexpected, redundant role for Wnt signaling in guiding DTC migration along the D/V axis. These findings indicate that Netrins and Wnts have two major functions in guiding migrations. One is to provide instructive polarity information along the axis of their gradation and the other is to help guide migrations orthogonal to the axis of their gradation by functioning redundantly with each other. These results demonstrate that Wnt signaling can function independently of, but redundantly with, Netrin signaling to promote D/V oriented DTC migration and that UNC-5 signaling can function independently of, but redundantly with, Wnt signaling to promote A/P oriented DTC migration. Furthermore, our finding that the unc-5 (ev644) hypomorphic allele has impaired A/P guidance, but almost intact D/V guidance, indicates that the unconventional function of UNC-6, UNC-5, and UNC-40 in DTC phase 3 A/P polarity may be genetically separable at the level of UNC-5 function from conventional instructive UNC-6 signaling [36]. These results provide a conceptually novel view of how A/P and D/V guidance mechanisms can be regulated in vivo. Notably, simultaneous compromise of both Wnt and Netrin signaling pathways caused the nearly complete penetrance of posterior DTC phase 2 D/V and phase 3 A/P migration defects, demonstrating that the sum of Wntless and Netrin signaling accounts for the entire phase 2 and phase 3 guided migrations of this cell. This together with the markedly increased frequency of AVM or PVM D/V guidance defects in the mig-14; unc-6 double mutants, compared to the unc-6 nulls, identifies Wnt signaling as the long-sought mechanism postulated to function in parallel to Netrin in regulating D/V guidance in migrating cells and axons in C. elegans and possibly across different species. The phenotypes observed in the mig-14; unc-5 and mig-14; unc-6 double mutant are reminiscent of defects observed in src-1 mutants. Furthermore, src-1 mutant defects are suppressed by loss-of-function mutations in Rho family GTPases [38]. This suggests the possibility that both the Netrin and the Wnt signaling pathways are either regulated by SRC-1 or converge on SRC-1 to regulate small GTPase activity known to be critical for A/P polarity establishment [39], [40]. SRC-1 also binds integrins, which are involved in the regulation of the phase 3 turn [25], [41]. It would be interesting to explore further how these three signaling pathways (Wnts, Netrin, and integrins) are integrated to regulate the turning of the DTC, which must involve coordination of the cytoskeletal rearrangements and adhesion processes. Netrin signaling components are known to have a greater impact on the migration of posterior DTCs [5]. The data presented here reveal a more comprehensive contribution of the Netrin guidance system to D/V migration of the anterior DTC that is evidently masked by Wnt redundant functions. We speculate that anterior, like posterior DTC guidance, is likely to be fully governed by Netrin and Wnts. Although anterior Wnts, like the CWNs and the anterior Wnt regulator SFRP-1, have greater effects on anterior DTC migration, these effects are not fully penetrant even when a Netrin signaling component is simultaneously impaired. Which of the Wnts other than the CWNs govern anterior DTC migrations remains to be determined. The difference in response of the anterior and posterior DTCs seems to be dependent on the composition of Wnts graded oppositely along the A/P axis. This difference in Wnt responsiveness is likely necessary to facilitate the mirror image migration of the anterior and posterior DTCs. We identified redundant roles for Wntless or Wnt signaling components and UNC-5, while in a variety of compromised genetic backgrounds we observed mutual suppression of lin-44 and unc-5 mutant phase-3 A/P polarity defects. These findings indicate that UNC-5 cooperates with some Wnts and opposes the function of other Wnts to maintain a fine balance of activities required for proper A/P polarity. Similar interactions occur between the different Wnts. Various combinations of Wnt double mutants resulted in synthetic enhancement of phase 3 A/P polarity reversals revealing redundancies between different Wnt signaling components, while on the other hand EGL-20 and CWN-1 display opposing functions. A fine balance of Wnt signaling was similarly reported to be required for migration and positioning of other cell types in C. elegans [14], [24], [32], [42]. Our results reveal a contribution of UNC-5 to this balance and imply the existence of a complex regulatory network of interactions between the different Wnts and UNC-5 that determines the phase 3 A/P polarity of DTC migration (Figure 8D; diagrammed in Figure S5). We propose that the balance between Wnts and UNC-5 determines whether anterior or posterior polarities are established (and hence the direction taken on the A/P axis), whether the cell halts, or whether the cell reorients to the D/V axis. The ability of UNC-5 to oppose some Wnts can also explain in principle how an sfrp-1 lof mutation, which is predicted to up-regulate interacting Wnts, is able to enhance rather than suppress unc-5 (RNAi) -induced DTC migration defects. Our data demonstrate that UNC-6 signaling has a role in A/P guidance that is redundant with Wnt mediated A/P guidance, while at the same time Wnts have a role in D/V guidance that is redundant with UNC-6 and UNC-5 mediated D/V guidance. These observations raise a perplexing question: How can Wnt and UNC-6/Netrin signals contribute to D/V and A/P guidance, respectively, when they appear to be graded along orthogonal axes (i. e. , A/P and D/V axes, respectively) of the body wall? For the Netrin-redundant contribution of Wnts to D/V guidance, we propose that one or more A/P graded Wnts inhibits the possibility of leading edge formation at the anterior and posterior ends of the cell or growth cone, thereby facilitating D/V guidance - possibly by inhibiting or excluding the polarity establishment machinery from the A/P poles and effectively localizing it to the center of the cell (along the D/V axis) where is can be employed for D/V guidance (Figure 8). This alone would not provide polarity information for the ensuing phase 2 ventral to dorsal migration; however, a possibly analogous situation is provided by early gastrulation in the C. elegans embryo. In this case, ingression of endodermal precursors is regulated by a Wnt-frizzled signaling pathway that induces (what could be considered a ventral to dorsal) cell movement by activating apically localized myosin II contraction, thereby effectively squeezing the cell into the embryo' s interior [43]. Polarity information in this case may be provided by an intrinsic polarity of the moving cells that is simply activated by Wnt signaling. We propose that this may be akin to the unconventional regulation of phase 2 DTC migration by Wnt signaling. Ventral to dorsal migration is further driven by the conventional instructive ability of an UNC-6/netrin gradient to mediate apparent repulsion by signaling through an UNC-5-dependent receptor mechanism ([36] and see below). It is worth noting that an unc-5 mutation behaves like a Wnt signaling mutation in its effects on A/P polarity. Therefore, whatever functions we attribute to Wnt signaling may also be attributed to UNC-5 signaling. This raises the possibility that unconventional UNC-5 signaling contributes to ventralward AVM axon guidance (and perhaps also DTC D/V guidance) by inhibiting anterior and posterior leading edges as proposed above for Wnts. This role of UNC-5 is strictly redundant with the role of EGL-20/Wnt and is separate from UNC-5' s instructive ability to mediate repulsion away from ventral sources of UNC-6. An example of apparent A/P bipolar inhibition was recently published in a study of HSN axon guidance, where it was observed that Wnt signaling components function to exclude UNC-40/DCC localization from the anterior and posterior poles of the HSN growth cone [37]. This report lends additional support to the mechanistic specifics of our model. In a minor variation of the bipolar inhibition model, differential inhibitory effects of UNC-5 on anterior and posterior poles may determine the A/P polarity of axon and DTC phase 3 migrations by skewing the balance between UNC-5 and Wnt inhibitory signals toward one pole. It is tempting to speculate that bipolar inhibition of anterior and posterior leading edges could also serve as a general mechanism for regulating cessation of cell migration along a single axis as well as reorientations from one axis to another, thus determining either cell positioning along the A/P or D/V axes at the end of a migratory path, or changes in trajectory from one axis to another in complex pathfinding processes. The dual function of UNC-5 in A/P and D/V guidance also raises the untested possibility that both Wnt and Netrin signaling pathways converge on the UNC-5 receptor and that this receptor assimilates information from both cues. The role of Netrin and its receptors is not limited to cell and axon guidance - it contributes to a wide range of biological processes, including organogenesis, synaptogenesis, dendritic self-avoidance, cell adhesion, angiogenesis, cell survival, tumor formation and metastasis [44]–[48]. Like Netrins, Wnts control a variety of developmental processes including cell migration and axon guidance, synaptogenesis, polarity establishment, cell fate determination, mitotic spindle reorientation, and are also involved in tumorigenesis and various human diseases [4], [10], [13], [49]–[51]. Here we show that Wnts and Netrin signaling components share redundant functions, which are not readily revealed except by impairing both pathways simultaneously, suggesting that they might be substantially involved in more processes and to a greater extent than currently appreciated. The finding of shared functions suggests that Wnt and Netrin signaling mechanisms could be co-regulated. One putative co-regulator is the DAF-12 steroid hormone receptor, which is required for all DTC reorientations [52], [53]. DAF-12 is responsible for the transcriptional activation of UNC-5, just prior to the reorientation of the DTC from the A/P to the D/V axis [36], which in turn helps drive UNC-6 dependent ventral to dorsal phase 2 DTC migration. The nearly full penetrance of the mig-14; unc-5 double for phase 2 and phase 3 DTC defects raises the distinct possibility that one or more Wnt signal transduction components is co-regulated by DAF-12 along with the UNC-5 receptor not only at the first, but possibly also at the second turn of the DTCs. Interestingly, SFRPs, known to function as Wnt regulators, contain a cysteine rich domain (CRD), which is highly homologous to the Frizzled family CRD, but also contain a Netrin-related motif (NTR domain) [54]. It is an interesting possibility that the SFRPs may function in some cases to co-regulate these two fundamental pathways. Redundancy between Wnt and Netrin signaling components extends beyond DTC and growth cone migration to include functions essential for viability. We have observed that double mutants of unc-6 (ev400) with mig-14 (ga62) are inviable. These results suggest that certain Wnts and UNC-6 may also function redundantly to regulate at least one essential developmental process critical for viability. Furthermore, the mig-14; unc-6 double mutants displayed a dysfunctional vulva and a fully penetrant egg-laying defect. Given the involvement of Netrin and integrins in anchor cell invasion [44], which is reminiscent of their involvement in DTC phase 3 A/P polarity [25], it is possible that the egg laying defect is the result of redundancy between Wnt and Netrin function in regulating anchor cell invasion. Taken together, our observations imply that during normal development as well as in some pathological conditions, Wnts and Netrins may have functions that are not apparent due to their redundant output - a notion that is important to consider in order to fully elucidate the underlying mechanisms governing these processes. Furthermore, the observation that Netrins and Wnts have shared functions in A/P and D/V guidance is an important potential precursor to understanding how polarity information from these two guidance systems is integrated to generate defined migratory patterns. Our data provides a novel conceptual view by which D/V and A/P polarity establishment may be effectively one in the same, D/V polarity being, in part, the culminating result of bipolar inhibition of polarity along the AP axis.
Standard procedures were used for the culture, maintenance and genetic analysis of C. elegans [55]. All strains were grown at 20°C for analysis, unless indicated otherwise. Mutant strains and transgenic lines used in this study are listed in Table S1. Strains not isolated in our laboratory were obtained from the Caenorhabditis elegans Genetics Center (University of Minnesota), or as indicated in the Acknowledgements section. When necessary, double mutants were verified by PCR; primers are listed in Table S1. unc-5 (RNAi) constructs were generated by cloning a 574 bp EcoRI fragment spanning nucleotides 563–1137 of unc-5 into the pPD129. 36 L4440 vector [56]. In vitro transcribed RNA (Ambion MEGAscript kit) was then injected into young adult hermaphrodites by standard procedures [57]. F1 progeny of the injected worms were analyzed as L4 larvae or adults and compared to the respective non-injected strains. DTC migration patterns or axon pathfinding were scored by mounting 1 mM levamisole-treated animals (L4 or adult stage) on 2% agarose pads for observation using Differential Interference Contrast (DIC) and fluorescence microscopy (Leica DMRA2 or DMRB microscope). All strains assayed in Figure 2 and described in Table S2 carried the gly-18p: : gfp transgene to mark the DTCs. gly-18p: : gfp rarely affects D/V or A/P guidance of the DTC (Table S2). The polarity reversal phenotype is highly dependent on the incubation temperature. Care was taken to analyze all comparable strains under the same growth conditions, therefore, a control strain grown under the same conditions was included in each set of experiments; data from several independently generated lines were analyzed and the data pooled. gli-18: : gfp was also used to mark the CAN neurons (Figure 6), while mec-7: : gfp or mec-4: : gfp were used to label the mechanosensory neurons (Figure 7). In order to assess the frequency of the egg laying defects and brood sizes, young hermaphrodites (no older than the L3 larval stage) of unc-6 (ev400) or mig-14 (k124) mutants or the mig-14 (k124); unc-6 (ev400) double mutants were cloned and followed over a period of about 5 days. Worms were scored as having an egg-laying defect only if they fully failed to lay eggs. In these cases the parent becomes a “bag” of trapped larvae that eventually eat their way out. To assess brood sizes, the progeny of each cloned worm was counted. It should be noted that the reported brood sizes of the mig-14 (k124); unc-6 (ev400) are an over-estimation of the actual viable propagating progeny, as not all larvae develop fully to the adult stage. Therefore the synergistic effect between these two mutations is likely even greater than presented. Standard errors of the proportion (SE) were calculated assuming a binomial distribution of the observed proportion and the actual sample size. Statistical tests were carried out using a standard (two-tailed) comparison of two proportions (Z test). All P values represent the probability that the measured frequency of the phenotype is the same for the two strains being compared. A P-value of less than 0. 05 is considered significant. All comparisons described as significant in the Results section were based on this criterion. | While ample information was gathered in past decades on identifying guidance cues and their downstream mediators, very little is known about how the information from multiple extracellular cues is integrated within the cell to generate normal patterning. Netrin and Wnt signaling pathways are both critical to multiple developmental processes and play key roles in normal development as well as in malignancies. The UNC-6/Netrin guidance cue has a conserved role in guiding cell and growth cone migrations along the dorso-ventral axis, whereas Wnts are critical for determining polarity and guidance along the antero-posterior axis. In this study we show that these two signaling pathways function redundantly in both antero-posterior and dorso-ventral guidance as well as in processes essential for viability. Furthermore, we demonstrate that a fine balance between Wnt and Netrin signaling pathways is critical for proper polarity establishment and identify Wnt signaling as one of the long sought mechanisms that signal in parallel to Netrin to promote dorso-ventral guidance of cells and axons in Caenorhabditis elegans. These findings pave the way to unraveling the broader roles of Wnt and Netrin signaling pathways and provide a conceptually novel view of how antero-posterior and dorso-ventral guidance mechanisms are orchestrated. | Abstract
Introduction
Results
Discussion
Materials and Methods | invertebrates
cell motility
caenorhabditis
neuroscience
animals
gene function
animal models
developmental biology
mutation
caenorhabditis elegans
model organisms
organism development
molecular development
molecular genetics
body plan organization
morphogenesis
pattern formation
research and analysis methods
developmental neuroscience
cellular neuroscience
cell biology
axon guidance
gene regulatory networks
cell migration
genetics
nematoda
biology and life sciences
computational biology
organisms | 2014 | Netrins and Wnts Function Redundantly to Regulate Antero-Posterior and Dorso-Ventral Guidance in C. elegans | 11,567 | 327 |
The site frequency spectrum (SFS) has long been used to study demographic history and natural selection. Here, we extend this summary by examining the SFS conditional on the alleles found at the same site in other species. We refer to this extension as the “phylogenetically-conditioned SFS” or cSFS. Using recent large-sample data from the Exome Aggregation Consortium (ExAC), combined with primate genome sequences, we find that human variants that occurred independently in closely related primate lineages are at higher frequencies in humans than variants with parallel substitutions in more distant primates. We show that this effect is largely due to sites with elevated mutation rates causing significant departures from the widely-used infinite sites mutation model. Our analysis also suggests substantial variation in mutation rates even among mutations involving the same nucleotide changes. In summary, we show that variable mutation rates are key determinants of the SFS in humans.
The distribution of allele frequencies across segregating sites, commonly referred to as the Site Frequency Spectrum (SFS), is a central focus of population genetics research as it can reflect a wide range of evolutionary processes, including demographic history as well as positive and purifying selection [1–8]. Until recently, the SFS was usually measured in samples of tens or hundreds of people, but advances in sequencing technology have enabled the collection of sequence data at much larger scales [9–14]. Notably, the Exome Aggregation Consortium (ExAC) recently released high quality, exome-wide allele counts for over 60,000 people [12]. Large sample sizes are valuable because they make it possible to detect many more segregating sites, and to estimate the frequencies of rare variants. For example, the recent dramatic expansion of human populations leaves little signal in the SFS in small samples [15], but is readily detected in large samples, where there is a huge excess of low frequency variants compared to model-predictions without growth [13,14,16,17]. Similarly, large samples enable the detection of deleterious variants that are held at very low frequencies by purifying selection [18–22]. In this paper, we extend the SFS by considering the SFS conditional on the observed alleles at a given site in other species (specifically, other primates in our analysis). Our original motivation was that this could allow us to measure the effects of sequence context on the selective constraint of missense variants. In general, sites with strong levels of average constraint across mammals tend to be less polymorphic within humans [16,23,24] but, to the best of our knowledge, there has not been extensive consideration of the joint distribution of the substitutions across other lineages and the human SFS. In particular, we hypothesized that if an identical substitution has occurred independently in a closely related species—e. g. , in a great ape—then this is strong evidence that the same variant is unlikely to be deleterious in humans. However, an identical substitution in a more distantly related species may be much less informative, as substitutions at other positions within the same gene may change the set of preferred alleles due to epistatic interactions [25–31] (Fig 1). For example, it has been shown that, in a handful of cases, likely disease-causing variants in humans are actually wildtype alleles in mouse, presumably rendered harmless by parallel substitutions at interacting positions [26]. As we show below, the human SFS varies greatly depending on patterns of substitutions in other species. In part, this does appear to be due to differences in fitness effects; however a more important factor is mutation rate variation across sites. Under the widely-used infinite sites model, the SFS is independent of mutation rate; but in the ExAC dataset we observe a clear breakdown of this model. Mutation rates are known to vary across sites due to a variety of different mechanisms, leading to differences between CpGs, transitions and transversions, as well as additional effects that correlate with broader sequence context, replication timing, transcription, recombination rate and chromatin environment [32–39]. We show here that mutation rates are much more variable than generally appreciated, and that rates at some sites are high enough to generate substantial deviations from infinite sites predictions. The main ExAC paper [12] also recently reported that the SFS varies substantially across mutation types, and also noted that this implies departures from the infinite sites model, especially for CpG transitions. In summary, our results suggest more variation in mutation rates across sites than is generally appreciated, and further that the infinite sites model provides a poor fit for population genetic analyses in large modern data sets.
In this section we consider whether mutation rate variation may contribute to the observed trend across cSFS. Under the standard infinite sites assumption, the SFS is independent of mutation rate. However, we conjectured that in the very large sample size of ExAC, the infinite sites assumption may no longer be a good model for the data [12]. To examine this, we stratified the human SFS by mononucleotide mutation types (as well as the dinucleotide mutation type CpG->TpG), for which there are well-characterized differences in mutation rates. For this analysis we focused on intronic sites, to reduce potential effects of selective constraint. We found that the different mutation types have significantly distinct spectra. The fraction of rare variants among CpG->TpG mutations (36%) was roughly half that of non-CpG transitions (71%, see Fig 3A). Similarly, non-CpG transitions have higher mutation rates than transversions and indeed, the SFS for transitions is also skewed towards higher frequencies than transversions (Fig 3B). Overall, the fraction of rare variants in the subsample of Europeans was significantly negatively correlated with germline mutation rates estimated from the deCode project dataset [45] (weighted linear regression p = 4. 9x10−6, see Fig 4A and S1 text). If multiple hits are prevalent within the ExAC sample, then some of them should occur in different subpopulations. Higher mutation rates should then lead to excess sharing of low frequency variants among subpopulations. To verify the occurrence of recurrent mutations, we examined the sharing between the European and East Asian ExAC subsamples. Indeed, at low frequencies, non-CpG transitions exhibited a higher sharing rate than transversions, and CpG transitions exhibited much higher sharing rate than non-CpG sites (For example, for sites with a minor allele count of 10, we get a t-test p < 2. 2x10−16 for both comparisons; see Fig 3C, and a similar analysis performed in Fig 2d in the main ExAC paper [12]). As an additional test of whether mutation rate affects the fraction of rare variants, we turned to sites in transcribed regions. It is known that in such regions, A->G and A->T mutations occur at higher rates on the template (non-coding) strand than on the non-template (coding) strand, due to the effects of transcription-coupled repair or other transcription-associated mutational asymmetries [46–48]. Indeed, as predicted from these rate asymmetries, we observed a 1% difference between the template and the coding strands in the fraction of rare variants in introns (t-test p < 2. 2x10−16 for A−>G, p = 6. 0x10−7 for A−>T). C->T mutations also exhibit a small but significant difference (t-test p = 3. 0x10−4) between the strands, even though, to our knowledge, no previous work has observed a rate asymmetry for C->T mutations (Fig 3D). Similarly, we hypothesized that the SFS at CpG sites might also depend on chromatin environment. Specifically, CpG sites experience high mutation rates only when they are methylated [49–52]. We thus examined the effect of chromatin states in H1 human embryonic stem cell lines, inferred by ChromHMM [53] (as a proxy for germline chromatin states) on the SFS across different mutation types. Methylation levels are expected to be low in active regions including promoters and enhancers and high in repressed regions such as heterochromatin. Indeed, we find highly significant differences in the SFS at CpGs (see S1 text for details), consistent with this expectation: i. e. , fewer rare variants in heterochromatin, where methylation levels are high. In contrast, the other mononucleotide mutations showed only modest variation across chromatin states (Fig 3F). Finally, we found that recombination rate is also negatively correlated with the fraction of rare variants (Pearson correlation p < 2. 2x10−16, and see Fig 3E and S5 Fig). This is consistent with the postulated positive correlation between recombination and mutation rates [54,55]. However, linked selection—which is expected to be more pervasive in regions of low recombination—could also contribute to this trend [56–58]. Overall, the SFS variation patterns across chromatin states, recombination rates, and strands, underscores that heterogeneity in mutation rates does exist within mutation types, and that it has a substantial effect on the SFS. These observations on mutation rate variation led us to conclude that the infinite-sites model provides a poor fit for these large-sample human polymorphism data. We therefore investigated finite-sites mutational models. Below, we describe the fit of various mutational models while using previously-inferred population genetic models of European demography. In particular, we eventually used a modified version of the demography inferred by Nelson et al. [14] (see Materials and Methods for the other demographic models considered). The assumed demography provides a good fit for the SFS of sites with the lowest mutation rates. We asked how well different finite-sites models account for the observed relationship between de-novo mutation rates and the SFS. First, we considered the Jukes-Cantor model, which uses a 4 x 4 uniform mutation transition matrix [59]. But we were surprised to find that this finite sites model barely improved the fit to the SFS across the range of estimated mutation rates (Fig 4A). In our simulations, the probability of obtaining more than one mutation on the genealogy of a segregating site is low enough that the finite sites SFS is similar to the infinite sites SFS, even at the relatively high mutation rate estimated for CpGs. We hypothesized that we might achieve a better fit if we consider the fact that some sites have higher intrinsic mutation rates than the mean for the particular nucleotide change at that site; this notion has received increasing support in the recent decade from both evolutionary and family-based studies of human mutation rates [32–35,37,60–62]. We therefore augmented the Jukes-Cantor model by incorporating additional variation in mutation rates across sites belonging to each mutation type (see Materials and Methods). The augmented Jukes-Cantor model with within-mutation-type variation fitted the data well, including the large difference in SFS between CpG and non-CpG sites (Fig 4A). The augmented model suggests that 3% of mutations within a mutation type have a mutation rate of over 5 times the mean rate for that type. This estimate is close to the level of mutation rate variation inferred by Hodgkinson and Eyre-Walker [63]. It is natural to wonder what effect recurrent mutations may have in smaller samples. Small samples have the disadvantages of increased noise and limited temporal resolution of analysis. For example, in demographic inference, larger samples are essential for detecting the signal of recent rapid growth of the human population [17,64,65]. Interestingly, we found that samples much smaller than ExAC may also create an unappreciated bias, as we describe next. We examined the effect of subsampling the SFS of the European ExAC sample to a smaller number of individuals (see S1 text). SFS differences between non-CpG transitions and transversions remained roughly the same, even with a sample of a few hundred people. Conversely, the difference between CpG and non-CpG sites changed dramatically for smaller samples. For samples smaller than 1500 people, there appears to be more rare variation in CpG than non-CpG transitions (Fig 4B, S4 Fig). This finding exemplifies that if one category of sites has substantially more rare variation in the population than a second category, the sample SFS may actually exhibit more rare variation in the second category. Therefore, a comparison of the amount of rare variation across categories of sites may yield different orderings, depending on the sample size. Finally, we returned to the species trend across cSFS that we described earlier (Fig 2C). Given the previous observations on SFS differences between mutation types, we asked whether the trend across substituted-species cSFS we described earlier (Fig 2C) could be explained by differing compositions of the various mutation types. Indeed, most of this trend is due to the fact that CpG transitions make up a higher fraction of sites for more closely related substituted species (Spearman ρ = −0. 9, p = 0. 08, and see S2 Fig). Since CpG transitions are depleted of rare variants, this results in the cSFS skewness trend. Namely, the fraction of rare variants is strongly negatively correlated with the fraction of CpG transitions across substituted species (Pearson r = −0. 997, p = 9. 7x10−6 for nonsynonymous mutations; r = −0. 999, p = 9. 9x10−7 for synonymous mutations, see Fig 5C). Why is the fraction of CpG transitions negatively correlated with the relatedness of the substituted-species to humans? Below, we suggest how this could be explained through the mutational mechanism of CpG transitions, which leads to different substitution dynamics on evolutionary timescales than the dynamics at non-CpG sites [66,67]. Substituted-species sites likely experienced two independent mutations at the site during primate evolution, and are therefore enriched for hypermutable sites [60,61,63]. A simple model that we develop in S1 text supports this intuition. In this model, we initially assumed a “uniform molecular clock” regime in which substitutions accumulate at the same yearly rate across the primate phylogeny. Under this assumption, differences in the distribution of mutation rates between substituted-species categories should be vanishingly small (S6 Fig). However, recent work [66,68–71] has demonstrated that while the “uniform clock” assumption is valid for some mutation types—importantly, CpG transitions—the yearly substitution rates of other mutation types depend heavily on life-history traits such as generation time [70,72,73], and thus vary extensively across primates. Notably, Moorjani et al. have pointed out that this difference leads to variable mutational spectra across primates [68]. We therefore augmented our model by including two mutation categories: mutation types that follow a “uniform clock”, and mutation types with rates that depend on generation times (S1 text, and see Fig 5A). The model predicts an enrichment of uniform-clock mutations for substituted-primates with longer generation times. Notably, this translates into a prediction of an enrichment of uniform-clock mutations—like CpG transitions—in substituted species more closely-related to humans (with the exception of orangutan, which is thought to have the longest generation time among the primates considered, although it was only estimated in females [74]). Examining the expected distributions of mutation rates in substituted-species sites, this enrichment leads to a skew towards higher mutation rates for more closely-related substituted-species (Fig 5B). Overall, this model provides an explanation by which mutational mechanisms underlie the observed correlation between the relatedness of the substituted species and the skew of its cSFS towards common variants. Finally, we asked whether additional causes beyond mutation rate variation might also contribute to the species trend across cSFS. To this end, we used a logistic regression model to examine whether the probability of the variant being rare is associated with the relatedness of the substituted species (see Materials and Methods). A model that included only the relatedness of the substituted species showed a perfectly correlated ordering of the two (Fig 6A, CpG transitions were excluded from this analysis). We then turned to examine whether this correlation persists after controlling for mutational composition differences between substituted-species categories. We controlled for the effect of mononucleotide mutation types on the probability of the variant being rare (Fig 6B). We then further refined the mononucleotide mutation types by using their two flanking nucleotides, and estimated another model with these finer mutation type categories (Fig 6C). The trend persisted even after controlling for mutation type, most noticeably for nonsynonymous sites, which likely involve the strongest purifying selection pressures (Spearman ρ = 1, p = 0. 016 for the ordering of substituted-species coefficients for both models). We repeated the analysis while including CpG transitions, and found that the perfect correlation persisted (Spearman ρ = 1, p = 0. 016 for both models; S12 Fig). There are a few possible drivers of the residual trend observed in Fig 6B and 6C. First, there may be additional variation in mutation rate that is not accounted for by the 3-mer mutational context [32–39]. However, we note that the estimates in Fig 6 change only slightly between panels, suggesting that the unaccounted variation in mutation rates does not greatly bias the estimates. Furthermore, in S1 text we demonstrate that, in non-CpG sites, mutation rate distributions are expected to be very similar across substituted species categories. Therefore, additional mutation rate variation is not expected to contribute to the correlation between the relatedness of the substituted species and the fraction of rare variants. Another possibility is that differences in selection pressures across substituted-species categories contribute to the cSFS trend. It is possible that substituted-species sites of more closely related species are on average less deleterious. Furthermore, they may be less deleterious particularly for humans. More-related species have, on average, more similar context on which the mutation occurs. When the sequence context of the substituted species is similar to that of humans, the fixation of the human-minor allele in the substituted species suggests that the mutation is benign for humans. As sequence context diverges, epistatic effects may come into play and change the selective effect of the mutation [28,75,76]. In S1 text, we investigate the effect of sequence context divergence more directly (see S8 Fig).
Our analysis showed a significant correlation between the probability that a variant is rare in humans and the relatedness of another species in which the same mutation occurred. This trend was largely driven by mutation rate variation, which we have observed to be a primary determinant of the human SFS. The large effect that mutation rate variation has on the human SFS could have a major impact on any future work involving human polymorphism datasets with large sample sizes. For example, most demographic inference algorithms that use the SFS as a summary statistic [e. g. 6,65,77] rely on the infinite-sites model, which is evidently not a valid assumption for large samples. Adjusting demographic inference schemes to include the effects of recurrent mutations on the SFS (for examples of recent efforts towards this goal, see [78–82]) has the potential to significantly improve inference accuracy. We have also seen that the trend across cSFS persisted even after tri-nucleotide mutational composition was taken into account. This remaining correlation is consistent with differences in selection pressures across substituted-species categories. Substitutions in other lineages have proven to be highly informative for understanding deleterious effects in the contemporary human genome; among numerous features that have been considered, the strongest predictors of the pathogenicity of a mutation are species divergence features [83–86]. Nevertheless, methods used to predict the deleteriousness of a mutation at a site typically rely on a single summary of how variable a site is across the phylogenetic tree. Our analysis suggests that the location of a mutation on the evolutionary tree is informative of how deleterious the same mutation is for humans. It is our hope that the integration of divergence patterns and sequence context into methods that predict the fitness or health effects of human mutations could increase accuracy and predictive power.
For polymorphism data, we downloaded single nucleotide polymorphism (SNP) data from version 0. 2 of the Exome Aggregation Consortium database [40]. This database is a standardized aggregation of several exome sequencing studies amounting to a sample size of over 60,700 individuals and approximately 8 million SNPs. For each SNP we extracted upstream and downstream 30 nucleotides in the coding sequence of the human reference genome hg19 build. For simplicity, we excluded sites that are tri-allelic (6. 5% of all SNPs) or quad-allelic (0. 2% of all SNPs). For divergence data, we used the following reference genome builds downloaded from the UCSC genome browser [41]: chimpanzee (panTro4), gorilla (gorGor3), orangutan (ponAbe2), gibbon (nomLeu1), macaque (rheMac3), and baboon (papHam1). We used the UCSC genome browser’s liftOver program to align each ExAC SNP along with its 60bp sequence context to the six aforementioned reference genomes. We used the baboon reference genome solely for the ascertainment of all other substituted-species categories (rather than including a substituted-baboon category in the analysis). For gene annotations, we downloaded the refGene table of the RefSeq Genes track from the UCSC genome browser. For each SNP in our data, we extracted all gene isoforms in which the position was included. We kept all ExAC SNPs that fell in a coding exon, intron or untranslated region. We excluded from the analysis non-autosomal SNPs, SNPs that had multiple annotations corresponding to different transcript models, and SNPs with a sample size of less than 100,000 chromosomes. After applying the filters, we were left with 6,002,065 SNPs. For recombination rates, we downloaded the sex-averaged recombination rate map constructed by Kong et al. [87], which estimates rates at a resolution of 10kbp bins. In order to construct an upper bound on the probability of a human polymorphic site being ancestrally shared with another species, we consider the case of a selectively neutral polymorphism shared with chimpanzees. A polymorphism observed in the human sample at the current time is an ancestral polymorphism at the time of the human-chimpanzee split only if there are at least two lineages ancestral to the human sample at the human-chimpanzee split time. Leffler et al. [42] assume a constant human effective population size of 10,000 people throughout history, and estimate a probability of about 1. 0x10−5. In S1 text, we augment Leffler et al. ’s approximation with more complex demographic models for recent human history and derive an upper bound of 1. 4x10−5 for this probability. Multiplying this probability by the number of exonic sites (3,531,936) in our data, we get an expected number of 49 sites in our data that are ancestrally shared with chimpanzees. However, our derivations are based on a pre-out-of-Africa effective population size (NepOOA) of 10,000 people. Very little is known about human demographic history prior to the out-of-Africa event, and as we show in S1 text, the probability of an ancestral polymorphism rises very quickly with increasing NepOOA. Estimates of NepOOA range between 7300 [88] and 12,500 [89] people and are continually revised as estimates of human mutation rate and demographic history are refined. With NepOOA=12,500, we get an upper bound of 283 polymorphisms in the dataset that are expected to be shared between human and chimpanzee, which compose at most 1. 2% of the substituted-chimpanzee sites. The upper bound for other species, or sites under purifying selection, should be even smaller, and are overall too few to affect our results. We therefore conclude that ancestral polymorphisms are too few to significantly affect our analysis. To get a theoretical expectation for the fraction of rare variants under different mutational models, we used various software for computing the expected sample SFS of 33,750 diploid individuals, corresponding to the size of the non-Finnish European subsample in the ExAC dataset. For all mutational models, which we describe below, we generated predictions under various demographic models from recent literature: Gazave et al. [90] (model 2 in their work), Tennessen et al. [16] and Nelson et al. [14]. For the infinite-sites model, we computed the expected sample SFS analytically using fastNeutrino [65]. The infinite-sites model corresponds to an upper bound for the fraction of rare variants, but nonetheless predicted a fraction of rare variants much lower than that observed in data (75%-78%) for all non-CpG mutations under the Gazave et al. (59%) and Tennessen et al. (60%) demographies. The Nelson et al. model, which was inferred using a larger sample size of 11,000 people predicted 75% of biallelic polymorphisms would be rare under the infinite-sites model. In order to fit the highest observed fraction of rare variants for non-CpG sites in the ExAC data, we modified the parameters of the most recent epoch of exponential growth in Nelson et al. We estimated these parameters using fastNeutrino [65] on all A->C intronic mutations from ExAC. The inferred parameters were: current effective population size of 4,009,877 diploids, and an exponential growth onset time of 119. 47 generations in the past with a growth rate of 5. 38% per generation. The more ancient demographic parameters were fixed to the same values as in the model of Nelson et al. We assume multiple mergers (non-Kingman merger events) have negligible effect on the SFS since the sample size is significantly smaller than the current effective population size. A similar demographic model [91] with a four-fold smaller current effective population size exhibited a relative difference of only about 1. 3% and 0. 3% in the proportion of singletons and doubletons respectively for a comparable sample size of 50,000 people. Hence, we felt confident in using the Kingman coalescent for drawing genealogies. For the finite sites model, we first simulated independent coalescent trees using ms [92] and then generated 1kb non-recombining sequences for each coalescent tree using the desired recurrent mutation rate with the 4 x 4 Jukes-Cantor model of mutation [59]. We used the program Seq-Gen [93] to drop recurrent mutations on coalescent trees drawn from ms. We used mutation rates in a uniform logarithmically spaced grid of 40 points ranging from 10−9 to 5. 3x10−5 mutations per basepair per generation per haploid. For each value of the mutation rate, we simulated enough sequence data so that at least 100,000 biallelic polymorphic sites were available to reliably estimate the expected fraction of rare variants. If we indicate whether a variant is rare by Y, then for each mutation rate μ, the expected fraction of rare variants is E[Y|S=1; μ], where S is an indicator variable indicating whether a site is polymorphic and, specifically, biallelic. Finally, we considered a model with additional, within-mutation-type heterogeneity in mutation rate. Specifically, we considered a model in which sites of a particular mutation type (e. g. , C->A sites) have a mean mutation rate μ as before, but the mutation rate itself, M, is no longer fixed (and equal to μ), but rather a random variable with mean μ. Let f (M|S = 1; μ) be the probability density function of M in a site with mean mutation rate μ conditional on it being biallelic. Then, by the law of total expectation we have: E[Y|S=1; μ]=∫E[Y|M, S=1]f (M|S=1; μ) dM. By Bayes’ rule, f (M|S = 1; μ) is determined by both the within-mutation-type distribution of mutation rates, g (M; μ), and the probability of a site with mutation rate M being biallelic, as follows: f (M|S=1; μ) =P (S=1|M) g (M; μ) P (S=1; μ). Therefore, E[Y|S=1; μ] =∫E[Y|S=1, M]P (S=1|M) g (M; μ) ∫P (S=1|M′) g (M′; μ) dM′dM. For a large range of M, we have already estimated E[Y|M] as described above. From the same simulations we have estimated the probability of a site with mutation rate M being a biallelic polymorphism, P (S = 1|M). Lastly, the distribution of mutation rates due to within-mutation-type variance was modeled using a lognormal distribution: log10M; μ ~ N (log10μ−σ22ln (10), σ2). The mean parameter in the lognormal distribution above ensures that E[M] = μ. σ was arbitrarily chosen to be 0. 57 (red line in Fig 4A). Notably, Hodgkinson et al. also fit a lognormal distribution of mutation rates to their dataset of co-occurrence of SNPs in chimpanzees and humans, and estimate a similar value of σ = 0. 83 for non-CpG mutations [63] and σ = 0. 8 CpG transitions (personal correspondence). We tested whether the species trend across cSFS is due solely to the effect of mutation rate variation. We used a logistic regression model to examine whether a residual substituted-species trend remains after controlling for mutation type. Let Y be a binary-valued random variable indicating whether a variant is rare, μ→ be a vector of mutually exclusive indicator (dummy) variables for each mutation type, s→ be a vector of mutually exclusive indicator variables for the divergence pattern for the variant (substituted in one of the primates or human-private) and Z be an indicator of whether the variant is nonsynonymous (we only considered coding variants). We fitted the logistic regression model logit (P (Y=1|μ→, s→, Z) ) =β0+β→μ⋅μ→+ (1−Z) β→ssyn⋅s→+Zβ→sns⋅s→, where the parameters β0, β→μ, β→ssyn, and β→sns were learned from the data. We tested whether the coefficients β→ssyn, β→sns exhibit a trend across s, i. e. whether the probability of the variant being rare is associated with the relatedness of the substituted species. When ignoring the mutation rate effect (i. e. fixing β→μ≡0), the β→sns estimates were perfectly anti-correlated with the relatedness of the substituted species to human, consistent with the observation in data (Fig 6A). We then allowed for an effect for the mutation type by estimating β→μ for the different categories of mononucleotide mutation types (Fig 6B). We also estimated a model with a finer resolution of mutational categories, further partitioning the mononucleotide mutation types by their two flanking nucleotides (Fig 6C). For nonsynonymous sites, which likely involve the strongest purifying selection pressures, the trend persisted even after controlling for mutation rate variation (Spearman ρ = 1, p = 0. 016 for both mononucleotide correction and for the correction including flanking nucleotides context). | The site frequency spectrum (SFS, i. e. , the distribution of allele frequencies) is a summary of natural variation, used to study demographic history and natural selection. Here, we extended the SFS by conditioning on phylogenetic divergence patterns. We refer to this extension as the “phylogenetically-conditioned SFS” or cSFS. Exploring the determinants of the cSFS revealed two main findings. First, we found that mutations that have been independently fixed in another species are more likely to be benign for contemporary humans if the fixation occurred in a closely related species. The background on which a mutation occurs is therefore an important feature to consider when predicting the fitness consequences of a mutation. Second, we found that the SFS is substantially affected by repeat mutations within the human population. The extent of repeat mutations implies that some sites must have particularly high mutation rates, beyond the known variation across the different possible nucleotide changes. Our observations contradict the “infinite sites” mutation model, which is commonly used in population genetic analyses, and imply that future SFS-based analyses of human populations should account for mutation rate variation. | Abstract
Introduction
Results
Discussion
Materials and Methods | demography
computational biology
vertebrates
animals
mammals
primates
substitution mutation
mutation
genome analysis
mammalian genomics
epigenetics
dna
chromatin
dna methylation
research and analysis methods
chromosome biology
gene expression
biological databases
chromatin modification
dna modification
animal genomics
people and places
biochemistry
cell biology
nucleic acids
database and informatics methods
genetics
biology and life sciences
apes
genomics
amniotes
genomic databases
organisms
orangutans | 2016 | Mutation Rate Variation is a Primary Determinant of the Distribution of Allele Frequencies in Humans | 7,574 | 260 |
Currently, there is a growing interest in ensuring the transparency and reproducibility of the published scientific literature. According to a previous evaluation of 441 biomedical journals articles published in 2000–2014, the biomedical literature largely lacked transparency in important dimensions. Here, we surveyed a random sample of 149 biomedical articles published between 2015 and 2017 and determined the proportion reporting sources of public and/or private funding and conflicts of interests, sharing protocols and raw data, and undergoing rigorous independent replication and reproducibility checks. We also investigated what can be learned about reproducibility and transparency indicators from open access data provided on PubMed. The majority of the 149 studies disclosed some information regarding funding (103,69. 1% [95% confidence interval, 61. 0% to 76. 3%]) or conflicts of interest (97,65. 1% [56. 8% to 72. 6%]). Among the 104 articles with empirical data in which protocols or data sharing would be pertinent, 19 (18. 3% [11. 6% to 27. 3%]) discussed publicly available data; only one (1. 0% [0. 1% to 6. 0%]) included a link to a full study protocol. Among the 97 articles in which replication in studies with different data would be pertinent, there were five replication efforts (5. 2% [1. 9% to 12. 2%]). Although clinical trial identification numbers and funding details were often provided on PubMed, only two of the articles without a full text article in PubMed Central that discussed publicly available data at the full text level also contained information related to data sharing on PubMed; none had a conflicts of interest statement on PubMed. Our evaluation suggests that although there have been improvements over the last few years in certain key indicators of reproducibility and transparency, opportunities exist to improve reproducible research practices across the biomedical literature and to make features related to reproducibility more readily visible in PubMed.
There is a growing interest in evaluating and ensuring the transparency and reproducibility of the published scientific literature. According to an internet-based survey of 1,576 researchers in Nature, 90% of respondents believe that there is either a slight or significant crisis of reproducibility in research [1]. However, multiple recent efforts are attempting to address some of the existing concerns [2–6]. These initiatives, as well as previous proposals by several stakeholders to change scientific practice, may be resulting in genuine improvements in the transparency, openness, and reproducibility of the scientific literature. A survey of a random sample of biomedical articles published from 2000–2014 suggested that the literature lacked transparency in important dimensions and that reproducibility was not valued appropriately [7]. For instance, protocols and raw data were not directly available, and the majority of studies did not disclose funding or potential conflicts of interest. Furthermore, over half of the articles in the sample claimed to present some novel discoveries and the vast majority did not have subsequent studies that were attempting to replicate part or all of their findings [7]. These results suggested that there is significant room for improvement with regard to reproducible research practices. Furthermore, the study provided baseline data to compare future progress across key indicators of reproducibility and transparency. Since 2014, there have been new or intensified efforts to promote open science practices across the biomedical literature. Although it is unlikely that individual interventions have single-handedly resulted in drastic changes, these efforts may cumulatively reflect a gradual shift toward the adoption of a culture that embraces transparency and replication. For instance, in January 2015, the Institute of Medicine issued a report that recommended that all stakeholders in clinical trials “foster a culture in which data sharing is the expected norm, ” and that funders, sponsors, and journals promote and support data sharing [8]. The International Committee of Medical Journal Editors (ICMJE) also proposed a policy requiring data sharing as a condition of publication, even though no formal policy changes have been enacted [9,10]. Other stakeholders have also supported raw data sharing [2] and some journals have started requesting full protocol sharing [11], since access to detailed protocols is necessary to allow study procedures to be repeated [12]. Several fields are paying more attention to replication, especially after the findings of reproducibility checks demonstrated concerning results in psychology [13] and cancer biology [14,15]. Furthermore, a growing number of journals have started to require reporting guidelines and disclosure statements, and commercial and nonprofit organizations, such as the Open Science Framework (http: //osf. io), have introduced new infrastructure supporting research transparency. Additional efforts have also tried to improve the disclosure and visible indexing of information related to transparency and reproducibility. In 2017, PubMed, which is run by the United States National Library of Medicine (NLM) at the National Institutes of Health (NIH), started including funding and conflicts of interest statements with study abstracts. Although this information is often disclosed in the full text of journal articles, many research consumers do not have a subscription to all of the journals catalogued in PubMed. To our knowledge, it is unknown whether information about key transparency indicators is easily accessible to the general public on PubMed and whether this information was available prior to 2017. These and other recent open science initiatives, or even simply the wider sensitization of the scientific community over the past 20 years, may have improved the reproducibility and transparency of the biomedical research over the last few years. However, to our knowledge, there is no evidence on whether progress has been made on all, some, or none of the indicators that have been proposed as being important to monitor [5,7]. Given the importance of examining the progress of reproducibility and transparency in the scientific literature, we sought to build upon our previous analysis [7] and to assess the status of reproducibility and transparency in a random sample of biomedical journal articles published between 2015 and 2017. Here, we evaluate the proportion of studies reporting sources of public and/or private funding and conflicts of interest, sharing protocols and raw data, and undergoing rigorous independent replication and reproducibility checks. We also investigate what can be learned about these reproducibility and transparency indicators from widely accessible open access data provided on PubMed.
Among the 155 randomly selected articles published between 2015 and 2017, we excluded 6 non-English language articles. Of the remaining 149,68 (45. 6% [95% confidence interval, 37. 5% to 54. 0%]) were publications in the research field of Medicine, with smaller numbers in the fields of Health Sciences (n = 28), Biology (n = 13), Infectious Disease (n = 16), and Brain Sciences (n = 24). Among 120 articles that were published in a journal with a 2013 impact factor, the median impact factor was 3. 1 (interquartile range, 2. 0–4. 7). The majority of publications had some form of empirical data (118 of 149 [79. 2% (95% confidence interval, 71. 6% to 85. 2%) ]—n = 104 excluding case studies and case series, in which protocol and raw data sharing may not be pertinent, and n = 97 excluding also systematic reviews, meta-analyses and cost-effectiveness analyses in which replication in studies with different data would not be pertinent). Among the 149 eligible articles, there was one (0. 7% [0. 0% to 4. 2%]) cost-effectiveness or decision analysis, 14 (9. 4% [5. 4% to 15. 6%]) case studies or case series, four (2. 7% [0. 9% to 7. 2%]) randomized clinical trials, six (4. 0% [1. 6% to 8. 9%]) systematic reviews and/or meta-analyses, and 92 (61. 7% [53. 4% to 69. 5%]) “other” articles with empirical data (including cross-sectional, case-control, cohort, and various other uncontrolled human or animal studies). Approximately one-fifth (20. 8% [14. 8% to 28. 4%]) of the sample was classified as research without empirical data or models/modeling studies. There were 64 (43. 0% [35. 0% to 51. 3%]) with a PubMed Central reference number (PMCID), of which 37 were also PubMed Central Open Access (PMCOA). Nearly one-third (46,30. 9% [23. 7% to 39. 0%]) of the 149 biomedical articles did not include information on funding. There were 78 articles (52. 3% [44. 0% to 60. 5%]) that were publicly funded, either alone or in combination with other funding sources. Of these, three received National Science Foundation (NSF) support and 25 had NIH funding, either alone or in combination with other funding sources. Among the 149 articles, there were 52 (34. 9% [27. 4% to 43. 2%]) that did not include a conflicts of interest statement. However, there were 87 (58. 4% [50. 0% to 66. 3%]) that specifically reported no conflicts of interests and 10 (6. 7% [3. 4% to 12. 3%]) that included a clear statement of conflict. Excluding case studies or case series and models/modeling studies, in which a protocol would not be relevant, one (1. 0% [0. 1% to 6. 0%]) of the 104 articles with empirical data included a link to a full study protocol. This article was a systematic review that stated that “methods for study inclusion and data analysis were prespecified in a registered protocol (PROSPERO 2015: CRD42015025382) ” (PMID: 27863164) [16]. There was also one clinical trial (27391533) and two prospective cohort studies (25682436 and 28726115) that referenced a ClinicalTrials. gov identifier (i. e. , an NCT number). For two of the studies (27391533 and 28726115), the month and year in which sponsors or investigators first submitted a study record to ClinicalTrials. gov were the same as the reported study start dates. For one of the observational studies (25682436), the first ClinicalTrials. gov study record date was approximately 11 years after the disclosed study start date. There were 31 (29. 8% [21. 4% to 39. 7%]) articles that included supplemental materials, including methods sections, videos, tables, survey materials, and/or figures, either as a detailed appendix at the end of the article or online. However, none of the supplementary materials allowed for a reconstruction of a full protocol. Furthermore, none of the articles mentioned any sharing of scripts/code. There were 19 (19 of 104,18. 3% [11. 6% to 27. 3%]) articles that discussed some level of publicly available data (Table 1). While 13 provided data set identifiers or accession codes, there were four articles that included supplementary excel data files. Although another article mentioned that all relevant data were within the supporting information files, the supplementary files did not contain any raw data (26413900). Among the 97 biomedical articles with empirical data, excluding case studies and case series, systematic reviews/meta-analyses, and cost effectiveness/decision analyses studies, only five (5. 2% [1. 9% to 12. 2%]) were inferred to be replication efforts trying to validate previous knowledge. Over half (56,57. 7% [47. 3% to 67. 6%]) claimed to present some novel findings. Although 10 (10. 3% [5. 3% to 18. 6%]) articles had statements of both study novelty and some form of replication, 26 (26. 8% [18. 6% to 36. 9%]) had no statement or an unclear statement in the abstract and/or introduction about whether the article presented novel findings or replication efforts. Of the 97 biomedical articles with empirical data, there were two articles that had at least some portion of their findings replicated. One of the replicating articles used an “almost comparable study design but over a longer period” and included some patients with different characteristics (Index article: 24415438, replication: 27363404) [17]. The second was a partial replication effort with a longer follow-up (Index article: 27067885, replication: 27241577). Only one article was included in a subsequent systematic review. As shown in Table 2, there were no statistically significant differences between PMCOA and non-PMCOA articles or between articles with a PMCID or without a PMCID. However, there was a suggestion for fewer articles having a statement of no conflicts of interest (p = 0. 014) and more articles including a statement pertaining to data sharing (p = 0. 049) in the PMCOA group than in the non-PMCOA group. Furthermore, there was a suggestion that articles without PMCIDs were less likely to mention funding and to have public funding than in the PMCID group (p = 0. 015). Among the 590 articles published between 2000 and 2017 in eligible research fields directly related to biomedicine, 520 were non-PMCOA articles. Among the 520 non-PMCOA articles, 81 (15. 6% [12. 6% to 19. 1%]) had a PMCID, thus a PDF is available for each individually. However, full text XML (i. e. , Extensible Markup Language) for these articles cannot be downloaded in bulk. Therefore, 439 articles did not have a full text available in PubMed. Of the 439 eligible articles, 184 listed funding sources at the full text level. Nearly two-thirds (115 of 184,62. 5% [55. 0% to 69. 4%]) included some funding information under the “Publication type, MeSH terms, Secondary source ID” tab on PubMed (e. g. , “Research Support, Non-US Gov’t”). There were 39 (21. 2% [15. 7% to 28. 0%]) additional articles in which PubMed provided at least one specific funding source (i. e. , a specific grant number). None of the articles disclosed competing interests under a “Conflict of interest statement” tab on PubMed. Among the 263 articles in which protocol or data sharing would be relevant (excluding articles without some form of empirical data, model/modeling studies, and case studies or case series), there was one systematic review with a registered protocol on PROSPERO that included a University of York Centre for Reviews and Dissemination (CRD) number at the abstract level. There were six articles that either referenced their clinical trial’s identifier, included a link to ClinicalTrials. gov, or stated that a Clinical Trials repository link was available on the journal website. Five of these articles also had clinical trial identifiers at the PubMed level. Among the 11 articles that discussed supplementary data, database identifiers, or claimed that data were available upon request, two referenced data identifiers or accession numbers in their abstract (PMID: 22224476 [GenBank links under the “Publication type, MeSH terms, Secondary source ID” tab on PubMed], 27871817 [included links to the SILVA Database under the “LinkOut—more resources” tab on PubMed]). Of 252 articles with empirical data (excluding case studies and case series, systematic reviews/meta-analyses, and cost effectiveness/decision analysis studies) published between 2000 and 2017, five did not have an abstract on PubMed. Among the eight articles classified as partial or full replication studies based on information provided in the abstract and/or introduction, four had enough information in the abstract alone to establish whether they were replication studies. Approximately half (55 of 123,44. 7% [35. 8% to 53. 9%]) of the articles claiming to present some novel findings based on the abstract and/or introduction could be classified as novel according to the abstract only. Of the 10 articles that had statements of both study novelty and some form of replication, only four could be classified based on the abstract only. A comparison of articles published from 2000–2014 versus 2015–2017 revealed some distinctive patterns (Table 3). Articles published between 2000 and 2014 were less likely to include information related to funding (p = 1. 4 × 10−5). The proportion of articles including information on funding increased over time, with apparently more rapid changes occurring after 2014 (Fig 1). While recently published articles were more likely to contain conflicts of interest statements (p = 2. 5 × 10−13), the proportion of articles with information on conflicts of interests seems to have increased steadily over time (Fig 1). Availability of data substantially increased in 2015–2017 (p = 9. 7 × 10−8), with the proportion of articles including a statement regarding data sharing increasing since 2015 (Fig 2). However, there were no major changes in availability of full protocols. Furthermore, there were more replication attempts published in recent years (either alone or combined with addition novel analyses) (p = 3. 0 × 10−4) (Fig 3). Although the proportion of articles reporting novel findings has remained fairly constant since 2000, there has been a decrease in the proportion of studies with either no or an unclear statement in the abstract and/or introduction about whether there were any novel findings or replication efforts. As expected, fewer articles published in 2015–2017 had already been incorporated in systematic reviews and meta-analyses (given the limited time span available) (p = 8. 9 × 10−4). Open access (PMCOA articles and those with PMCID) proportionally increased substantially in 2015–2017 (p = 6. 7 × 10−8 and p = 1. 2 × 10−7, respectively).
Our study has certain limitations. Our evaluation relied on published biomedical research information. Therefore, it is possible that additional protocols, raw data, and clarifications on conflicts or funding could be established by contacting authors, journals, or sponsors. Second, our study relied on published records. This means that we based our determination of novelty on the information reported by investigators. For instance, it is possible that authors may have tried to spin their manuscript as being more novel than it really is in order to ensure publication. Although we used our best judgment to classify articles and two authors discussed uncertainties before agreeing upon a final classification, certain decisions were more subjective [7]. Moreover, since only one author conducted the data abstractions, we were unable to calculate any inter-rater reliability metrics. However, the primary abstractor for these data was the coprimary abstractor in a previous study evaluating the same indicators of transparency and reproducibility in articles published in the biomedical literature [7]. Therefore, the primary abstractor of the current evaluation has had extensive experience analyzing these indicators and had an already streamlined process to do so. Nevertheless, when determining study novelty and replication for articles from diverse biomedical fields, difficulty arose assessing whether study results were actually groundbreaking, full or partial replication efforts, or being fully replicated by subsequent studies. In order to account for these limitations, all uncertainties were discussed by two investigators (JDW and JPAI). Third, we did not perform any sample size calculations. Our study evaluated multiple indicators that were all equally important, and they varied substantially in the proportion to which they were satisfied already by the articles in the 2000–2014 sample. The number of annual published biomedical articles increases at approximately 5% per year, and our sample ensured that 2015–2017 would be as well represented as previous years, accounting also for an increase in the volume of published literature over time. Fourth, we acknowledge that the sampling method for the recent set of articles was not identical to the sampling method for the original set of articles. However, when we applied the new enhanced field-classification method based on article-level classification to the original set of 441 articles, we found that 421 were in common between the original and new classifications. With approximately 95% overlap in biomedical definitions between the two samples, we are confident that our population of articles from which the sample was drawn and the sampling methods are comparable. Fifth, our analyses were based on a random sample of 149 biomedical articles published between 2015 and 2017. Therefore, we were unable to account for potential differences in reporting practices across various fields and subdisciplines. Future evaluations should assess these indicators within specific fields. Improvements over time may reflect improvements within specific fields, across many/all fields, and/or an increased representation of the most transparent fields in the more recent literature. Sixth, it is worth noting that we focused on key indicators of reproducibility and transparency that have been proposed as important to monitor. In particular, these indicators were established based on a series of five papers about research published in the Lancet [5]. However, these indicators serve as a proxy for transparency and reproducibility and do not capture all potential areas where open science advances may have been made. Finally, an additional limitation is that this study required manual examination of publications and coding of data. We are hopeful that algorithmic means to extract similar information from full text sources (such as PMCOA) can be developed to enable larger scale analyses in the future. Our empirical evaluation of biomedical articles published between 2015 and 2017 suggests that progress has been made improving key indicators of reproducibility and transparency. We found that a greater proportion of articles included information on funding, had a conflicts of interest disclosure statement, discussed or publicly shared some portion of their data, and claimed or were inferred to be replication efforts trying to validate previous knowledge. While clinical trial identification numbers and funding details were often provided on PubMed, details related to data sharing and conflicts of interest statements were generally not disclosed. Although numerous efforts to improve reproducibility have already been adopted by researchers, journals, and funders, additional efforts will be necessary to continue to sensitize key stakeholders in the research enterprise of the importance of continuing to improve these indicators over time.
We used a sampling process to generate a new random sample of 155 articles published between 2015 and 2017 and indexed in PubMed. We did not perform any sample size calculations since our study evaluated multiple indicators that were all equally important, and they varied substantially in the proportion to which they were satisfied already by the articles in the 2000–2014 sample. Our sample of 155 articles ensured that 2015–2017 would be as well represented as previous years, accounting for the fact that the number of annual published biomedical articles increases approximately 5% per year (Table 4). The sample of 155 articles for the years 2015–2017 was at least 1. 5 times that for any other 3-year period from 2000–2014 (Table 4). Articles classified as a “Journal Article” in PubMed were considered and then ordered randomly. Articles in scientific fields not directly related to biomedical research (defined as Biology/Biotechnology, Medicine, Infectious Disease, Health Sciences, and Brain Sciences) [39] were excluded. Even though these fields may sometimes have repercussions for biomedicine, their transparency practices may differ systematically, and separate evaluation efforts would be necessary [7]. All non-English language articles were then excluded and one investigator (JDW) independently characterized the new sample into seven study categories, as previously described (Box 1) [7]. We also considered a previous sample of 441 English language journal articles published between 2000 and 2014 [7] for a comparison against the newer articles and for combined analyses of indicators in terms of open source data. Sampling for the recent set (2015–2017) of papers was done in a manner to produce a set that, given data availability, was as similar as possible to the original set (2000–2014) to enable comparison. Both sets were chosen randomly based on PubMed identification (PMID) numbers. Although both samples were limited to articles considered to be in biomedical fields, in the current analyses, we used an enhanced field classification process based on article-level classification [40], which allowed for a better categorization of both the 155 new articles and the 441 previous articles. Furthermore, the recent sample was limited to “articles” only. While this information was not in place for the original set, only 19 of the 441 articles in the original set were designated as “letters” in PubMed rather than articles [7]. In order to determine whether there are different reporting practices among free full text articles, we identified the subset of articles made available through PubMed Central (PMC), a digital repository that archives publicly accessible full text biomedical and life science journal articles. Availability of free access in PMC was based on assignment of a PMC identifier (i. e. , PMCID versus non-PMCID articles). We also classified articles based on whether there was a publicly available XML version of the full text of the article in the open access subset of PMC (i. e. , PMCOA versus non-PMCOA articles). The XML of the full text of roughly 1. 7 million PMCOA articles is available in bulk, which allows for algorithmic analyses of the data at scale, as opposed to one at a time. Since 2015, PMCOA comprises roughly half of PMC articles and over 20% of all PubMed articles [41]. We aimed to compare the key indicators of reproducibility across the different article types. In order to maintain consistency with our previous evaluation [7], we determined the 2013 impact factor of each publication’s journal. The journal name for the eligible article was searched in InCites Journal Citation Reports. No information was recorded for journals without a 2013 impact factor. As previously reported in our survey of biomedical research published between 2000 and 2014 [7], full articles with data and analyses were examined for statements of conflicts of interest, funding disclosures, and publicly available full protocols and data sets. In particular, we reviewed the final versions of the articles available online. For published articles without data and analyses, only statements of conflict and funding were investigated, since protocols, data sets, and reproducibility were not relevant [7]. These indicators, which were assessed in a previous evaluation of articles published in the biomedical literature between 2000 and 2014 [7], have been proposed as being important to monitor in relation to transparency and reproducibility. According to “Increasing value and reducing waste in research design, conduct, and analysis” [5], one of five papers on “Research: Increasing value, reducing waste” published in the Lancet, there are several key issues necessary to improve the research process. Under the recommendations section, the authors note that it is necessary to monitor the proportion of research studies “with publicly available (ideally preregistered) protocol and analysis plans, and proportion with raw data…, ” “without conflicts of interests, as attested by declaration statements and then checked by reviewers, ” and “undergoing rigorous independent replication and reproducibility checks” [5]. As suggested during peer review, we also determined the proportion of articles with (1) statements related to the sharing of script/code by searching the full text of the articles for the words “supporting, ” “supplement, ” “appendix, ” “code, ” and “script, ” respectively, and (2) any supplemental materials. The abstracts and introductions were then evaluated for patterns of reproducibility (e. g. , whether the study claimed to be a replication effort [Box 2]), as previously described [7]. Web of Knowledge was utilized to determine whether subsequent citing articles had tried to replicate the analyses and whether data were included in systematic reviews and/or meta-analyses. For all articles without a PMCID (i. e. , non-PMCID articles) published between 2000 and 2017, we repeated all assessments of indicators of reproducibility and transparency using information reported at the PubMed level. Key indicators of transparency and reproducibility, such as funding and conflicts of interest statements, are often disclosed in the full text of journal articles. However, many research consumers do not have a subscription to all of the journals catalogued in PubMed. Therefore, we evaluated whether PubMed can be used to identify these indicators among the subset of articles without a free full text. Since articles in the PMCOA subset are of particular interest to meta-researchers who want to download information en masse, we repeated our analyses stratified by whether the XML of the full text was publicly available (i. e. , PMCOA versus non-PMCOA articles). We examined the title, abstract, “MeSH terms” tab, and the “LinkOut—more resources” tab on PubMed for each article. This captures the metadata-level information that is available in PubMed. We note that article-level metadata can also be downloaded from PubMed in bulk in various formats and that these metadata are amenable to algorithmic mining of information. Using descriptive statistics, we characterized the indicators of transparency for the period 2015–2017. A priori established Fisher’s exact tests were used to examine differences between the 2000–2014 and 2015–2017 samples, PMCOA and non-PMCOA articles, and PMCID and non-PMCID articles; all statistical tests were two-tailed. As suggested during peer review of our work, we also analyzed potential changes over time for certain indicators of reproducibility and transparency. In particular, we plotted 3-year moving proportions for indicators with an adequate number of events against time. For instance, for the year 2013, we calculated the proportion of articles with a data sharing statement between 2012 and 2014. These analyses can explore more gradual changes that could have occurred. Analyses were performed using R (Version, 3. 2. 3: The R Project for Statistical Computing). We used the P < 0. 005 threshold for statistical significance [2,42], calling results with P values 0. 05 to 0. 005 suggestive. | Currently, there is a growing interest in ensuring the transparency and reproducibility of the published scientific literature. According to a previous evaluation of 441 biomedical articles published from 2000–2014, the majority of studies did not share protocols and raw data or disclose funding or potential conflicts of interest. However, multiple recent efforts, which are attempting to address some of the existing concerns, may be resulting in genuine improvements in the transparency, openness, and reproducibility of the scientific literature. In this study, we investigate the reproducibility and transparency practices across the published biomedical literature from 2015–2017. We analyze reporting of public and/or private funding and conflicts of interests, sharing protocols and raw data, and independent replication and reproducibility checks. We also investigate what can be learned about reproducibility and transparency indicators from open access data provided on PubMed. Our evaluation suggests that although there have been improvements over the last few years in some aspects of reproducibility and transparency (e. g. , data sharing), opportunities exist to improve reproducible research practices across the biomedical literature and to make features related to reproducibility more readily visible in PubMed. | Abstract
Introduction
Results
Discussion
Materials and methods | open science
open access publishing
research funding
publication practices
replication studies
research design
research and analysis methods
sequence analysis
meta-research article
bioinformatics
government funding of science
biological databases
research assessment
sequence databases
science policy
systematic reviews
database and informatics methods
reproducibility
scientific publishing | 2018 | Reproducible research practices, transparency, and open access data in the biomedical literature, 2015–2017 | 6,516 | 240 |
The X chromosome (chrX) represents one potential source for the “missing heritability” for complex phenotypes, which thus far has remained underanalyzed in genome-wide association studies (GWAS). Here we demonstrate the benefits of including chrX in GWAS by assessing the contribution of 404,862 chrX SNPs to levels of twelve commonly studied cardiometabolic and anthropometric traits in 19,697 Finnish and Swedish individuals with replication data on 5,032 additional Finns. By using a linear mixed model, we estimate that on average 2. 6% of the additive genetic variance in these twelve traits is attributable to chrX, this being in proportion to the number of SNPs in the chromosome. In a chrX-wide association analysis, we identify three novel loci: two for height (rs182838724 near FGF16/ATRX/MAGT1, joint P-value = 2. 71×10−9, and rs1751138 near ITM2A, P-value = 3. 03×10−10) and one for fasting insulin (rs139163435 in Xq23, P-value = 5. 18×10−9). Further, we find that effect sizes for variants near ITM2A, a gene implicated in cartilage development, show evidence for a lack of dosage compensation. This observation is further supported by a sex-difference in ITM2A expression in whole blood (P-value = 0. 00251), and is also in agreement with a previous report showing ITM2A escapes from X chromosome inactivation (XCI) in the majority of women. Hence, our results show one of the first links between phenotypic variation in a population sample and an XCI-escaping locus and pinpoint ITM2A as a potential contributor to the sexual dimorphism in height. In conclusion, our study provides a clear motivation for including chrX in large-scale genetic studies of complex diseases and traits.
Genome-wide association studies (GWAS) have discovered a wealth of loci associated with complex phenotypes with almost 5,800 significant associations for more than 500 different phenotypes reported in the NHGRI GWAS catalog [1] (accessed August 13,2013). These GWAS discoveries are, however, concentrated on the autosomes leaving the sex chromosomes, especially the relatively large X chromosome (chrX), underrepresented; while chrX contains approximately 5% of genomic DNA, hence being comparable in size to chromosome 7, and encodes for more than 1,500 genes, only around 20 unique significantly associated X-chromosomal loci in total are recorded in the catalog. For instance, there are hundreds of known autosomal loci for height, BMI and blood lipids, but only one significant height locus has been identified in chrX, and this in individuals of African ancestry, and no X-chromosomal associations for these other highly polygenic phenotypes have been reported. Nevertheless, almost 1% of genetic variance in height and BMI has been shown to be accountable to chrX SNPs [2], demonstrating that common genetic variation in chrX contributes to complex phenotypes. A likely explanation for the dearth of association findings in chrX is that the chromosome is often neglected in GWAS: Wise et al. recently surveyed all published GWAS from 2010 and 2011 and found that only 33% of these studies had included chrX analyses [3]. While some association studies have opted for including chrX, such as recent genetic screens on sex-hormone binding globulin levels [4] and Grave' s disease [5], removal of non-autosomal data appears to be a common practice in GWAS [6], [7]. There are many potential reasons for the exclusion of chrX in GWAS, as outlined by Wise et al [3], a major contributor being that the analysis pipeline applied for autosomes is not directly applicable to chrX analyses. While women carry two copies of chrX, men are hemizygous for the chromosome. The allele dosages between the sexes are balanced by random X chromosome inactivation (XCI) that silences one of the two chromosomes in women, hence requiring the allele coding for chrX markers to be adjusted accordingly for the analyses. However, XCI does not evenly cover the whole of chrX, but approximately 15% of the loci in the chromosome completely escape from XCI and in further 10% of the sites the silenced chrX is variably active in women, although the expression from the inactivated copy of chrX is often lower than from the active chrX [8]. The incomplete XCI adds another layer of analytical challenges, yet at the same time it also makes chrX particularly interesting to study, as the regions of incomplete dosage compensation are among the genomic contributors to the differences between gene dosages in men and women. As such, these loci could partly explain phenotypic sexual dimorphisms and additionally contribute to the phenotypic characteristics observed in chrX aneuploidies. Given the underutilization of chrX data in previous studies and hence the potential for novel biological discoveries, we aimed at surveying the contribution of the chromosome to complex traits. To this end, we expanded the marker set for chrX by imputing the non-pseudoautosomal region of chrX in almost 25,000 Finnish and Swedish individuals from seven discovery and one replication cohort (Table 1) by utilizing the recently released comprehensive reference panel from the 1000 Genomes Project [9]. We focused our chrX-wide screen on twelve quantitative anthropometric and cardiometabolic phenotypes for which hundreds of autosomal, but no X-chromosomal, loci have been identified in GWAS of individuals of European ancestry, namely height, body-mass-index (BMI), waist-hip-ratio (WHR), systolic and diastolic blood pressure (SPB and DBP), C-reactive protein, fasting insulin and glucose, total, LDL and HDL cholesterol (TC, LDL-C and HDL-C) and triglycerides (TG). By using a linear mixed model, we show that the variation in chrX influences the levels of many of these complex phenotypes and in an association analysis identify and replicate three new associated X-chromosomal loci, one for fasting insulin and two for height, hence demonstrating the value of assessing chrX associations. Further, we find strong evidence for a lack of dosage compensation in one of the two associated height loci by applying a meta-analysis that allows for sex heterogeneity in effects and by a formal statistical model comparison between the different dosage compensation models given the observed data.
We first estimated the proportion of variance in each of the twelve phenotypes accountable jointly to the common and low-frequency (minor allele frequency (MAF) >1%) SNPs in chrX using a linear mixed model [2], [10] in the study samples for which the individual-level genotype data were available (Materials and Methods). ChrX variants were estimated to contribute to the levels of many of these phenotypes (Table 2): under the model of equal variance in males and females (see Materials and Methods for discussion about the models), more than 0. 5% of the variance in height, SBP, HDL-C, fasting glucose and insulin appear to be due to chrX variation, hence motivating the search for associated variants in chrX. The highest estimate for X-linked variance (1. 4%, P-value = 2. 00×10−6) was observed for height, a highly heritable and polygenic phenotype. For the other phenotypes the statistical significance of the estimates was not overwhelming, a result of the available sample size and lower trait heritability, but also the estimates for SBP, HDL-C and fasting insulin were significantly different (P-value<0. 05) from zero. In these four phenotypes, which showed non-zero X-linked variance, on average 4. 0% (range 2. 4%–6. 0%) of the estimated whole-genome genetic variance was attributable to chrX while the corresponding value over all twelve traits was 2. 6%. Following the work of Yang et al. [2] we calculated the variance estimates under three different models for dosage compensation, i. e. , equal variance between men and women, full dosage compensation and no dosage compensation (Table S1). The differences between the model fit were small (as measured by likelihood-ratios) and none of the models was consistently favored above the other two. This is likely due to our sample size being limited for such comparisons, but may also reflect the differences in the genetic architecture of the various loci in chrX that contribute to the variance for each phenotype. In order to identify X-linked loci contributing to the phenotypic variance we assessed associations between directly genotyped and imputed chrX SNPs and the twelve phenotypes across seven discovery cohorts (N = 19,697; Table 1). As the majority of the loci in chrX are subject to XCI, we adopted an allele coding that is consistent with the full dosage compensation model, i. e. we treated hemizygous men as equivalent to homozygous women (Materials and Methods). Within each cohort the associations were first studied separately for males and females using SNPTEST [11] and the results were subsequently combined in a fixed-effects meta-analysis in GWAMA [12]. By analyzing the associations of up to 405,411 polymorphic high-quality SNPs we identified three associated loci (P-value<5×10−8): two for height (both in Xq21. 1) and one for fasting insulin (in Xq23) (Table 3, Table S2, Figure 1). We followed-up these findings in an independent replication set and found further evidence for association in all three loci (discovery and replication combined up to N = 24,729), with all lead SNP P-values<6×10−9 (Table 3). In the more strongly associated height locus the associated SNPs (lead SNP rs1751138, joint, i. e. , discovery and replication combined, P-value = 3. 03×10−10, MAF = 0. 36) are located approximately 35 kb upstream of ITM2A (integral membrane protein 2A), a gene implicated in early cartilage development [13], [14]. We observed that the minor A allele of rs1751138, which is associated with shorter stature, is also associated with an increased expression of ITM2A in whole blood (P-value = 6. 23×10−14, N = 513; Materials and Methods), providing further evidence for ITM2A as a functional candidate gene for this association. The second region associated with height spans FGF16, ATRX and MAGT1. The lead SNP (rs182838724, joint P-value = 2. 71×10−9, MAF = 0. 30) is intronic within ATRX, a gene associated with the X-linked alpha thalassaemia mental retardation syndrome (ATR-X), a rare condition manifesting itself as profound developmental delay often accompanied by several other distinct characteristics including skeletal abnormalities in 90% and short stature in two thirds of the affected individuals [15]. As the two height lead SNPs map only 2 Mb apart, we confirmed that the associations are independent of each other by conditioning the association analysis on the lead SNP of the ITM2A locus. The conditional analysis did not attenuate the height signal in the ATRX region, yet here the most associated SNP (rs34979608, joint P-value = 1. 52×10−9, r2 with rs182838724 = 0. 91; Table S3, Figure S1) maps outside ATRX, 4 kb downstream of MAGT1, a gene encoding a magnesium transporter. Both of the height associations are present already in childhood (P-value for ITM2A = 1. 58×10−5, P-value for ATRX = 0. 00955, N = 3287, subset of two of the study cohorts, ages 8–10; Table S4), but for the ATRX locus the association appears weaker in children than in the same individuals in adulthood (beta in childhood = 0. 059, beta in adulthood = 0. 092, P-value for difference in effect sizes = 0. 043; Table S4) suggesting additional influence of puberty. In the third associated locus, the dosage of the minor G allele of rs139163435 (MAF = 0. 071) was robustly associated with lower levels of fasting insulin across all cohorts (joint P-value = 5. 18×10−9). The lead SNP maps to an apparent gene desert: the gene closest to the association, SLC6A14, a possible candidate gene for X-linked obesity (OMIM: 300444), lies more than 500 kb away. Up to 25% of the chrX loci may not be subject to complete dosage compensation [8], but in these regions also the inactivated X chromosomes are transcriptionally fully or partially active. When adopting an allele coding that is consistent with the full dosage compensation model, as we did in the cohort-level analyses, the genetic effects in these incompletely dosage compensated loci are expected to be larger in absolute value in women than in men. The fixed-effects meta-analysis does not, however, account for potential sex heterogeneity. Therefore, we complemented the fixed-effects analysis by performing a meta-analysis that treats the male and female-specific genetic effects separately, a so-called sex-differentiated meta-analysis (Materials and Methods), in order to indicate loci showing incomplete dosage compensation and to potentially also facilitate the discovery of new associations. Allowing for different effect sizes between males and females in the meta-analysis pinpointed no further loci, yet the ITM2A lead height SNP was more strongly associated in this sex-differentiated analysis (joint P-value = 3. 26×10−12; Table 3). Pointing to lack of dosage compensation, in this locus the allelic effects were estimated to be more than twice the size in women compared to men when coding hemizygous men equal to homozygous women (standardized beta in females: 0. 093, se: 0. 014; beta in males: 0. 035, se: 0. 009; P-value for the difference in the male and female effects = 2. 85×10−4; Table 3). For the other two new loci there was no indication of heterogeneity in the male and female effects (Table 3). Accordingly, the proportion of variance explained was approximately twice the size in men compared to women in the ATRX (0. 22% vs. 0. 14% for height) and Xq23 (0. 51% vs. 0. 19% for fasting insulin) loci, as expected under the model of random XCI, but not for the ITM2A SNP (0. 12% vs. 0. 39% for height) (explained variances calculated under the assumption of full dosage compensation; Table 3; Materials and Methods). The observed deviation from the full dosage compensation model in the ITM2A locus was not driven by differences in allele frequencies or sample sizes, as these are similar between men and women (Tables 1 and 3). To verify the observation of lack of dosage compensation in the ITM2A locus for height, we formally investigated how the models of full dosage compensation and no dosage compensation explain the data in the three associated loci pinpointed in our chrX-wide association analysis. To this end we devised a novel application of a Bayesian model-comparison framework [16] (Materials and Methods), to quantify the relative support for each of the dosage compensation models. While the Xq23 and ATRX associations were highly consistent with full dosage compensation there was clear evidence towards complete escape from the inactivation in the ITM2A locus: Assuming that both models are equally probable a priori, the posterior probabilities for the no dosage compensation model are 0. 07 for rs13916345 in Xq23 (fasting insulin), and 0. 18 for rs182838724 near ATRX (height), but 0. 99 for rs1751138 near ITM2A (height) (Figure 2). As lack of dosage compensation in the ITM2A locus in women should be reflected in the level of ITM2A expression, given the two actively transcribed X chromosomes, we evaluated the sex difference in the level of the ITM2A expression probe that had earlier showed a significant cis-effect with the lead SNP of the locus. The average level of ITM2A expression in whole blood was observed to be higher in women (P-value = 0. 00251; Figure S2; Materials and Methods), thus providing further support for incomplete XCI in this locus.
Motivated by the underrepresentation of reported GWAS discoveries in chrX, we investigated the association of chrX to twelve anthropometric and cardiometabolic traits in more than 24,500 individuals using a high-resolution map of non-pseudoautosomal chrX SNPs. Our data demonstrate that SNPs in chrX are associated with many of the studied phenotypes, including the three novel loci pinpointed in our chrX-wide association analysis. Additionally, our discovery of lack of dosage compensation for height near ITM2A not only highlights the value of accounting for potential sex heterogeneity when assessing chrX associations, but also manifests that some of the X-linked loci may contribute to sexual dimorphisms, in this case to the height difference between men and women. The contribution of common X-chromosomal SNPs to a few complex phenotypes have been explored previously: Yang et al. showed that between 0. 57% and 0. 82% of the variance in height, BMI and von Willebrand factor is explained by the SNPs in chrX [2]. In our study, we extended the estimates to further ten anthropometric and cardiometabolic phenotypes, and also use a more comprehensive set of SNPs, and similarly demonstrate that a small but non-negligible proportion, up to 1. 4%, of the total variance in many of the twelve phenotypes studied appears attributable to chrX. While the variance estimates for autosomes have been observed to be proportional to the chromosome length [2], our estimates for chrX were on average 2. 6% of the total genetic variance estimate, and therefore below what would be expected based on the proportion of the genomic DNA contained within chrX, i. e. , approximately 5%. However, given the smaller population size and lower mutation rate, chrX is genetically less diverse than the autosomes, and indeed our estimates appear to be more in line with around 3. 4% of the SNPs in the 1000 Genomes reference being X-chromosomal [9] (Figure S3). This implies that around 3% of all GWAS discoveries could be hidden in chrX and hence with the inclusion of chrX in GWAS further X-chromosomal loci for complex traits will be discovered. As thus far chrX has only infrequently been included in GWAS of the twelve traits studied, it is unsurprising that the three loci now discovered, one for fasting insulin and two for height, represent the first ones reported in chrX for these phenotypes in European populations. Two earlier GWAS conducted in individuals of African ancestry discovered a height locus in Xp22. 3 [17], [18], yet the associated lead SNP is monomorphic in Europeans and the region shows no signal in our analyses. One of the early height GWAS including mainly European samples showed a suggestive association (P-value = 3×10−6) [19] mapping 8. 6 kb from the ITM2A association discovered in our analyses (r2 = 0. 728 between the lead SNPs), however, in that study the finding failed to replicate and thus never reached the formal threshold for significance. While further studies are required to elaborate the causative variants underlying the association with fasting insulin in Xq23, in both of the height loci there are plausible candidate genes that could be responsible for the observations. The first height association spans FGF16, ATRX and MAGT1 in a gene rich region. Both ATRX and MAGT1 have been implicated in mental retardation syndromes, and interestingly the syndrome associated with mutations in ATRX is often accompanied with skeletal abnormalities and short stature in the affected individuals [15]. Another candidate is FGF16, which is a member of fibroblast growth factor family shown to play various roles in developmental processes including morphogenesis. Mouse studies proposed a crucial role for FGF16 in cardiac morphogenesis [20] and, recently, nonsense mutations in FGF16 were demonstrated to associate with congenital limb malformations [21], suggesting the involvement of the gene in human skeletal development. In the second height locus, ITM2A is a functional candidate: The association signal maps just 35 kb upstream of the transcription start site of ITM2A, a gene known to be involved in cartilage development [13], [14]. Additionally, our eQTL analysis provided further evidence showing that the height-associated variants in this locus also influence the expression of ITM2A. Interestingly, high expression of ITM2A in adipose tissue stem cells has been proposed to inhibit the initiation of chondrogenesis in these cells [22]. As the allele associated with shorter stature associated with increased expression of ITM2A, this suggests the allelic effect to height could be mediated through the capacity to generate cartilage and bone. Given the unique nature of chrX compared to autosomes, i. e. , males being hemizygous for the chromosome and the partly incomplete XCI, which could give rise to sexual dimorphisms, we analyzed our association data also by accounting for potential heterogeneity in genetic effects between sexes. This led to the discovery of considerable sex-difference in the genetic effects for height near ITM2A, while for the other two associated loci there was no such evidence. A large-scale scan on 270,000 individuals for sexual dimorphisms in autosomal genetic effects for various anthropometric phenotypes identified significant sex differences only in associations for waist phenotypes and none was observed for height [23]. Additionally, the few previous height GWAS that had significant or suggestive discoveries in chrX [17]–[19] did not, to our knowledge, evaluate for potential sex heterogeneity in the chrX associations. Therefore, the sex difference in genetic effects for height in the ITM2A locus here is likely the first demonstration of its kind and hence warrants the investigation for potential sex heterogeneity in associations for other X-linked loci. The difference in the genetic effects for males and females near ITM2A appeared fully consistent with no dosage compensation in this locus. While we cannot completely exclude the possibility that such a difference can also arise through some other type of sex-by-SNP interaction effect, we gained further confirmation for the lack of dosage compensation for height in ITM2A by quantifying the evidence for the two dosage compensation models in each of the associated loci using a new model comparison framework. Additionally, the gene expression data that showed higher expression of ITM2A in women, as expected if two X chromosomes are transcriptionally active in this locus. Similarly, expression of ITM2A was previously observed to be female-biased in monocytes and in the same study the cis-eQTL for ITM2A showed sex heterogeneity [24]. Furthermore, strongly speaking for the role of impartial silencing of this locus, in a comprehensive survey into XCI, ITM2A was found to be among the 10% of chrX genes, which variably escape from inactivation, i. e. , are expressed from the inactivated copy of X (Xi) in a majority of women [8]. For comparison, the expression of ATRX and MAGT1 in the other height locus (FGF16 was not included in the survey) from Xi is fully silenced [8]. Given the converging evidence from our association study, statistical model comparison of the association data, gene expression data and literature, it seems likely that our observations of sex heterogeneity in the genetic effects for height near ITM2A are due to incomplete dosage compensation in this locus. Therefore, our study likely provides, to the best of our knowledge, the first link between an XCI-escaping gene and phenotypic variation in non-syndromic individuals. This discovery has several plausible implications. As increased expression of ITM2A links with shorter stature, the greater dosage of ITM2A in women compared to men may explain some of the substantial sex difference observed in height. In addition to the sex differences in the overall expression levels, incomplete dosage compensation also causes the genetic variation in the population to have different effects on the trait distribution of males and females. Assuming that rs1751138, the ITM2A lead SNP, is the causal variant for the observed height association, we estimate that the observed 36% frequency of the height decreasing allele accounts for 1. 5% of the current difference in mean height between men and women in the Finnish population, when compared to a population that was monomorphic for the major allele at this SNP (Materials and Methods). Besides contributing to sexual dimorphisms, XCI-escaping genes are candidates for causing the abnormalities in chrX aneuploidies. Hence, our findings also highlight ITM2A as a potential contributor to the height phenotype often observed in individuals with an unusual number of X chromosomes [25]. To conclude, our findings illustrate the value of including the chrX in large-scale genetic studies and provide a motivation to assess the chrX associations in larger sample sizes, particularly for traits where we estimated chrX to explain part of the trait variation. We anticipate that such studies will identify further loci that contribute to the heritability of complex traits, as well as increase our understanding of their genetic architecture and underlying biology. As evidenced by our observations in the ITM2A locus, studying chrX association opens avenues for the discovery of links between phenotypes and loci that escape from XCI. Such associations bear potential to bring insights into the biological bases of sexually dimorphic traits such as complex diseases with different incidences between males and females.
The study included individuals from seven discovery cohorts: The Northern Finland Birth Cohort 1966 (NFBC1966), The Cardiovascular Risk in Young Finns Study (YFS), The COROGENE Study (COROGENE), Helsinki Birth Cohort Study (HBCS), the Health 2000 GenMets Study (GenMets), the Diabetes Genetics Initiative (DGI) and a prospective cardiovascular disease (CVD) case-control sample from the FINRISK collections (PredictCVD). In the replication stage a further subset of individuals from the FINRISK collections (FR) was included. A summary of the cohort characteristics is given in Table 1. All participants gave an informed consent and the data was de-identified for all analyses. NFBC1966 is a birth cohort study of children born in 1966 in the two northernmost provinces of Finland originally designed to focus on factors affecting pre-term birth, low birth weight, and subsequent morbidity and mortality [26]. The blood sample for the DNA extraction and all phenotype data (except the childhood height measurements) used in the present study were collected at a follow-up visit when the participants were 31 years of age. The COROGENE cohort includes acute coronary syndrome patients who underwent coronary angiogram between June 2006 and March 2008 in the Helsinki University Central Hospital [27] and matched controls from the Helsinki-Vantaa region participants of FINRISK 1997,2002, and 2007 surveys performed in Finland every five years since 1972 [28]. In the current study, the COROGENE cases were only included in the analyses of height and body-mass-index. The DGI sample consists of individuals from Sweden and Finland and was originally designed to identify loci associated with type 2 diabetes (T2D) [29]. The sample consists of patients with T2D; gender, BMI, age and geographically matched controls and discordant sibships. The current analysis was performed on 3142 individuals, including all individuals from the original analysis and individuals previously excluded after having been identified as belonging to pairs of samples identified as cryptic first degree relatives [29]. GenMets is a subset from the Health 2000 survey collected in 2000–2001 to obtain information on the most important public health problems in Finland [30]. The cohort includes metabolic syndrome cases and their matched controls aged 30 years and above. YFS is a longitudinal follow-up study of children and adolescents from all around Finland initiated in 1980 to study the cardiovascular risk from childhood to adulthood [31], and the data for the present study is from the 27-year follow-up when the participants were 30–45 years of age. The PredictCVD study comprises of incident cardiovascular disease cases and matched controls selected from the 1992,1997,2002 and 2007 FINRISK surveys [28]. For the analyses of the current study, the 96 individuals included and genotyped in both COROGENE and PredictCVD samples were excluded from the COROGENE data set. The FR sample used for replication includes a random, yet non-overlapping with COROGENE and PredictCVD, subset of individuals from FINRISK surveys from 1997 and 2002. The following genotyping arrays were used for genotyping the study cohorts: Illumina 370K array for NFBC1966, Illumina 610K array for COROGENE and GenMets, custom generated Illumina 670K array for YFS and HBCS, Illumina 770K array for PredictCVD, Illumina HumanCoreExome-12v1-0 for FR and Affymetrix GeneChip Human Mapping 500K Array set for DGI. The quality control procedures included removing closely related individuals (PI_HAT >0. 1) by analyzing pairwise IBD relationships for all individuals in five Finnish discovery cohorts together and have been described previously in detail [29], [32]. The imputation of non-pseudoautosomal chrX variants into the study cohorts was performed on the cleaned data in each cohort using IMPUTE version 2. 2. 2 [33], [34]. The reference panel used in the imputation was the integrated variant set release (v3) released in March 2012 (http: //mathgen. stats. ox. ac. uk/impute/data_download_1000G_phase1_integrated. html). The data were split into genomic regions of ∼5 Mb (with 250 kb (DGI) or 1 Mb (other cohorts) buffer region), using effective population size of 20000 (DGI) or 11418 and k value of 80. Within each cohort, all twelve phenotypes were adjusted for males and females separately using age (not in NFBC1966) and ten first principal components as covariates. Additionally, WHR and blood pressure measurements were adjusted for BMI. The residuals from the linear regression were inverse normal transformed to have mean 0 and standard deviation 1, and the normalized residuals were then used as phenotypes in the variance estimate and association analyses. In cohorts where the information was available, individuals on lipid-lowering medication were excluded prior the covariate adjustment for the blood lipids (TC, TG, LDL-C and HDL-C), blood-pressure medication was an exclusion criterion for systolic and diastolic blood pressure and diabetes medication for glucose and insulin. Non-fasting individuals were excluded from the analysis of glucose and insulin, and pregnant women were only included in the analysis of CRP and height. All phenotype adjustments and data normalization were done in STATA/SE 12. 1 or R (version 2. 11. 1). The sample size for each phenotype in the discovery cohorts, i. e. , the number of individuals with both genotype and phenotype information, are given in Table S5. We estimated how much phenotypic variance a panel of 217,112 high quality SNPs from chrX (IMPUTE2 info >0. 8, MAF >0. 01) explain using the linear mixed model approach implemented in GCTA (v. 1. 13) [10]. This analysis included six of the seven discovery cohorts (NFBC1966, COROGENE, GenMets, YFS, HBCS and PredictCVD) for which we had access to the individual genotype data. Following the work of Yang et al. [2], we applied three models for dosage compensation: full dosage compensation (FDC), no dosage compensation (NDC) and equal variance in both sexes (EV). However, none of the models was consistently favored over the other two across the traits (Table S1). In the main text we report the variance estimates from the EV model because those best capture the average genetic contribution of chrX to the population: in FDC and NDC models the variances between males and females are different and for both models the mean of the sex-specific variances is close to the single value given by the EV model in our data. Furthermore, since the traits have been normalized to have a variance of 1 within each sex, the models that assume different genetic variance between the sexes (namely FDC and NDC) should also allow different residual variances between the sexes but this has not been implemented in GCTA (v. 1. 13). We report the results from all three models using imputed SNPs in Table S1. For a comparison, we also estimated the genetic variances using 9,517 directly genotyped SNPs in chrX but did not find notable differences from the results with the imputed data (Table S6). For another comparison, we estimated the variance explained by the 319,445 directly genotyped SNPs with MAF>0. 01 in the autosomes (Table 2). If the variants contributing to the traits were uniformly distributed across the genome, then we would expect that chrX genetic variance is about 3% of the autosomal genetic variance, as about 3% of the genetic variation in our data is in the X chromosome. Figure S3 plots estimated autosomal genetic variance against chrX one. All mixed model analyses excluded individuals in such a way that none of the remaining pairs of individuals had an estimated relatedness coefficient r >0. 05 and the same trait values were used as with the association analyses. For comparison, the analyses were also carried out by using r = 0. 025 as the relatedness cut-off (Table S7). The associations between variants and phenotypes were tested for all available genotyped or imputed SNPs in chrX separately for men and women in each cohort encoding genotypes {0,2} in men and {0,1, 2} in women, i. e. , assuming that one of the two X chromosomes in women is fully inactivated. In DGI the analysis was performed using a linear mixed model that can account for sample structure, as implemented in EMMAX [35]. In other cohorts the analyses were performed in SNPTEST [11] (version 2. 4. 0) assuming an additive genetic model and using expected genotype counts. In each cohort the results were filtered prior the meta-analysis to include only good-quality variants (SNPTEST info in women >0. 4) and variants with more than three copies of the minor allele (minor allele count in women >3) resulting in between 323,564 and 383,337 variants per cohort in the discovery set. The cohorts and genders were combined in a fixed-effects (inverse variance weighted) meta-analysis and in a sex-differentiated meta-analysis, which combines the results allowing for the allelic effects to differ between men and women and also conducts a meta-analysis separately for the results from men and from women, both meta-analysis options implemented in GWAMA [12]. Genomic control was applied for the association results of each study to account for P-value inflation arising from residual population structure in the data or other confounding factors. The meta-analysis summary was also corrected with the genomic control lambda for those phenotypes where there was indication of inflation (lambda >1. 0): The genomic control lambdas for the twelve phenotypes were between 0. 94–1. 13 in the discovery analysis (Table S8). For each phenotype the meta-analysis results were further filtered to include only the SNPs for which the association result was available from at least two of the input files, hence association results were available for up to 405,411 SNPs. Adopting the genome-wide significance P-value threshold, P-value<5×10−8, three loci showed significant associations in the discovery analysis. In these associated loci the lead SNPs were imputed with high quality (imputation info >0. 80) in all cohorts. To investigate whether the observed associations with height near ITM2A and ATRX are present already in pre-puberty, the association of the chrX SNPs with childhood height was studied in a subset of individuals from NFBC1966 and YFS cohorts (N (males) = 1204 and N (females) = 1189 in NFBC1966; N (males) = 417 and N (females) = 478 in YFS) who had height measurements available both from pre-puberty (ages 8–10 years for NFBC1966 and 9 years for YFS) and adulthood. The height measurements were adjusted for age (in NFBC1966) and ten first principal components, the residuals were inverse normal transformed and subsequently the association of the SNP dosage with the transformed residuals was studied in SNPTEST using an additive model of association. The SNP-phenotype associations for the two height measurements were studied in both cohorts, separately for the genders, and the results were combined in a fixed-effects and a sex-differentiated meta-analysis using GWAMA applying genomic control correction. The association results for the height lead SNPs with childhood and adulthood height in these individuals are given in Table S4. A subset of the COROGENE cohort (N = 513) had both genome-wide SNP data and gene expression data from whole blood, assayed using Illumina HumanHT-12v3 Expression BeadChips, available, as described previously [36]. The associations between the three lead SNPs and all gene expression probes within 1 Mb of the SNPs was studied in SNPTEST using an additive model of association. For the one significantly associated expression probe (ITM2A/ILMN_2076600) we further evaluated whether the level of expression was different between the sexes by comparing the unadjusted expression in males and females using Student' s t-test in R. As males carry only one copy of the X chromosome, the genotype variances of the SNPs in the non-pseudoautosomal region of chrX differ between women and men. Assuming the model of complete inactivation of one the X chromosomes in women, and hence coding the X-linked alleles {0,2} in men and {0,1, 2} in women, the genotype variance in men is twice that in women: 2P (1-P) for females and 4P (1-P) males, where P denotes the allele frequency of the SNP. Thus, the estimation of the variance explained from the meta-analysis summary statistics should be evaluated for the genders separately using the formulas 2P (1-P) bF2 for women and 4P (1-P) bM2 for men, where bF and bM denote the standardized effect sizes in women and men, respectively. We applied a Bayesian framework [37] to compare full dosage compensation (FDC) and no dosage compensation (NDC) models at the top SNPs of the three associated regions. We used the estimated effect sizes bF (allelic effect in females) and bM (effects in males when two genotypes are coded 0 and 2) together with their standard errors to approximate the likelihood function as in [16]. The two models differ in their prior specification: FDC: bF∼N (0, s2), bM∼N (0, s2) and cor (bF, bM) = 1 NDC: bF∼N (0, s2), bM∼N (0,0. 25 s2) and cor (bF, bM) = 1 Where s2, the variance of the prior effect size in females, depends on the allele frequency of the SNP and is chosen so that with 95% probability the studied SNP explains less than 1% of the variance of the trait. Intuitively, according to the FDC model bF = bM whereas according to the NDC model bF = 2bM. Bayes factors between the two models can be computed using the approximate likelihood approach [16] and the posterior probabilities of the NDC model are shown in Figure 2 under the assumption that each model is equally likely a priori. Suppose that each copy of the minor allele (‘a’ with frequency fa) at a particular SNP on chrX causally affects the trait value by ‘b’ both in males and in females, i. e. , there is no dosage compensation. To simplify notation, we assume that the mean trait value of males with genotype ‘A’ is the same as the mean trait value in females with genotype ‘AA’, and denote this mean by ‘m’. The overall mean trait values in males, ‘mM’, and in females, ‘mF’, are mM = (1−fa) •m + fa• (m+b) = m + fa•b, and mF = fAA•m + fAa• (m+b) + faa• (m+2b) = m + (fAa+2faa) •b = m + 2fa•b The difference in the sex-specific means is mM − mF = m + fa•b − (m + 2fa•b) = − fa•b So the effect of allele ‘a’ is either to increase (if b<0) or decrease (if b>0) the male-female difference in the mean trait value by |fa•b|, compared to the situation where only allele ‘A’ was present in the population. Application to rs1751138, the lead SNP of the association near ITM2A: b = −0. 555 cm (se = 0. 0734 cm). This is a fixed-effects estimate of the allelic effects in quantile normalized height in females −0. 092559 (0. 013934) and males −0. 071802 (0. 019274), multiplied by an estimate of the standard deviation of height in Finland, 6. 5 cm. Thus by introducing the minor allele (A) with frequency fa = 0. 36 in the population, the male-female difference in mean height increases by 0. 36*0. 555 cm = 0. 20 cm. As the mean difference in height between the sexes in Finland is about 13. 7 cm, the variation at this SNP accounts for 0. 20 cm/13. 7 cm = 0. 0146≈1. 5% of that difference. | The X chromosome (chrX) analyses have often been neglected in large-scale genome-wide association studies. Given that chrX contains a considerable proportion of DNA, we wanted to examine how the variation in the chromosome contributes to commonly studied phenotypes. To this end, we studied the associations of over 400,000 chrX variants with twelve complex phenotypes, such as height, in almost 25,000 Northern European individuals. Demonstrating the value of assessing chrX associations, we found that as a whole the variation in the chromosome influences the levels of many of these phenotypes and further identified three new genomic regions where the variants associate with height or fasting insulin levels. In one of these three associated regions, the region near ITM2A, we observed that there is a sex difference in the genetic effects on height in a manner consistent with a lack of dosage compensation in this locus. Further supporting this observation, ITM2A has been shown to be among those chrX genes where the X chromosome inactivation is incomplete. Identifying phenotype associations in regions like this where chrX allele dosages are not balanced between men and women can be particularly valuable in helping us to understand why some characteristics differ between sexes. | Abstract
Introduction
Results
Discussion
Materials and Methods | genome-wide association studies
x chromosome inactivation
genetics
epigenetics
biology | 2014 | Chromosome X-Wide Association Study Identifies Loci for Fasting Insulin and Height and Evidence for Incomplete Dosage Compensation | 10,320 | 299 |
Apicomplexan parasites are global killers, being the causative agents of diseases like toxoplasmosis and malaria. These parasites are known to be hypersensitive to redox imbalance, yet little is understood about the cellular roles of their various redox regulators. The apicoplast, an essential plastid organelle, is a verified apicomplexan drug target. Nuclear-encoded apicoplast proteins traffic through the ER and multiple apicoplast sub-compartments to their place of function. We propose that thioredoxins contribute to the control of protein trafficking and of protein function within these apicoplast compartments. We studied the role of two Toxoplasma gondii apicoplast thioredoxins (TgATrx), both essential for parasite survival. By describing the cellular phenotypes of the conditional depletion of either of these redox regulated enzymes we show that each of them contributes to a different apicoplast biogenesis pathway. We provide evidence for TgATrx1’s involvement in ER to apicoplast trafficking and TgATrx2 in the control of apicoplast gene expression components. Substrate pull-down further recognizes gene expression factors that interact with TgATrx2. We use genetic complementation to demonstrate that the function of both TgATrxs is dependent on their disulphide exchange activity. Finally, TgATrx2 is divergent from human thioredoxins. We demonstrate its activity in vitro thus providing scope for drug screening. Our study represents the first functional characterization of thioredoxins in Toxoplasma, highlights the importance of redox regulation of apicoplast functions and provides new tools to study redox biology in these parasites.
Apicomplexan parasites are global killers of animals and humans. Most apicomplexans possess a plastid, the apicoplast. The apicoplast is essential for parasite survival [1] throughout their complex life cycles and has no equivalent in humans. Accordingly, apicoplast functions and pathways of biogenesis are sought after as promising drug targets for diseases like toxoplasmosis and malaria [2]. The apicoplast was acquired via endosymbiosis, whereby a eukaryotic auxotroph took up an autotrophic alga (Fig 1A). Subsequent reduction of the algal organelles and gene transfer from the algal nuclear genome to the auxotroph genome resulted in integration of the alga as an organelle. A similar pathway gave rise also to the complex plastids found in a divergent group of organisms of ecological or medical importance. These include the cryptophytes, heterokonts, haptophytes, dinoflagellates, and chromerids [3]. The apicoplast and related plastids have multiple compartments: the two inner compartments originate from the algal primary plastid. The next compartment out is the periplastid compartment (PPC). The PPC is remnant of the algal cytosol, and in the cryptophyte plastid the PPC still hosts a relic of the algal nucleus, named the nucleomorph [4]. Finally, the outermost compartment comes from the host and is either of phagosomal [5] or ER [6] origin. Most apicoplast proteins are nuclear encoded, co-translationally translocated into the ER and trafficked from there to the apicoplast outermost membrane and through the subsequent compartments to their destination within the organelle. The conservation of this elaborate architecture necessitates a mechanism that ensures that proteins fold into their functional forms in their appropriate compartment and that they are kept in a conformation compatible with their translocation through the membranes bounding these compartments. The formation of disulphide bonds from dithiols affects the conformation of proteins and thus plays an important regulatory role in their sorting to their target cellular compartments and in their correct function when in these compartments. Proteins with thioredoxin domains (Trxs) mediate disulphide-dithiol dynamics in target proteins in response to compartmental redox states. The thioredoxin (Trx) fold contains a double cysteine active site (cysteine-X-X-cysteine/CXXC) through which the disulphide exchange occurs. In this reaction, the N-terminal cysteine serves as a nucleophile, creating an intermediate mixed disulphide species with the substrate. Next, a nucleophilic attack by the C-terminal cysteine of the Trx CXXC motif on its N-terminal cysteine results in an oxidized CXXC motif and in the release of the reduced substrate (Fig 1B illustrates this process). A well-studied example of the role of Trxs in controlling protein folding and sorting is the family of protein disulphide isomerases (PDI) that mediate the folding and sorting of secretory proteins in the ER [7]. Redox mediated protein folding also takes place in the mitochondrial intermembrane space via the oxidoreductase Mia40 and the oxidase Erv1 [8]. We have previously identified two apicoplast Trxs in Toxoplasma gondii (TgATrxs). TgATrx1 is a resident of the apicoplast periphery and is also found in vesicles in the cytosol, around the apicoplast and at low levels in the ER [9]. TgATrx2, also found at the apicoplast periphery, was suggested to be a PPC resident, which is compatible with the presence of its orthologues in the cryptophyte nucleomorph genomes [10]. The roles of both TgATrxs are unknown. Here we examine the hypothesis that TgATrxs may play a role in the control of apicoplast biogenesis by mediating disulphide exchange with proteins destined to different compartments of the apicoplast. We show that both TgATrxs are essential for parasite growth. In line with our hypothesis both TgATrxs depend on their CXXC active site for function, but the depletion of each TgATrx results in a defect in a different apicoplast biogenesis pathway. Finally, TgATrx2 and its malaria orthologues have features that are divergent from canonical Trxs. Nevertheless, recombinant TgATrx2 exchanges disulphide in vitro. We have utilized this ability to generate an in vitro activity assay that lays the foundations for the development of a platform for drug screens in the future.
To consider the importance of ATrx1 and ATrx2 to plastid function and correlate their presence with complex plastid evolution, we searched for homologues of both sequences in apicomplexans and other eukaryotes with complex plastids. TgATrx1 homologues were found throughout apicomplexans (coccidia, piroplasms and hemazoans) and were only lacking from Cryptosporidium spp. , a genus that has lost its plastid. Homologues with greatest similarity to apicomplexan ATrx1s were found in chromerids, dinoflagellates and heterokonts. Molecular phylogenies of ATrx1 sequences resolved the apicomplexans with strong support as a monophyletic group along with chromerids and a subset of heterokont taxa (Fig 2, green shading). This ATrx1 clade excluded dinoflagellate sequences and other heterokonts, including diatoms (Fig 2). Notably, ATrx1 proteins from the clade containing apicomplexans, chromerids and heterokonts all harbor predicted plastid-targeting bi-partite pre-sequences, whereas the outgroup proteins including those of dinoflagellates and more complete representation of heterokonts are all predicted as cytosolic proteins. These data suggest a plastid ATrx1 occurs throughout apicomplexan/chromerid plastid radiation. Further, the phylogeny suggests that a gene duplication within heterokonts allowed for the evolution of the plastid form, and that this was the source of the apicomplexan plastid protein. Plastid ATrx1 has either been lost from dinoflagellates or was never gained, and similarly a cytosolic paralogue in apicomplexans is absent, presumably lost. Curiously, while ATrx1 is common to all apicomplexans with plastids, the protein in piroplasms and hematozoans has lost the otherwise conserved CXXC motif (Fig 2). ATrx2 is also broadly represented in apicomplexans and other eukaryotes with red-derived plastids, including being encoded in cryptophyte nucleomorphs. Unlike TgATrx1 homologues, all TgATrx2 homologues are predicted to occur in plastids. Further, in contrast to ATrx1s, the ATrx2 CXXC motif (CDHC or CEYC) is conserved throughout apicomplexans and chromerids (S1 Fig). The conserved portion of ATrx2 is short and thus ATrx2 phylogenetics were unresolved. Our previous studies showed that both TgATrx1 and TgATrx2 are residents of the apicoplast periphery. We showed that TgATrx1 may localize to several peripheral compartments, to vesicles in the cytosol and at low levels to the ER [9], while TgATrx2 is likely confined to the PPC [10]. To better resolve the location of these two proteins we have employed specific sub-compartmental markers [10] and examined their co-localization using super resolution microscopy. Consistent with our previous observations, signals for both proteins surround the stromal compartment visualized by CPN60 [11]. However, TgATrx1 and TgATrx2 only partially co-localize with each other and their staining patterns are different (Fig 3, S2 Fig). TgATrx2 tightly co-localizes with the PPC marker PPP1, but shows only partial overlap with the outer compartment marker 201270 (the product of TGME49_201270 previously designated 101270 and 001270 [10]) which lends support to its proposed localization in the PPC (Fig 3, S2 Fig). TgATrx1 signal shows incomplete overlap with PPP1 but fully overlaps with 201270 and presents additional staining (Fig 3, S2 Fig). The additional signal may correspond to the vesicles previously observed by electron microscopy [9]. Taken together the differences identified in the phylogenetic distribution and localization of the two TgATrxs suggest distinct functions. The localization further provides additional support to the previous suggestion [10] that TgATrx2 serves a role that is relevant to the PPC. To assess the function of both TgATrxs in T. gondii we engineered conditional mutant lines for each of them, in which a tetracycline-regulatable promoter drives the expression of theTgATrx1 gene or the TgATrx2 gene as described previously [10] and as illustrated in Fig 4A. For the TgATrx1 gene this manipulation utilized the TATiΔKu80 line [10] as the parental line. For the TgATrx2 gene the parental line was a TATiΔKu80 background where the TgATrx2 coding sequence was endogenously C-terminally tagged with three HA epitope tags as previously described [10]. The resulting lines are named TATiΔKu80PIATrx1 and TATiΔKu80PIATrx2-3HA. Both proteins were downregulated upon addition of anhydrotetracycline (ATc) (Fig 4B and 4C). The previously reported multiple forms of TgATrx1 [9] were observed by Western blot prior to addition of ATc, and all species fell below the detection levels by 48 hours of ATc treatment (Fig 4Bi). Two forms are detected for TgATrx2, likely corresponding to the protein before and after cleavage of its targeting presequences, which occurs upon apicoplast import. Both forms fell below the detection level at 48 hours of ATc treatment (Fig 4Bii). To test for growth phenotypes upon TgATrx1 or TgATrx2 depletion, plaque assays were performed. In both cases depletion resulted in loss of plaque formation indicating a severe growth defect in the absence of either of these proteins (Fig 4D). To assess the importance of the putative CXXC active site of each TgATrx we tested the ability of either wild type TgATrx, or a mutant form with the second active site cysteine substituted with an alanine (CXXA), to rescue the knockdown phenotypes (see Fig 1B for mechanism). For each regulated line, we constitutively expressed TgATrx1 or TgATrx2 minigenes, bearing either the wild type CXXC (TgATrx1CXXC or TgATrx2CXXC) or a mutated form CXXA (TgATrx1CXXA or TgATrx2CXXA). For both conditional mutant lines, a constitutively expressed copy with the wild type CXXC rescued the growth phenotype upon ATc treatment, whereas the active site mutant did not (Fig 4E). Immunofluorescence analysis confirmed that the complementing proteins reached the apicoplast (Fig 4F). ATrx2 sequences have two features that are distinct from classical thioredoxins. First, while thioredoxins are typically small proteins (e. g. human Trx1 and Trx2 are 12 and 18 kDa respectively [12]), ATrx2 orthologues are larger, e. g. TgATrx2 is ~50kDa with N- and C-terminal extensions to its conserved Trx fold. Second, in classical thioredoxins (e. g. human Trx1 (hTrx1) ) the two amino acids of the CXXC motif are typically glycine and proline, and this affects the Trx redox potential [12]. In TgATrx2 and all its orthologues an acidic and an aromatic residue (CDHC or CEYC) are found (S1 Fig) [10]. To validate that TgATrx2 can engage in disulphide exchange we tested its activity in vitro. Purified recombinant 6xHis-tagged TgATrx2 (S3 Fig) was incubated with reduced recombinant hTrx1. The resulting samples were alkylated with 4-acetamido-4' -maleimidylstilbene-2,2' -disulfonic acid (AMS), thus adding ~510 Dalton per thiol, resulting in a shift in the migration of proteins with reduced thiols compared to oxidized thiols upon separation by SDS-PAGE. TgATrx2 showed such a migration shift upon incubation with hTrx1 (S3 Fig), suggesting that the two exchanged disulphides. We further demonstrated that recombinant TgATrx2 can reduce insulin in an in vitro insulin turbidity assay [13] (S3 Fig). These data demonstrate that both TgATrx1 and TgATrx2 provide essential functions for T. gondii growth in culture. The CXXC motifs are essential elements implying that disulphide exchange is key to these functions. In addition to the typical TgATrx1 staining at the apicoplast periphery and vesicles around the apicoplast, the mutant TgATrx1CXXA was found in other cellular foci (Fig 4F). This was not the case for TgATrx2CXXA mutant which maintained the same localization as its wild type parent (Fig 4F). Co-staining with an ER marker imaged by super-resolution microscopy showed that this additional signal is adjacent to and in some cases co-localizing with the ER (S4 Fig). This suggests that the active site mutation enhances ER accumulation beyond the light ER localization observed previously only by electron microscopy [9]. The altered localisation of TgATrx1 upon mutation of its CXXC active site raises the possibility that its overall depletion could cause a general protein import defect. Therefore, we investigated the import of known apicoplast stromal (LytB) and peripheral (PPP1) markers under TgATrx1 depletion using the previously described import assay [10,14]. The targeting presequences of these proteins are cleaved upon arrival at their destination. When import is compromised, the un-processed precursor accumulates at the expense of the mature form, and this precedes the point of complete organelle loss [10,14]. Newly synthesized PPP1 showed minor precursor form accumulation at 48 hours of ATc treatment, and by 72 hours no mature protein was detectable (S5 Fig). Likewise, after 72 hours of TgATrx1 depletion, LytB mature form was undetectable (S5 Fig). We examined apicoplast numbers during TgATrx1 depletion. Scoring plastids in 100 parasites showed that apicoplast loss was evident at 24 hours and continued gradually (S5 Fig). Note that the complete loss of mature proteins occurs at a time point where 68% of the parasites still have an apicoplast (Fig 5). In contrast, in the case of TgATrx2, the mature forms of both LytB and PPP1 were still detected at 72 hours of depletion, despite similar levels of apicoplast loss as seen with TgATrx1 depletion at this time point (S5 Fig). This suggests that unlike TgATrx1, when TgATrx2 is depleted newly synthesized apicoplast proteins can continue to target the organelles that are still intact and reach maturation. These experiments indicated a protein import defect in TgATrx1-depleted cells that was not seen with depletion of TgATrx2. We assessed how other pathways of apicoplast biogenesis and maintenance were affected under TgATrx2 depletion. We examined apicoplast genome maintenance by measuring the relative copy number of the apicoplast TogoCr29 gene (which encodes the large subunit rRNA) normalized to the copy number of the nuclear act1 (TGGT1_209030) gene via qPCR [14–16]. At 24 hours of ATc treatment, no significant change was observed of TogoCr29 compared to the untreated control (S5 Fig). At 48 and 72 hours, TogoCr29 copies were 60% and 47% of the untreated control. This decrease is more rapid than the observed decrease in apicoplast numbers (Fig 5), suggesting a reduction in TogoCr29 gene copies in the treated parasites. Next, we assessed apicoplast transcription. We performed RT-qPCR using total RNA extracted at different time points of ATc treatment. In each reaction, we compared the expression levels of apicoplast-encoded TogoCr29 mRNA to the nucleus-encoded Act1 mRNA. Some decrease in expression (to 89%) was observed at 24 hours, followed by a steep drop to 18% and 13% at 48 and 72 hours (S5 Fig). This reduction largely precedes plastid loss, which was at 69% at 72 hours (Fig 5). These data suggest that apicoplast transcription is decreased upon TgATrx2 depletion. No tools are currently available to directly measure translation in the apicoplast. For comparison, we measured apicoplast genome maintenance and gene expression also under TgATrx1 depletion. At 72 hours of ATc treatment, the apicoplast to nucleus genome copy number ratio was equivalent to that measured in untreated parasites (S5 Fig), indicating no effect on apicoplast genome maintenance. Some decrease in apicoplast gene expression was observed, however, it was significantly milder than that observed following TgATrx2 depletion (S5 Fig). Fig 5 summarises the measurements of the different pathways at the three time points. The first pathways to reach their lowest values are protein import for TgATrx1 depletion and gene expression for TgATrx2 depletion, and this occurs at 48 hours of ATc treatment. At 72 hours, all pathways show defects likely due to secondary effects. Changes in plastid gene expression can be linked to changes in redox state (reviewed e. g. in [17,18]). To test the apicoplast stroma redox state under TgATrx2 depletion we employed the redox sensitive GFP (roGFP) molecules [19,20] in Toxoplasma. These are GFP molecules with two engineered cysteines that can form a disulphide bond, and with two fluorescence excitation peaks at 385 nm and 470 nm. The 385/470 nm ratio increases upon oxidation and decreases upon reduction of roGFP [21,22]. Since these reporters have not been used in Toxoplasma before, we first tested the suitability of two roGFP variants, roGFP1 and roGFP1-iL [21,22] to report on the parasite’s cytosolic redox state. The two roGFP proteins were transiently expressed and cytosolic localization was confirmed by live fluorescence microscopy (Fig 6A). The change in 385/470 nm ratio was measured in live cells at steady state and during the addition of oxidizing and reducing agents, to assess the dynamic range of each probe. roGFP1-iL is fully reduced in the parasite cytosol at steady state (Fig 6B) and is thus not suitable to assess any potential fluctuation in redox conditions in this compartment. roGFP1 is predominantly, but not fully, reduced (Fig 6Ci), and thus is suitable for measuring redox changes. Neither the parental strain nor the TgATrx2 conditional knockdown, showed a difference in the cytosolic redox state between ATc treated (72 hours) and non-treated parasites as measured by the roGFP1 assay (Fig 6Ci, ii). To measure the apicoplast redox state, each roGFP variant was fused to the apicoplast-targeting sequence of ferredoxin NADP reductase (FNR) [23], and apicoplast localization was confirmed by fluorescence microscopy (Fig 6D). FNR-roGFP-iL is fully reduced in the apicoplast stroma and thus not suitable to follow redox fluctuation in this compartment (Fig 6E). In contrast, FNR-roGFP1 showed that the steady state ratios lie within this probe’s dynamic range, making it suitable to measure apicoplast redox changes (Fig 6F). Upon incubation of TATiΔKu80PIATrx2-3HA with ATc, the dynamic range of FNR-roGFP1 changed within 48 hours (Fig 6F). Changes in the dynamic range of a roGFP-based probe due to high compartment oxidation have been observed when another redox probe, Grx1-roGFP2, was used to measure apicoplast redox state in Plasmodium parasites in response to drug treatment [24]. We suggest that the change herein is similarly the result of apicoplast oxidation due to TgATrx2 depletion. Thus, the first identifiable and most prominent phenotype of TgATrx2 depletion is reduction of apicoplast transcription, followed by a reduction in apicoplast genome maintenance. The onset of transcription defect coincides with a potential redox stress in the apicoplast stroma. We reasoned that identification of TgATrx2 substrates via pull-down might shed further light on its role in controlling apicoplast functions. In some cases, the interaction of Trxs with their substrates can be identified by pull-down of the wild-type (CXXC containing) Trx (e. g. [25]). More transiently associated Trx substrates can be identified using substrate trap mutants. Trap mutants lack the second cysteine of the CXXC motif, which is responsible for resolving the sulfhydryl bonds between Trx and substrate, thus stabilizing the mixed disulphide intermediate between them (illustrated in Fig 1B). This allows pull-down of the covalently coupled substrate along with Trx (e. g. [25,26]). TgATrx2 is not abundant (http: //toxodb. org/toxo/) potentially limiting the identification of its interactors via pull-down at the native levels of expression. To overcome this limitation, we attempted to isolate stable transfectants overexpressing TgATrx2CXXA. This was not successful. Therefore, we engineered an inducible system in which the coding region for TgATrx2CXXA-Myc is separated from a strong promoter by the fluorescent protein KillerRed ORF flanked by LoxP sites (Fig 7A). When integrated into T. gondii expressing inducible Cre-recombinase [27], rapamycin addition results in KillerRed excision and expression of TgATrx2-Myc (Fig 7B and 7C). Pull-down experiments were performed in triplicate with parasites expressing inducible TgATrx2CXXA-Myc or TgATrx2CXXC-Myc for 72 hours, and with the parental line expressing no Myc-tagged protein (control for non-specific interaction with the Myc-trap beads). Fig 7D shows a representative Western blot of one of these pull-down experiments. The whole eluate for each pull-down was analysed by mass spectrometry (S1 Table lists all MS results). Table 1 lists the identified proteins that appeared more than once, with more than two peptides, in the triplicates of TgATrx2CXXA-Myc or TgATrx2CXXC-Myc expressing lines while absent from the triplicates of parental line control; and that were identified with Mascot probability score 20 or above. Ten proteins comply with these criteria: three were pulled down by both forms, one by TgATrx2CXXA-Myc only, and six by TgATrx2CXXC-Myc only. These 10 proteins include a hypothetical protein, a predicted cation efflux transporter and five proteins with predicted functional domains and homology regions shared with factors controlling gene expression and translation (Table 1). Finally, two ribosomal proteins (RPs) were found, RPS25 and RPL4, however, these were previously assigned to cytosolic ribosomes [28] and are likely contaminant in the experiment. Among the putative interactors of TgATrx2CXXC-Myc was TgATrx2 itself. We confirmed this interaction by co-immunoprecipitation analysis, using a cell line co-expressing TgATrx2-HA and TgATrx2-Myc. Anti-Myc antibody recovered the HA-tagged copy and this interaction was decreased by reduction of the disulphide bonds (Fig 7E). These findings suggest that TgATrx2 forms oligomers. To test an additional potential interactor, we selected TGME49_292320, which encodes for a putative tRNA guanine transglycosylase, and which was consistently represented strongly in the pull-down experiments (Table 1), and generated a line where it is endogenously tagged with HA. Immunofluorescence analysis found TGME49_292320 mainly in the parasite nucleus, however with an occasional but reproducible additional apicoplast signal (Fig 7F). Pull-down of TgATrx2CXXA-Myc introduced into this line validated the interaction between the two proteins by Myc-pull-downs recovering HA-tagged TGME49_292320 and vice versa (Fig 7G). Taken together TgATrx2 interactions and the phenotype of its depletion tie its role to the control of the expression of apicoplast genome encoded genes.
The notion that ATrx1 and ATrx2 have roles in separate pathways is supported by their divergent sequences and phylogenetic histories. The distribution and predicted targeting of ATrx2 orthologues suggest a stable old association with plastids derived from red algae. The universal requirement for transcription regulation control in plastids is consistent with ATrx2’s broad maintenance. Plastid ATrx1, on the other hand, has a more restricted distribution. Dinoflagellates lack the plastid ATrx1, but they also lack one plastid outer compartment as only three membranes surround their plastids. Divergence in plastid protein trafficking pathways between apicomplexans and dinoflagellates is, therefore, consistent with a role of TgATrx1 in trafficking. The pattern of inheritance of ATrx1 sequences provided an unexpected perspective on the apicomplexan plastid origin. Plastid gain through endosymbiosis in apicomplexans is argued to have occurred after divergence from ciliates [32], but the source of the plastid remains an open discussion point. Recent plastid phylogenomic analysis suggested that the chromerid/apicomplexan plastid was gained by tertiary endosymbiosis of a non-diatom heterokont [33]. ATrx1 phylogeny indicates that this apicomplexan/chromerid plastid protein was gained from such heterokonts, congruent with this scenario. Lack of plastid ATrx1 paralogues in dinoflagellates could indicate loss of ATrx1 along with loss of one plastid membrane, or that the source of the dinoflagellate plastid is different from that of apicomplexan/chromerids. The loss of the active site CPPC motif in haemosporins and piroplasms is puzzling and might indicate a diverged function of this plastid ATrx1 and participation of alternative Trx proteins in these taxa. TgATrx1 depletion results in the reduction of mature apicoplast proteins which starts when TgATrx1 is fully depleted and reaches below detection levels prior to complete organelle loss (Fig 5, S5 Fig). This response is the same as mutants of different apicoplast import components [10,11,14,34,35]. We therefore suggest that TgATrx1 is directly involved in the control of apicoplast protein import. The previous detection of TgATrx1 within vesicles and in the ER via electron microscopy [9] raises the possibility that it functions in trafficking from the ER to the apicoplast outer membrane. The observed enhanced retention in foci that co-localize with the ER upon active site mutagenesis (S4 Fig) is consistent with this notion. However, regarding the latter observation it cannot be excluded that this retention may be the result of misfolding. It is possible that TgATrx1 is part of the machinery that separates ER-derived vesicles with apicoplast protein content from vesicles destined elsewhere. Further analysis is required to test this hypothesis. Two main mechanisms are proposed in the literature whereby disulphide exchange controls protein import: (1) the exchange leads to conformational changes of translocation pore-complex components thus controlling their permeability [36–38]. (2) Disulphide exchange modulates the conformation of the proteins in transit thus affecting their transport competence. For example, in mammalian and yeast cells, members of the protein disulphide isomerase (PDI) family, which contain Trx domains, modulate the folding of secretory proteins in the ER to achieve forms suitable for their onwards transition [7]. It has been suggested that the outermost apicoplast compartment is of ER origin [6]. The second mechanism may apply for TgATrx1 which might act in an analogous manner to PDIs, engaging with substrates that are flowing from the ER to, and perhaps also through, the outermost apicoplast compartment. TgATrx1 might assist in controlling their conformational state, thus regulating their onward translocation to the inner compartments. The partial accumulation of TgATrx1CXXA mutant protein in the ER could be explained by this model whereby trapped TgATrx1-substrate complexes might be incompetent for forward trafficking from the ER. Put in the context of the above hypothesis of TgATrx1’s role in vesicle sorting, it is possible that TgATrx1 escorts proteins destined to the apicoplast within ER-derived vesicles while keeping them in a translocation competent conformational state. A possibility that TgATrx1 controls import via the first mechanism is also feasible. In that case, loss of translocon permeability upon TgATrx1CXXA over expression would result in build-up of a protein backlog in the ER. Redox regulation of protein import via Trx proteins as means to integrate endosymbiotic organelles into the cell metabolism has been discussed for plant chloroplasts and for mitochondria [39]. Our data suggest that redox control is also active in coordinating apicoplast function in the parasite cell. The first defect detected upon TgATrx2 depletion is reduction of apicoplast transcription followed by reduction in genome copy number (Fig 5, S5 Fig). Redox regulation of gene expression is well documented. Transcription factors (recently reviewed in [40]), ribosomal proteins, elongation factors and helicases are targets of Trx proteins in different organisms [40–44] including Plasmodium [45]. Likewise, ATrx2 pull-downs identified putative substrates with likely roles in gene expression: two candidates have potential roles in transcription (TGME49_259250,310290) and five in translation (TGME49_210360,231140,309120,295070,292320) (Table 1). A mechanistic model of how the exchange between Trx and gene expression factors controls their function is proposed in some cases. The function of the transcription factors NF-κB is redox modulated in two ways. The DNA binding activity of NF-κB is stimulated by reduction of a disulphide bond via hTrx1 [46]. Additionally, the translocation of NF-κB from the cytosol to the nucleus requires its dissociation from I-κB, which is mediated by hTrx1 [47,48]. Since we suspect that TgATrx2 is a PPC resident and transcription occurs in the apicoplast stroma, we hypothesise that it may affect the translocation of its substrates. For example, TgATrx2 may mediate the folding of its substrates or their interaction with chaperones or escort molecules thus controlling their translocation from the PPC towards the stroma (this hypothesis is illustrated in the model in Fig 8). One of the TgATrx2 substrates that we identified (TGME49_292320) is a predicted tRNA guanine transglycosylase (tGT) (Table 1). tGT typically modifies tRNAs with queuine, which affects the efficiency and fidelity of translation [49]. The dual targeting to the nucleus and the apicoplast and reciprocal co-immunoprecipitation (Fig 7) validate the interaction identified by the trap experiments and thus provide support for the role of TgATrx2 in the control of gene expression via disulphide exchange with translation components. The pull-down experiments detected inter-molecular disulphide-bonded forms of the wild-type but not substrate-trapping mutant of TgATrx2 (Table 1), raising the possibility that the oligomerization observed is blocked when TgATrx2 and its substrate (s) are trapped. This suggests that the interactions with substrates and with other TgATrx2 molecules occur via the CXXC motif and that they compete. Alternatively, the C-terminal extension may mediate oligomerization in a way that is sensitive to substrate trapping. Dimerization is observed in other Trx proteins. For example, the bacterial PDI DsbC operates as a homodimer and acts both as a chaperone and disulphide bond facilitator. Loss of dimerization reduces both these activities [50]. The pull-down experiments further identified an interaction of TgATrx2 with the ER chaperone BiP (Table 1). BiP has no cysteines thus the interaction cannot be mediated directly via disulphide bonds. Some PDIs have non-covalent interaction with BiP [25,51]. It is proposed that PDIs resolve non-native disulphides in substrates that become misfolded and targeted to BiP. For this to take place with TgATrx2, BiP should target the apicoplast peripheral compartment where TgATrx2 resides. Like in other organisms, BiP in Toxoplasma likely localizes to the ER [52,53] however its potential presence in compartments of the apicoplast has not been fully assessed. We have used roGFP molecules for the first time in Toxoplasma (Fig 6). roGFP-iL, which has high reduction potential, was fully reduced at steady state in both the cytosol and apicoplast stroma. On the other hand, roGFP1, which has low reduction potential, was partially oxidized in the cytosol and the apicoplast, making it suitable for studies of those compartments. The observation of the apicoplast being reduced corroborates a previous study that used the redox sensitive dimerization of acyl-carrier-protein (ACP) as a redox indicator [54]. A study in P. falciparum, which used a glutathione-specific sensor, also based on roGFP, similarly revealed a reducing environment in the apicoplast stroma [24]. The change in the dynamic range of FNR-roGFP1 upon TgATrx2 depletion (Fig 6) was unexpected and may indicate oxidation of the stroma [24]. Since the observed translation defect appeared simultaneously with this change, it is hard to determine the order of events. It is unlikely that TgATrx2 directly effects the stromal redox state as it does not seem to reside in the stroma (Fig 3iv). The hypothesis that TgATrx2 depletion affects the transport of translation and transcription components into the stroma, suggests that impaired gene expression directly or indirectly leads to the potential oxidation. We have shown that two apicoplast-specific Trxs are essential for its function and parasite survival. The ATrx2 orthologues have a conserved non-canonical CXXC motif, which is different from hTrx1 and 2 in the identity of the two middle residues. Moreover, the C-terminal extension found in ATrx2 orthologues is not typical for human thioredoxins. These features are conserved in the ATrx2 orthologues from Plasmodium spp and other disease-causing apicomplexans, making ATrx2 a particularly attractive drug target candidate. This is supported by findings from the whole genome screen for genes important for growth in blood-stages, that P. berghei ATrx2 is likely essential [55]. The in vitro activity assay established here could be used to develop a platform for inhibitor screening [13]. Furthermore, some anti-malarials currently used in the clinic cause redox stress in the Plasmodium apicoplast [24]. Thus, targeting redox-sensitive apicoplast pathways offers exciting prospects for combination therapy using ATrx-targeting drugs.
T. gondii tachyzoites were grown in human foreskin fibroblasts (HFF, obtained from ATCC, catalogue number #CRC1041). per standard techniques [56]. Where relevant, we added anhydrotetracycline (ATc) to the growth medium at a final concentration of 0. 5 μg/mL. For plaque assays, fresh monolayers of HFF were infected with parasites in the presence or absence of 0. 5 μg/mL ATc for 7 days. Fixation, staining and visualization were performed as previously described [56]. Design of conditional knockdown vectors: we constructed a vector to guide the insertion of the Tet-inducible promoter [57] between the putative start of ATrx1 and ATrx2 to their putative promoter as described before [10]. Fragments corresponding to the upstream (2877/1047bp) and downstream (1050/582bp) region of the ATrx1/ATrx2 start codons respectively were amplified by PCR using primers 1–8 (S1 Table lists all primers) and inserted into the NdeI and the BglII/AvrII restriction sites of pDT7S4PPP1. 1myc [10] respectively. Following transfection into TATiΔKu80 or TATiΔKu80PIATrx2-3HA strains, integrants were selected using 1 μM pyrimethamine and confirmed by PCR using primers 9–17 (S1 Table). Complementation vectors were generated by cloning ATrx1 and ATrx2 minigenes into BglII and AvrII restriction sites within pUPRT_ (TUB) PPP1Ty [10]. The cysteine to alanine mutation was introduced by site-directed mutagenesis using primers 18–21 (S1 Table). These plasmids were transfected by electroporation into the TATiΔKu80PIATrx1 and TATiΔKu80PIATrx2-3HA lines and stable transgenic lines were selected in 5 μM FUDR. For expression of myc-tagged TgATrx2 in the lines with endogenous HA-tagged TgATrx2 and TGME49_292320, the TgATrx2 minigene was cloned between BglII and AvrII restriction sites within pDT7S4-PPP1-myc [10] using primers 22–23 (S1 Table) creating pDT7S4-ATrx2-myc. Stable lines were selected using 1 μM pyrimethamine and confirmed by immunofluorescence with an anti-Myc (Thermo-scientific) antibody. To create a switch-on system the fragment ATrx2-Myc-3’UTR was sub-cloned from the vector pDT7S4-ATrx2-myc with primers 24–25 (S1 Table) to the vector loxP-KillerRed-loxP-YFP (kindly given by Markus Meissner) [27]. The fragment was inserted between BglII and NotI restriction sites using primers ATrx2-BglII-loxP-MfeI-F and ATrx2-NotI-R, thereby replacing the original YFP sequence on the vector. Insertion between BglII and NotI removed the second loxP site on the original vector; therefore a new loxP site was introduced via the forward primer ATrx2-BglII-loxP-MfeI-F. These plasmids were linearized with ScaI and transfected by electroporation into the RH DiCre line (kindly given by Markus Meissner) [27]. The stable transgenic lines were selected using mycophenolic acid (25 μg/mL) and xanthine (50 μg/mL) one day after transfection. To generate cytosolic roGFP expressing vectors the roGFP1 and roGFP-iL-KDEL constructs [22] were used as a template with primers 28–29 and 26–27 respectively. For the latter, both the Erp57 signal sequence at the 5' end and the KDEL-coding sequence at the 3' end, used to direct the protein in the ER, were deleted by primer design. The resulting amplicons were each cloned into the Toxoplasma TUB8mycGFPMyoATy expression vector between EcoRI and PacI restriction sites downstream a Myc tag. The new TUB8-myc-roGFP-iL plasmid was transfected by electroporation into the TATiΔKu80PIATrx2-3HA line and the resulting transgenic parasite population was enriched by three rounds of passages in culture plus cell sorting [10] before isolation of stable clones. The new TATiΔKu80PIATrx2-3HA-roGFP1-iL line was confirmed by immunofluorescence performed on isolated clones. To generate apicoplast roGFP expressing vectors the FNR leader sequence was amplified from FNR-RFP [23] template using primers 30–31 and cloned using EcoRI/BstBI into the EcoRI site in each of the two above mentioned pTUB8roGFP vectors. Total parasite lysates (collected at 1500 g, 10 min, RT and lysed in 1x sample buffer), or immunoprecipitation fractions (105 or 107 parasites per lane respectively) were separated by SDS-PAGE and used for immunoblot analyses. After blocking in Odyssey block (LI-COR Biosciences) or in 1X TBS, Tween-0. 2%, 5% BSA, blots were probed with: monoclonal mouse anti-ATrx1 11G8 at 1: 5000 [9]; mouse anti-HA mAb at 0. 1 μg/mL (Covance); rat anti-HA (Sigma-Aldrich, 1: 50); Mouse-anti-His (Amersham (GE), 1: 1000); rabbit anti-c-Myc (Thermo Scientific, 1: 1,000); rabbit anti-Mic5 (gift of Dr. Vern Carruthers, 1: 10,000) and anti-actin. This was followed by goat anti-mouse Ig coupled to IRDye 800 (1: 10,000, LI-COR) or goat anti-rabbit Ig coupled to IRDye 680 (1: 10,000, LI-COR), or goat anti-rabbit or anti-mouse HRP conjugated (Promega, 1: 10,000). Parasites were grown within HFF on coverslips. Immunofluorescence assays were carried out as indicated previously ([14] or [58]). Antibodies and concentrations used were: rat anti-HA (Sigma-Aldrich, 1: 50); rabbit anti-c-Myc (Thermo Scientific, 1: 1,000); rabbit anti-CSP60 (Reff, 1: 1,000); FITC-coupled rat anti-HA (Roche, 3 μg/mL); rabbit anti-IMC1 (gift of Con Beckers, 1: 1,000); in Fig 3 the marker for the apicoplast stroma was the naturally biotinylated apicoplast luminal protein acetyl CoA carboxylase revealed by Texas Red coupled-streptavidin (Invitrogen, 1 μg/mL) [59]. Secondary antibodies: Cy2 goat anti-rabbit, Cy3 goat anti-Rabbit, Cy2 goat anti-mouse, Cy3 goat anti-mouse (all Jackson Immuno Research Laboratories, 1: 2,000). Images were either taken using a Delta Vision microscope, or SuperResolution Structural Illumination Microscopy (SR-SIM). For Deltavision, an RT deconvolution microscope with an Olympus UPlan/Apo 100x 1. 35 NA objective was used to view the slides. Images were deconvolved using softWoRx (version 3. 5. 1) using standard parameters and a conservative ratio algorithm. For SuperResolution stacks of 30–40 images were taken with increments of 0. 091 μm in a Zeiss Elyra SuperResolution microscope (Jena, Germany) with a 63x oil immersion objective and an immersion oil with a refractive index of 1. 518 (Zeiss, Germany). SuperResolution images were generated using ZEN software (version Zen 2012 SP1, Zeiss, Germany) and processed into their final form using FIJI software [60]. TATiΔKu80PIATrx2-3HA-roGFP1-iL or TATiΔKu80PIATrx2-3HA parasites transiently transfected with TUB8rogGFP1 were grown on HFF monolayers on coverslips in the presence or absence of ATc. Cells were rinsed three times with HEPES buffer (20 mM Hepes pH 7. 4 containing 130 mM NaCl, 5 mM KCl, 1 mM CaCl2,1 mM MgCl2,10 mM D-glucose), transferred to a microscope chamber and incubated in HEPES buffer at room temperature. A Zeiss Axio Observer A1 inverted microscope equipped with a 40X oil immersionFluor lens was used to image the cells. Fluorescence excitation light was generated by the Colibri illumination system, which alternated the excitation wavelength between 385 and 470 nm. Fluorescence emission at 510 nm was monitored by the computer-controlled AxioVision software. Regions of interest were determined manually including the background. Sequential images were collected every minute and exposure to excitation light was 80–200 ms/image. Cells were oxidized with 1 mM diamid and reduced with 10 mM DTT after 3 and 6 min, respectively. To determine the relative ratio of reduced and oxidized roGFPs, 385/470 nm ratios were determined from the fluorescence intensities of regions of interest after background subtraction. Western blot of transiently expressed proteins: TATiΔKu80iATrx1pi or TATiΔKu80PIATrx2-3HA parasites were grown with/out ATc for a given period, then transiently transfected with pBT_LytB or pTUB8-PPP1-HA [14] and let to grow for an additional 24 hours to reach the total desired time of treatment (for example for 72 hours +ATc time point, parasites were grown for 48 hours in ATc, transfected and then grown for an additional 24 hours in ATc). Transfected and treated parasites were collected, total parasite lysate was then separated by SDS-PAGE and blotted using anti-Ty or anti-HA antibodies as described above. In each experiment, parasites were cultured in triplicates with or without ATc for 24,48 or 72 hours). Nucleic acids were purified from parasite pellets obtained from fully egressed cultures using the RNeasy (for total RNA) and DNeasy (for gDNA) kits (QIAGEN) and following manufacturer’s instructions. DNA contamination was removed from RNA samples using the Turbo DNA-free kit (ThermoFisher Scientific) and samples were reverse transcribed to cDNA using the RETROscript kit (ThermoFisher Scientific). Concentrations of 20 ng of either gDNA or cDNA were then used in each qPCR reaction, which was set up with Power SYBR Green Master Mix (ThermoFisher Scientific) and using 300 nM of each primer. All qPCR reactions were performed using a 7500 Real Time PCR System (Applied Biosystems) using default temperature settings and performing a dissociation step after each run. Relative gene expression was determined using the double Δ Ct method [61]; the same methodology was also used to estimate differences in gene copy numbers as described by Ferreira and colleagues [15]. Immunopurification: RH DiCre loxP KillerRed loxP ATrx2 myc parasites were cultured in presence of 50 nM Rapamycin for 72 hours. Parasites were collected, rinsed with PBS supplemented with 20 mM NEM, and lysed for 30 min on ice in lysis buffer (10 mM Tris/HCl pH 8,150 mM NaCl, 0. 5 mM EDTA, 0. 5% (v/v) NP-40,20 mM NEM, 1 mM PMSF) supplemented with protease inhibitor cocktail (Roche). The lysate was span at 20,000 × g for 10 min at 4°C), and the supernatant submitted to denaturation with 0. 2% (v/v) SDS for 2 min at 100°C. The sample was pre-cleared with agarose beads (10% slurry) for 30 min at 4°C with rotation. The pre-cleared input was then applied to either Myc-Trap_A agarose beads (Chromotek) or HA epitope Tag Antibody (Pierce) agarose beads overnight at 4°C with rotation. The flow-through was collected and beads were washed three times in dilution buffer (10 mM Tris/HCl pH 8,150 mM NaCl, 0. 5 mM EDTA, 1 mM PMSF) supplemented with protease inhibitor cocktail. Substrates bound to the complex anti-myc-ATrx2-myc were released by incubation with 25 mM DTT for 10 min at RT, and the resulting eluate stored at -20°C. A second elution with 25 mM DTT for 10 min at 95°C allowed detachment of ATrx2-myc from the beads. Input, flow-through, washes and second elution samples were separated by SDS-PAGE and blotted using a rabbit anti-myc antibody (ThermoFisher Scientific, 1: 1,000). Mass spectrometry: Proteins were identified using nanoflow HPLC electrospray tandem mass spectrometry (nLC-ESI. MS/MS) at Glasgow Polyomics. Tryptic peptides, generated using the FASP procedure [62] were analyzed as previously described [63]. During result analysis, only peptides with Mascot score of 20 and above (namely the probability that this match might be a random event is 10−2 or lower) were included in the analysis. Alignments were generated using MAFFT [64], manually corrected and ambiguous sites removed. 364 sites were used. Maximum likelihood phylogenies were performed using RAxML v8. 1. 17 [65] using the best-fit model (LG+G+F) inferred by Prottest 2. 4 [66] and 100 resamplings for bootstrap calculations. Bayesian analyses were performed with MrBayes 3. 2. 6 [67] hosted on the CIPRES Science Gateway webportal (https: //www. phylo. org/, last accessed May 24,2017) [68]. | To survive, apicomplexan parasites must adjust to the redox insults they experience. These parasites undergo redox stresses induced by the host cell within which they live, by the host immune system, and by their own metabolic activities. Yet the myriad of cellular processes that are affected by redox changes and that may take part in maintaining the redox balance within the parasite are largely understudied. Thioredoxins are enzymes that link the redox state of subcellular environments to the functional state or the cellular trafficking of their substrate proteins. In this work, we identify two pathways that are controlled by two thioredoxins in the apicomplexan Toxoplasma gondii, and demonstrate that both are essential for parasite survival. We show that each of these enzymes contributes to the function of the apicomplexan plastid, the apicoplast, a unique parasite organelle with importance for drug discovery efforts. We thus highlight that part of the apicomplexan sensitivity to redox imbalance is specifically related to the apicoplast, and point at the importance of thioredoxins in mediating apicoplast biogenesis. Finally, our work raises the potential of apicoplast thioredoxins as new drug targets. | Abstract
Introduction
Results
Discussion
Materials and methods | taxonomy
plastids
parasite groups
plant cell biology
light microscopy
parasitology
apicomplexa
plant science
phylogenetics
data management
microscopy
sequence motif analysis
research and analysis methods
oxidation-reduction reactions
sequence analysis
computer and information sciences
bioinformatics
fluorescence microscopy
gene expression
chemistry
evolutionary systematics
eukaryota
dinoflagellates
cell biology
electrochemistry
database and informatics methods
genetics
biology and life sciences
chemical reactions
physical sciences
evolutionary biology
protists
organisms | 2018 | Two essential Thioredoxins mediate apicoplast biogenesis, protein import, and gene expression in Toxoplasma gondii | 13,356 | 301 |
Mechanisms underlying the dramatic patterns of genome size variation across the tree of life remain mysterious. Effective population size (Ne) has been proposed as a major driver of genome size: selection is expected to efficiently weed out deleterious mutations increasing genome size in lineages with large (but not small) Ne. Strong support for this model was claimed from a comparative analysis of Neu and genome size for ≈30 phylogenetically diverse species ranging from bacteria to vertebrates, but analyses at that scale have so far failed to account for phylogenetic nonindependence of species. In our reanalysis, accounting for phylogenetic history substantially altered the perceived strength of the relationship between Neu and genomic attributes: there were no statistically significant associations between Neu and gene number, intron size, intron number, the half-life of gene duplicates, transposon number, transposons as a fraction of the genome, or overall genome size. We conclude that current datasets do not support the hypothesis of a mechanistic connection between Ne and these genomic attributes, and we suggest that further progress requires larger datasets, phylogenetic comparative methods, more robust estimators of genetic drift, and a multivariate approach that accounts for correlations between putative explanatory variables.
The vast array of genome sizes is a pattern that begs for explanation [1], [2]. Haploid (1C) genome size (measured either in base pairs or mass, where 106 Kb ≈1 picogram) spans eight orders of magnitude: the known eukaryotic range is ≈2,249–978,000,000 Kb [3], while Archaea and Bacteria range from 491–5,751 Kb and 76–13,034 Kb, respectively [4]. Lynch and colleagues [5]–[7] have argued strongly for a central role for nonadaptive processes such as mutation and drift in the evolution of genome size and complexity. In contrast to proposed neutral and adaptive models of genome size evolution (see, e. g. [8], [9]), they outline a model positing that mutations increasing genome size are slightly deleterious. Under this model, lineages differ in effective population size (Ne) and, as a result, differ in the efficacy with which natural selection will counteract genome expansion. Thus, lineages with small Ne will experience drift towards larger genomes [7]. As support for their argument, they presented a comparative analysis of roughly 30 taxa, ranging from bacteria to angiosperms, fungi, and mammals. Among these taxa, they reported a statistically significant negative relationship between Neu (a composite parameter including effective population size and nucleotide mutation rate) and genome size. Strikingly, the relationship was quite strong: 66% of the variation in genome size was explained by Neu [7]. This is truly an astounding result, considering the widely divergent selective regimes, life histories, and modes of reproduction found across these diverse organisms. The Lynch & Conery model has sparked intense interest and >330 citations. Some objections on theoretical and methodological grounds have been voiced. Charlesworth and Barton [10] point out that Ne is confounded with many different aspects of organismal biology (e. g. , developmental rate, body size), and thus that both Ne and genome size may be correlated effects of one or more other causal factors. Daubin and Moran [11] outline several objections, including that taxon differences in mutation rates make Neu a poor proxy for Ne that estimates of Ne from silent-site nucleotide diversity in bacteria (as in [7]) are skewed by population subdivision and cryptic species, and further that such Ne estimates are overly sensitive to recent evolutionary history. Nevertheless, the idea that Ne drives genome size and complexity seems to have gained acceptance [12]–[14], with some going so far as to characterize it as “the principal explanatory framework for understanding the evolution of genome organization” ([12], p. 303). Here, we argue that such conclusions are premature without phylogenetic comparative analyses of genome size evolution. When species are used as data points, relationships between raw values of any two traits (e. g. , Ne and genome size) are difficult to interpret, as shared phylogenetic history means that assumptions of statistical independence are likely to be violated [15]–[17]. Special methods are required to recover independence of observations and to test for evolutionary associations between traits. Frequently, conventional (nonphylogenetic) analyses overestimate the strength of the association between traits relative to phylogenetic methods [18]. In an extreme case, a strong correlation in the raw data can be driven by a single association at the base of the phylogenetic tree, e. g. , it can reflect a single instance of correlated change in the traits, followed by uncorrelated changes and/or stasis in trait values during subsequent evolutionary history (Figure 1). In this study, we revisit the Lynch & Conery dataset with a phylogenetic perspective, taking advantage of new phylogenetic data and analysis tools.
A phylogenetic topology and reconstruction of genome sizes is presented in Figure 2, illustrating that close relatives have similar genome sizes. Initial simple linear regressions of genome size on Neu explored four branch length models and found that the phylogenetic generalized least squares (PGLS) model with all branches = 1. 0 provided a better fit than the nonphylogenetic ordinary least squares (OLS) model (Table 1). Subsequent analyses therefore used branch lengths of 1. 0. For all variables except intron number, phylogenetic models (PGLS) exhibited better fit than nonphylogenetic (OLS) models (Table 1). For genome size and gene number, estimation of the Ornstein-Uhlenbeck transformation parameter d indicated substantial phylogenetic signal (d = 1. 31 and 1. 16, respectively), and the resulting RegOU models fit significantly better than the OLS models (ln likelihood ratio tests (LRTs), χ2 = 5. 88, P = 0. 015 and χ2 = 7. 90, P = 0. 005, respectively). In comparing the two phylogenetic models, the RegOU model did not produce significantly better fit vs. PGLS (LRTs, χ2 = 1. 84, P = 0. 175 and χ2 = 0. 46, P = 0. 498 for genome size and gene number, respectively). Although there were strong negative relationships between Neu and six of the seven genomic attributes in nonphylogenetic regressions, the patterns disappeared when phylogenetic models were applied (Table 1). For example, the strong negative relationship between Neu and genome size (OLS, P<0. 001, Figure 3A) was replaced with a nonsignificant relationship under better-fitting phylogenetic models (PGLS, P = 0. 137, Figure 3B; RegOU, P = 0. 328). Similar patterns were evident for gene number, the half-life of gene duplicates, intron size, intron number, transposon number, and transposon fraction (Table 1).
Accounting for phylogenetic history substantially altered the perceived strength of the relationship between Neu and genomic attributes. In phylogenetic analyses, there were no consistent evolutionary associations between Neu and gene number, intron size, intron number, the half-life of gene duplicates, transposon number, transposons as a fraction of the genome, or overall genome size. Thus, a phylogenetically controlled reanalysis of the Lynch & Conery dataset [7] does not support the conclusion that Ne drives genome size patterns across the tree of life. The few existing comparative analyses of more phylogenetically restricted datasets either do not support or provide only equivocal support for the Lynch & Conery model. Whitney et al. [19] conducted a phylogenetically controlled analysis of 205 species of seed plants and found no association between Ne and genome size. Kuo et al. [20] analyzed 42 paired bacterial genomes, using the efficacy of purifying selection in coding regions to quantify genetic drift. Bacterial taxa experiencing greater levels of genetic drift – implying a smaller evolutionary Ne – had smaller genomes, a pattern opposite that predicted by the Lynch & Conery model as articulated in [7]. Finally, in putative support of the model, Yi & Streelman [21] reported a significant negative relationship between Ne and genome size in a phylogenetically corrected analysis of 33 species of ray-finned fish. However, this analysis has been challenged as artifactual. Gregory & Witt [22] argue that Pleistocene population bottlenecks and polyploidy shaped both Ne and genome size of fishes in such a way as to generate a non-causal correlation between Ne and genome size in this particular dataset. Future investigations of the role of genetic drift in determining genome size across the tree of life would benefit from several approaches. First, utilizing phylogenetic comparative methods, for which we advocate here, is an important step towards drawing robust inferences from species-level comparative analyses. Second, larger datasets would certainly increase confidence in our interpretations. While statistically nonsignificant, we note the relationships between Neu and genomic attributes (Table 1) are negative and thus are at least qualitatively consistent with the Lynch & Conery model, suggesting that power may be an issue. Furthermore, given that the Neu estimates in the current analysis required sequence data, species with small genomes relative to averages within clades are likely overrepresented; thus it would be important to ensure that species with large genomes are included in future analyses. Third, future studies would benefit from more robust estimates of genetic drift, as Neu estimated from silent-site diversity (as in [7] and the present reanalysis) has several undesirable properties. Because the mutation rate u differs among lineages [11], [23], [24], using Neu as a proxy for Ne could obscure any relationship between Ne and genome size. Further, Ne estimated from silent-site diversity may signal the effects of recent evolutionary events more than the long-term history under which genome size evolved [11]. Ka/Ks ratios (ratios of nonsynonymous to synonymous substitutions per site) are a promising alternative to Neu for estimating genetic drift [11], [20]. Finally, genome size is a complex trait that is unlikely to be explained by univariate analyses [10]. Phylogenetic comparative methods should be combined with multivariate models that are capable of distinguishing the contributions of highly correlated predictor variables. A recent analysis [19] is a step in the right direction: plant outcrossing rate and Ne were simultaneously examined in a multiple regression analysis of phylogenetically independent contrasts, allowing the partial contribution of each variable to be characterized. To make further progress on the population genetics of genome size and complexity, we clearly need phylogenetic comparative analyses of large datasets capable of distinguishing the contributions of Ne and its multiple correlates, including body size, developmental rate, and metabolic rate.
Data on Neu and genome sizes for 22 eukaryotic and 7 prokaryotic species were obtained from the Supporting Online Material of [7]. For a subset of these species, data on gene number, intron size, intron number, and the half-life of gene duplicates were also obtained from the same source. Data on total transposon number and fraction of the genome occupied by transposons were obtained directly from M. Lynch; these data combine counts of LTR, non-LTR, and DNA transposons and correspond to the fourth panel of Fig. 4 of [7]. All traits were log10 transformed prior to analysis; for total transposon number and transposon fraction, constants of 1. 0 and 0. 01, respectively, were added prior to log-transformation. A composite tree for the species was constructed in Mesquite v. 2. 71 [25] based on phylogenetic trees reported in [26]–[28]. As a visual heuristic, genome sizes were traced onto the phylogeny using the Parsimony Ancestral States method [29] with an assumption that all branch lengths equal 1. 0. All dependent variables were regressed on Neu using REGRESSIONv2. m [30] running in MATLAB v. 7. 9. 0. Three types of models were examined: ordinary least squares (OLS), phylogenetic generalized least squares (PGLS), and phylogenetic regression under an Ornstein-Uhlenbeck process (RegOU) [30], [31]. OLS is traditional ‘nonphylogenetic’ regression, which in effect assumes a star phylogeny in which all species are equally unrelated, and corresponds to the Neu vs. genome size analysis reported in [7]. PGLS assumes that residual variation among species is correlated, with the correlation given by a Brownian-motion like process along the specified phylogenetic tree (topology and branch lengths). PGLS is functionally equivalent to Felsenstein' s [15] phylogenetically independent contrast method [31]. Finally, the RegOU model estimates (via restricted maximum likelihood) the strength of phylogenetic signal in the residual variation simultaneously with the regression coefficients; the former is given by d, the Ornstein-Uhlenbeck transformation parameter. An OU evolutionary model is typically used to model the effects of stabilizing selection around an optimum [30]. When d = 0, there is no phylogenetic signal in the residuals from the regression model; when d is significantly greater than 0, significant phylogenetic signal exists [30], [32]. Following [33], starter branch lengths corresponding to all branches = 1. 0, Grafen' s arbitrary lengths, Pagel' s arbitrary lengths, and Nee' s arbitrary lengths were compared in PGLS and RegOU regressions of genome size on Neu. Based on their likelihoods, the models with all branches = 1. 0 achieved the best fit, and thus these branch lengths were used in all subsequent phylogenetic analyses. Model selection for each variable then proceeded in two steps. First, we compared the likelihoods of the PGLS model and the OLS model, with a higher likelihood taken as evidence of a better-fitting model. Second, we used ln likelihood ratio tests (LRTs) to compare the RegOU model with the PGLS and OLS models with 1 d. f. [30]. Given the issue of small sample sizes (see [32]) for most dependent variables and the fact that RegOU models require estimation of an extra parameter, RegOU models were examined only for genome size and gene number. | Genome size (the amount of nuclear DNA) varies tremendously across organisms but is not necessarily correlated with organismal complexity. For example, genome sizes just within the grasses vary nearly 20-fold, but large-genomed grass species are not obviously more complex in terms of morphology or physiology than are the small-genomed species. Recent explanations for genome size variation have instead been dominated by the idea that population size determines genome size: mutations that increase genome size are expected to drift to fixation in species with small populations, but such mutations would be eliminated in species with large populations where natural selection operates at higher efficiency. However, inferences from previous analyses are limited because they fail to recognize that species share evolutionary histories and thus are not necessarily statistically independent. Our analysis takes a phylogenetic perspective and, contrary to previous studies, finds no evidence that genome size or any of its components (e. g. , transposon number, intron number) are related to population size. We suggest that genome size evolution is unlikely to be neatly explained by a single factor such as population size. | Abstract
Introduction
Results
Discussion
Materials and Methods | evolutionary biology
genetics and genomics/population genetics
genetics and genomics
evolutionary biology/evolutionary and comparative genetics | 2010 | Did Genetic Drift Drive Increases in Genome Complexity? | 3,345 | 245 |
Two small quorum sensing (QS) peptides regulate competence in S. mutans in a cell density dependent manner: XIP (sigX inducing peptide) and CSP (competence stimulating peptide). Depending on the environmental conditions isogenic S. mutans cells can split into a competent and non-competent subpopulation. The origin of this population heterogeneity has not been experimentally determined and it is unknown how the two QS systems are connected. We developed a toolbox of single and dual fluorescent reporter strains and systematically knocked out key genes of the competence signaling cascade in the reporter strain backgrounds. By following signal propagation on the single cell level we discovered that the master regulator of competence, the alternative sigma factor SigX, directly controls expression of the response regulator for bacteriocin synthesis ComE. Consequently, a SigX binding motif (cin-box) was identified in the promoter region of comE. Overexpressing the genetic components involved in competence development demonstrated that ComRS represents the origin of bimodality and determines the modality of the downstream regulators SigX and ComE. Moreover these analysis showed that there is no direct regulatory link between the two QS signaling cascades. Competence is induced through a hierarchical XIP signaling cascade, which has no regulatory input from the CSP cascade. CSP exclusively regulates bacteriocin synthesis. We suggest renaming it mutacin inducing peptide (MIP). Finally, using phosphomimetic comE mutants we show that unimodal bacteriocin production is controlled posttranslationally, thus solving the puzzling observation that in complex media competence is observed in a subpopulation only, while at the same time all cells produce bacteriocins. The control of both bacteriocin synthesis and competence through the alternative sigma-factor SigX suggests that S. mutans increases its genetic repertoire via QS controlled predation on neighboring species in its natural habitat.
Horizontal gene transfer in prokaryotes is mediated via three distinct mechanisms comprising conjugation, transduction, and transformation [1,2]. Studying those mechanisms in detail is needed because they are among the reasons for the spread of antibiotic resistance and virulence determinants between bacteria [3]. Natural transformation, i. e. the uptake of extracellular DNA from the environment via genetic competence, is a powerful process able to expand and modify the gene inventory in both Proteobacteria and Firmicutes [4]. It requires a multi-protein complex localized in the cell membrane, many elements of which are highly conserved, and was studied in most detail in Bacillus subtilis and Vibrio cholerae [5,6]. In streptococci competence is a tightly controlled transient state whose activation involves detection of quorum sensing (QS) signaling peptides, which in streptococci are termed pheromones. In all streptococci studied so far the proximal master regulator of competence, and final receiver of the transduced signals, is the alternative sigma factor SigX (previously termed ComX). SigX binding to the RNA polymerase activates transcription of a core set of ~ 20 “late” competence effector genes [7]. They carry a nine bp cin-box in their promoter region and mediate the synthesis of proteins for DNA uptake and recombination [8]. The SigX regulon has been found in all streptococci sequenced to date, suggesting that genetic competence is ubiquitous in streptococci, although until now it could only be demonstrated experimentally in very few [9]. Although all streptococci use peptide pheromones for density dependent activation of sigX expression as well as bacteriocin synthesis, the details of signal transduction and integration differ widely between the different species. Two principal types of pheromones and signal detection mechanisms have been found [10]: Pheromones derived from pre-peptides carrying a double-glycine leader sequence are cleaved and exported by the membrane localized ComAB complex, and after an additional processing step the mature pheromone is detected in the extracellular environment through the transmembrane sensor histidine kinase of a two component signal transduction system that phosphorylates its corresponding response regulator. The phosphorylated response regulator induces transcription of the alternative sigma-factor sigX. The competence-stimulating peptide (CSP) of S. pneumoniae belongs into this group. The second type of pheromones was recently discovered and shown to be conserved in streptococci [7,11–13]. A pre-peptide is synthesized, exported and processed to yield a short hydrophobic mature peptide termed XIP (sigX inducing peptide), which is then re-imported by a permease. Its detection occurs by an intracellular transcription factor that interacts directly with the peptide signal [10]. The activated, dimeric regulator induces transcription of sigX by binding to its promoter region [12], much like the LuxR type QS regulators that are activated by binding of acylated homoserine lactons (AHLs), the QS signals of Proteobacteria. In S. mutans competence is regulated via both types of QS signaling peptides and the medium composition determines which signal is active. Moreover the medium also determines whether competence is induced uni-modally [14] or only in a subpopulation of cells [7,15]. The current understanding of competence development in S. mutans is shown in Fig 1. In complex media competence can only be induced by CSP (Fig 1A). The 46 amino acid CSP precursor is encoded by the comC gene and processed and exported by the comAB encoded ABC transporter, yielding the extracellular 21 residue CSP-peptide [16]. This is cleaved to its biologically active form by the protease SepM, thereby removing the 3 C-terminal amino acids [17]. Binding of CSP-18 to the histidine kinase ComD induces autophosphorylation of the protein and results in the transfer of the phosphorylgroup to a conserved aspartic acid residue of the cognate response regulator ComE. Phosphorylated ComE binds to two direct repeats in the promoter regions of genes encoding bacteriocins and their corresponding immunity proteins resulting in transcription of the “early competence genes” [18–20]. Via an unknown link comDE activation induces comRS expression and finally competence development in a subpopulation of the cells [7,15]. In the peptide free chemically definde medium (CDM), competence can only be induced by XIP, the second autoinducer of S. mutans (Fig 1B), via the ComRS system, that is also present in other mutans, pyogenic and bovis group streptococci [7,21]. It consists of an Rgg type transcriptional regulator (ComR) and a small XIP peptide, encoded by the comS gene [7]. The XIP pre-peptide is processed and the active seven residue pheromone is secreted into the environment by an unknown process. After accumulation extracellular XIP is internalized, most likely via the Opp permease, and binds to ComR in a proposed stoichiometry of 2: 1 (XIP: ComR) [7,12]. Resulting conformational changes promote dimerization of the XIP2/ComR complex and binding to DNA targets containing the ComR binding motif. In S. mutans binding motifs for ComR were found upstream of the comS and the sigX gene [7,12]. Thus XIP controls its own expression in a positive auto-regulatory feedback loop. XIP mediated competence induction is exclusively observed in chemically defined media and results in uni-modal expression of competence. The effect of the medium is hypothesized to be caused by differences in transport and degradation, respectively, of the autoinducers. In S. pneumoniae it was shown that CSP is degraded by the membrane bound protease HtrA, and that this degradation is reduced in the presence of unfolded proteins [22]. Therefore, it was hypothesized that the homologous HtrA of S. mutans might be inhibited from CSP degradation by small peptides present in complex medium, resulting in active CSP in complex but not in minimal medium [15]. Conversely, XIP is active in defined, but not in complex media. It was suggested that the import of XIP might be inhibited in complex media due to the clogging of the Opp permease [15]. Small amounts of peptides that are added to the cultivation medium completely eliminate the activity of XIP [15]. However, it remains unclear how under CSP induced conditions ComR might be activated by XIP, if this signal cannot be imported into the cell (Fig 1A). Bacteria often respond to environmental stimuli in a non-uniform manner; even in isogenic cultures and under homogenous conditions the appearance of multiple phenotypes (phenotypic variation) is observed [23]. Phenotypic variability has strong implications for the treatment of bacterial infections and is relevant for cellular differentiation [24]. Two distinct phenotypes occurring simultaneously are referred to as bimodal and were observed e. g. for lactose utilization, chemotaxis and persister cell occurence in E. coli [25] and for competence development and sporulation in B. subtilis [23,26,27]. It was previously shown that phenotypic heterogeneity enhances the overall fitness of the population under fluctuating conditions and helps bacteria to colonize different ecological niches within an ecosystem [23,28]. Thus phenotypic variability can be considered as a bet-hedging strategy and an evolvable trait. Signaling in bacteria is never discrete since stochastic fluctuations of the components that determine a cellular state occur [29]. This phenomenon is called noise and it is most pronounced for processes involving a limited number of molecules such as transcription and translation [30]. Noise is one critical determinant for the establishment of phenotypic variation [23,29]. Phenotypic heterogeneity can also be a result of the architecture of regulatory networks comprising positive feedback loops or toggle switches (two regulators that negatively regulate each other) which can amplify signals and respond nonlinearly to stimuli [23,24]. Noisy gene expression in conjunction with such networks can convert a graded signal into a bimodal response which is stably maintained in the population. An intriguing feature of the competence cascade in S. mutans is its bimodality under CSP induced conditions in THBY medium. It was hypothesized to be caused by the auto-regulatory feedbackloop of ComS [15]. A different hypothesis was developed by Lemme et al. [14] who found that the entire population expressed the bacteriocin related genes in an uni-modal way, while only a subpopulation expressed sigX and entered the competent state. However, bimodal expression of comE was observed at the same time. Since comE is supposed to act upstream of ComRS signaling, it was therefore hypothesized that comE expression might be the origin of bimodality [14]. Finally, an intriguing question is the autolysis of a fraction of the population after stimulation by XIP [31] or CSP [14,32,33]. In S. pneumoniae, it is the non-competent subpopulation which is killed by the competent siblings, a phenomenon termed fratricide [9]. The CbpD murein hydrolase which is responsible for this process has a homologue termed LytFsm in S. mutans. It has however been shown that lysis occurs in a fraction of the competent cells [14,34] and that LytFsm is a self-acting autolysin, not a fratricin [32]. The QS signaling cascade of S. mutans thus comprises two different signals, one that is detected in the external environment (CSP), and one that is detected intracellularly (XIP). They are linked with each other in different ways in different media. The system has a temporal dimension (“early” and “late” competence genes), and it can result in phenotypic heterogeneity of the population. Here we asked (1) At which level in the competence signaling cascade is bimodal gene expression triggered? (2) How is the ComE signaling cascade connected to the ComRS signaling cascade? (3) How can bacteriocin expression be unimodal, although comE expression is bimodal? (4) How is the medium effect on signaling mediated—is CSP really degraded in CDM, and how can an active XIP/ComR complex be formed in THBY? To answer these questions, it was necessary to follow the activation of the key genes along the signaling cascade on the single cell level. We developed a tool-box of integrative reporter plasmids carrying a fusion of the promoter of interest with a fluorescent protein. Those constructs were integrated into the chromosome of S. mutans. We used the wild-type, as well as strains that had key genes of the signaling cascade deleted, over-expressed, or modified. To monitor co-expression of “early” and “late” competence genes, we developed dual-fluorescent reporter strains that clearly demonstrated which of the cells induced upstream became finally competent. Population heterogeneity was observed directly under the microscope or using flow cytometry. RNA sequencing was used to determine gene expression of all involved components during the first 30 min post induction. The data result in a new understanding of the QS regulatory circuit of S. mutans that resolves the discordant observations described above.
The construction and experimental verification of all single and dual fluorescent reporter strains used in this study is described in detail in S1 Text and S1–S6 and S9 Figs. Using dual fluorescent reporter strains we analyzed signal propagation along the entire CSP-induced competence cascade in S. mutans, from the upstream comDE system to the central regulatory comRS module and the alternative sigma-factor sigX to the late competence gene lytFsm, a component of the transformasome. First the effect of the CSP inducer concentration was evaluated. S7 Fig (flow cytometry) and S8 Fig (microscopy) show that bimodal lytFsm expression was observed above a concentration of 20 nM CSP; the percentage of cells expressing lytFsm was not significantly changed between 200 nM and 100 μM CSP. Thus 2 μM CSP was chosen as inducer concentration in all following experiments. No significant influence on cell growth and viability was observed under these conditions while higher concentrations of CSP induce cell death and growth arrest [34]. The analysis was conducted 2. 5 h post induction because the expression of the late competence genes reaches its maximum after 2 h [14] and the full maturation of the fluorescent proteins requires additional 30 minutes [35,36]. Fig 2 shows that the comE gene is co-expressed with comS, sigX and the late competence gene lytFsm. Moreover and not surprisingly comS is coxpressed with sigX (S5B Fig), since both genes share the same ComR binding sequence in their 5`UTR [7]. , . All four genes are expressed in a bimodal way. The same coexpression patterns were also observed when we switched fluorophores for the two genes under study in the reporter strains (S9 Fig). To conclude, we observed an identical bimodal gene expression pattern for all genes in the signaling cascade. Two mechanistic explanations would be in accordance with the observed coexpression pattern and the bimodality of comE expression. Both models assume a basal expression of comE. To determine whether ComD/ComE or ComR/ComS represent the origin of bimodality during the CSP-induced competence development of S. mutans, we constructed overexpression strains strongly and constitutively expressing those proteins independent from their native promoters and thus equally across the entire population. Overexpression of the cellular component (s) representing the origin of bimodality should cause unimodal lytFsm expression (indicative of unimodal competence) across the population. We utilized plasmid pIB166 [37] and cloned the genes downstream of the strong constitutive Lactococcal P23 promoter. Plasmids were transformed into the LytFsm pAE03 reporter strain. LytFsm is a late competence gene [32] and it is the most strongly expressed late competence gene of the tested reporters (S4 Fig). We used the comRS overexpression strain in defined medium to test whether unimodal lytFsm expression is observed, as expected according to Masburn-Warren et al. [7]. LytFsm expression was unimodally induced even at low cell densities and without addition of external XIP (S10 Fig), indicating that endogeneous production of XIP is sufficient to induce competence unimodally. To analyse the origin of bimodality, the experiments were conducted in THBY under CSP inducing conditions. Fig 3 shows that for all overexpression strains CSP was needed to induce lytFsm expression. In a transcriptional signaling cascade from comDE to sigX it would have been expected that overexpression of one component should be sufficient to induce competence. The strong constitutive overexpression of comE or comD in the entire population did not enhance the percentage of lytFsm induced cells (Figs 3 and S11). Thus we can exclude that ComE or ComD is the origin of bimodality. Overexpression of only comR or only comS still resulted in a biphasic population behavior, although the proportion of LytFsm expressing cells was significantly enhanced compared to the WT or the comE overexpression strain. However, overexpression of both genes together (comRS) from the lactococcal P23 promoter resulted in unimodal expression of LytFsm, proving that comRS represents the origin of bimodality (Figs 3 and S11). ComR needs to be activated by XIP which is synthesized by comS [12]. Thus overexpression of comR or comS alone is not sufficient to induce sigX and the late competence genes. Both comR and comS are needed for the positive feedback loop to occur. Surprisingly, however, unimodal competence development in the comRS overexpression strain absolutely required CSP induction in THBY medium. No competence development was observed in the absence of CSP. A regulator upstream of comRS as assumed currently (Fig 1B) cannot be responsible for this observation, because overexpression of comRS would be able to overcome it. In complex medium it is thought that import of secreted XIP is impossible due to the block of the Opp peremease by small peptides (Fig 1A). The data suggest that CSP itself or induced cellular changes are able to bypass this block. Thus we tested how the percentage of cells expressing the late competence gene lytFsm was influenced by XIP supplementation to a CSP induced LytFsm reporter strain grown in complex medium and how the relative amounts of those two signals affect bimodal gene expression. Accordingly different combinations of CSP and XIP with concentrations ranging from 0. 2 μM to 20 μM of each inducer were tested. The results of the flow cytometric analysis of the LytFsm pAE03 reporter strain 120 minutes post XIP and/or CSP supplement are presented in Fig 4. As expected, XIP supplementation alone did not induce expression of the late competence gene lytFsm, regardless of the used concentration (Fig 4A). Conversely, and as expected, too, addition of CSP alone induced bimodal lytFsm expression. The percentage of cells expressing lytFsm was always constant, independent of the used CSP concentration in a range from 0. 2 μM to 100 μM (Figs 4B and S7 and microscopic images in S8 Fig). However, addition of both XIP and CSP in a ratio of 1: 1 resulted in expression of lytFsm (Fig 4C) while no expression had been observed with XIP alone up to a concentration of 20 μM. At the highest concentration of XIP and CSP tested here, unimodal expression of competence was observed. Lowering the CSP concentration to 0. 2 μM still resulted in unimodal lytFsm expression (Fig 4D). The XIP concentration strongly determines the percentage of lytFsm expressing cells, while the CSP concentration is of minor importance. We hypothesize that extracellular XIP is imported into the cell in a concentration dependent way, but only in the presence of CSP. How can this be achieved? The XIP uptake must be independent of Opp, since its deletion does not affect the CSP induced bimodal competence phenotype in complex medium [15]. We suggest that CSP induces import of XIP in an indirect way by inducing expression of bacteriocins, resulting in permeabilization of the cell. As relatively high XIP concentrations are required, it is likely that the import of the small XIP peptide is ineffective in THBY. To further substantiate this hypothesis we deleted the cipB gene in the LytFsm reporter strain background and tested whether addition of CSP or XIP or the combination of both petides still induces lytFsm expression. Strinkingly no induction of lytFsm expression was found in the CipB deletion strain, regardless which peptide or peptide combination was used for induction (S12 Fig). Accordingly, Perry et al. [34] observed that CipB deletion completely abolished CSP induced competence development. Since there is no direct regulatory link between comDE and comRS (see above), and since competence induction by CSP in complex medium appears to be an indirect effect, we hypothesize that CSP controls only bacteriocin synthesis, while competence is exclusively controlled via XIP. Previously only competence development has been tested in both CDM and THBYmedia, and it was shown that in CDM competence is not induced by CSP, regardless of the used concentration [15]. It was speculated that CSP might be proteolytically degraded by the HtrA protease in CDM, while in complex medium this protease was thought to be saturated by small peptides that are constituents of the medium, thus protecting CSP from degradation [15]. However, it was not yet tested if bacteriocin encoding genes are transcribed in CDM under CSP induced conditions. Using the CipB pMR1 reporter strains in the different deletion backgrounds we therefore tested whether CSP induction in CDM promotes transcriptional activation of bacteriocins. We show (Table 1 and S13 Fig) that CSP promotes a strong induction of cipB expression in CDM and thus is biologically active and not degraded. Induction of cipB expression occurs via ComDE, as deletion of either comD or comE completely abolished it. The induction of cipB expression was unchanged in the ΔcomRS, ΔcomS and ΔsigX background. An influence of those regulatory systems can thus be excluded. To summarize, CSP is active in CDM. It induces transcription of bacteriocin encoding genes via ComDE but does not induce competence via the ComRS system. To demonstrate that XIP signaling induces exclusively competence, while CSP controls only bacteriocin expression, we performed a time resolved analysis of the transcriptional response to these two autoinducers during the first 30 min after supplementation using RNA sequencing. Differentially expressed genes are presented in Fig 5 and are grouped according to quorum sensing related genes, mutacins, competence and others. CSP supplemention in CDM induced the expression of mutacins and their corresponding immunity proteins, while competence related genes were not found to be differentially expressed. The exact opposite behavior was observed for XIP stimulation in CDM, confirming our hypothesis. For the XIP signaling cascade it can be seen that the transcription of the central XIP responsive transcriptional regulator ComR was only very slightly induced, in accordance with the general observation that essential key regulators are unlikely to be strongly transcriptionally regulated. The primary targets of ComR, comS and sigX were instantly induced by XIP, representing the strongest responding transcripts. The up-regulation of the two genes already 5 minutes after XIP addition is in accordance with the post-translational activation of the ComR protein by XIP binding. Interestingly, XIP addition resulted in the activation of comDE expression for the later time-points (15 and 30 min). The CSP signaling cascade, by contrast, did not induce sigX or comS expression as expected. comDE expression was only very slightly induced upon CSP stimulation. However, mutacins, the primary targets of ComE, were strongly transcribed already 5 min post supplementation, suggesting post-transcriptional activation of the regulator ComE. Genes encoding transporters and proteases (ComAB, HtrA, Opp and SepM) as well as comC, encoding the CSP precursor, were not differentially expressed, regardless if XIP or CSP was used for stimulation. Both QS systems responded very fast, within the first 5 minutes, to their stimuli. Thus the results of the transcriptome analysis indicate that QS in S. mutans is initiated by the regulators ComE and ComR on the post-transcriptional level. The transcription of the target genes mutacins, comRS and sigX can already be observed 5–15 minutes after stimulation. The response of the “late competence” genes is slightly delayed compared to that of the mutacins, since the activation of SigX is additionally required. The expression of comE and cipB was confirmed by quantitative PCR (S1 Text and S15 Fig). To summarize our findings up to this point: The data show that comRS is the origin of bimodality. There is no direct signaling cascade from CSP to ComRS. CSP induces bacteriocins. Its role in competence stimulation in complex medium is an indirect one, most likely as a result of bacteriocin synthesis causing permeabilization of the cells and reimport of endogenously produced and secreted XIP. The co-expression analysis described above however raised the question, why comE is coexpressed with comS, sigX and the late competence gene lytFsm. We hypothesize that transcription of comE is controlled by ComR or SigX. If SigX or ComR would control comE transcription, then the XIP signal should induce comE expression independent of ComDE. We tested this hypothesis using the comE reporter strain in different gene deletion backgrounds under XIP induced conditions in CDM. Table 2 (and microscopic pictures in S16 Fig) shows that this is indeed the case. The transcriptional activation of comE is independent of internal production and external sensing of CSP as it also occurs in ∆comDE and ∆comC deletion strains. Deletion of comRS and surprisingly also sigX completely abolished transcriptional activation of comE. In the ΔcomS background, transcription of comE is slightly reduced, since the internal production of XIP is missing and cannot be completely restored by externally added XIP. In THBY, comE transcription was similarly dependend on ComRS and SigX (S1 Text and S5 Table). These data provide strong evidence for a direct regulatory role of SigX for the expression of comE. Since the targets of SigX contain a cin box in their 5`UTR [38], we analysed the promoter region of comE. We searched for the cin-box, a known binding motif for SigX [38], in the 5´ UTR of comE. Indeed we found a sequence 109 bp upstream from the start codon of comE which has only one mismatch with the consensus sequence of the cin-box (TGCGAATA, Fig 6). The late competence gene lytFsm, which is controlled by SigX [32,39] and is highly expressed (our data) also showed one mismatch at the same position (TCCGAATA), thus we conclude that the cin-box homologue identified upstream of comE is most likely functional. We then looked for a promoter sequence. Directly adjacent to the cin-box a -10 region (Pribnow box) was found (TATATT). A corresponding -35 region was not identified. The transcriptome analysis allowed to identify the transcriptional start site of the comE gene. It was localized directly downstream of the cin-box, in complete agreement with the -10 localization of the Pribnow box. Sequence analysis and RNAseq thus strongly suggest that SigX can bind to the 5`UTR of comE and induce transcription. If XIP induces comE expression via SigX it should also induce bacteriocin expression. We analysed the transcription of the mutacin V encoding gene cipB in the cipB pMR1 reporter strain with different gene deletion backgrounds in CDM medium under XIP induced conditions. If SigX controls comE and thus cipB transcription, then fluorescence of the reporter should be induced by XIP, but not in a ΔcomRS or ΔsigX background, and independent of CSP signaling. Table 3 (and microscopic images in S13 Fig) shows that the expression of cipB required the presence of comE, comRS and sigX, and was weakened in the ΔcomD background. As the deletion of comE completely abolished cipB expression we can exclude an influence of other bacteriocin regulatory systems. The results on the CSP induced comE and cipB reporter strains in THBY are shown in S5 Table and S6 Table. Thus we conclude that SigX indeed induces transcription of bacteriocins in S. mutans via ComE. This is a novel role for SigX, therefore we asked if bacteriocins could be found in the culture supernatant. As a positive control, we analysed CSP induced bacteriocin production in THBY. CSP induction caused clear inhibition zones of the overlaid indicator strains L lactis and S. sanguinis, which are sensitive for mutacins V and IV, in the wild-type and the ΔcomC and ΔcomS mutant, respectively (S17 Fig). Deletion of comC could be complemented by externally added CSP, and deletion of comS had no effect as expected. We then tested bacteriocin production on CDM agar plates. No inhibition of the indicator strains could be observed, neither for CSP nor for XIP stimulated cultures. Bactericidal activity was also not detected in the concentrated supernatants of XIP induced and planktonically growing S. mutans strains in CDM. Since bacteriocins were clearly transcribed under these conditions, various mechanism could account for the lack of bacteriocidal activity: Bacteriocin synthesis could be regulated posttranscriptionally, they might not be secreted, or they could be degraded. Interestingly also CSP was either not produced, degraded or not secreted (S18 Fig and S1 Text). Our data show that bacteriocin expression in S. mutans is controlled by two different mechanisms: (1) A fast unimodal post-transcriptional activation of ComE by CSP independent from the growth medium. It is likely caused by phosphorylation of the response regulator upon CSP detection by the ComDE two-component system. (2) A delayed activation of comE transcription by XIP in CDM mediated by SigX, and thus showing the same modality as SigX. To investigate the timing and modality of bacteriocin expression, we analyzed CSP and XIP mediated signaling using time-resolved flow cytometric analysis of the reporter strains for comE and its immediate target, cipB. The density plots of uninduced comE and cipB reporter strains (controls) can be found S19 Fig. Fig 7A shows that comE transcription was unimodal in CDM using XIP as an inducer (green density plots), but bimodal in THBY under CSP inducing conditions (red density plots). Exactly the same behavior had previously been observed for sigX [15]. Apparently the modality of sigX expression determines whether comE expression is uni- or bimodal. The induction of comE expression by XIP in CDM starts faster (30–60 minutes post XIP addition) than the CSP induced transcriptional activation of comE expression in THBY which begins 90–120 minutes post induction. CSP causes medium independent unimodal expression of bacteriocins, as also shown above. We have also shown above, that addition of CSP to THBY allows XIP signaling, but not because of a direct regulatory link as previously hypothesized. Instead an indirect effect is operating. We suggest that bacteriocins make the cells permeable and thus allow reimport of XIP into the producing cell. Thus in our experiment here, bacteriocin synthesis is induced by CSP, likely resulting in permeabilzation of the cell and reimport of XIP that is endogeneously produced. XIP then activates sigX, resulting in transcription of comE via binding of the SigX-RNA-Polymerase complex to the cin-Box. This explains why comE expression is delayed in THBY by about 90 min in comparison to CDM. Fig 7B clearly shows that the bacteriocin encoding gene cipB was strongly and instantaneously induced upon CSP stimulation in CDM in an unimodal way (blue density plots). Already 30 minutes after induction, which corresponds to the maturing time of the GFP+ fluorophore, the expression of cipB was significantly enhanced. Addition of XIP in CDM medium (yellow density plots) also induced cipB expression unimodally but more weakly and significantly delayed. It starts 90 minutes after addition of XIP and reaches its maximum 180 minutes post induction. Independent from the growth medium, induction of cipB expression occurs within 30 min post CSP supplementation (S20 Fig). Thus, the data are consistent with the proposed transcriptional control of comE by SigX. To observe post-transcriptional and transcriptional regulation in more detail, we then compared the time-course of expression of comE with that of its direct target cipB under CSP induced conditions in CDM directly (Fig 7C). CipB expression was observed instantaneously already 30 minutes post CSP supplement (blue density plots), while comE expression was unchanged and started to increase only 180 min post CSP addition (orange density plots). Microscopic analysis of the comE and cipB reporter strains confirmed that comE is not induced within the first 120 min while its target cipB is already fully induced 30 min post induction by CSP in CDM (S14 Fig and S1 Text). Thus CSP-induced activation of the bacteriocins via ComDE must be mediated via post-transcriptional regulation. The basal level of ComE transcripts is apparently sufficient for this reaction. We then investigated the influence of the key genes in the signalling cascade on timing and modality of comE and cipB expression by analyzing the fluorescence of the comE and cipB reporter strains in different gene deletion backgrounds. Fig 8 shows that transcriptional activation of comE requires comRS, comS and sigX while the transcriptional activation of the ComE target cipB is independent of these genes. Moreover transcription of comE is bimodal while the transcription of its target cipB is unimodal. These findings confirm that indeed two different mechanisms (post-transcriptional and transcriptional regulation) are operating that activate ComE thus fully supporting our hypothesis. To determine if phosphorylation of ComE controls transcription of the bacteriocin encoding genes we constructed mutants of comE where the aspartate residue at position 60 was replaced by a phosphomimetic, namely by glutamate (D60E) or alanine (D60A). Glutamate enhances the negative charge of the side chain and thus mimics a phosphorylated aspartate. Conversely, alanine cannot be phosphorylated and thus ComE should be inactive under all circumstances. We utilized the CipB pMR1 reporter strain in the ΔcomE deletion background and transformed it with replicative pIB166 [37] based plasmids carrying either the native or a mutated comE gene under the control of the native comE promoter. As a control, we introduced the native comE gene into the ΔcomE reporter strain, which restored the CSP responsive phenotype. We also introduced a comE overexpression plasmid into the reporter strain and found CSP independent unimodal cipB expression. Reporter strains carrying the empty plasmid showed no detectable fluorescence during growth or CSP induction (Fig 9). As expected, strain CipB pMR1 ΔcomE D60A did not fluoresce either under CSP induced or uninduced conditions. Due to the D60A mutation ComE could not be phosphorylated and looses its ability to induce cipB expression. In the CipB pMR1 ΔcomE D60E strain, which mimicked stable phosphorylation, CipB induction was independent of the presence of CSP, in contrast to the native comE gene. The fluorescence of the D60E mutant was lower than that of the CSP-induced wild-type. Hung et al. found that ComE phosphorylation affected dimerization of S. mutans ComE in vitro and that the D60E mutation prevents that dimerization [40]. To conclude here we demonstrate that phosphorylation of ComE is the mechanism responsible for the induction of bacteriocin expression under CSP induced conditions.
Our analysis results in a new understanding of the QS regulatory network in S. mutans which is depicted in Fig 10. The two signals CSP and XIP control different traits: CSP signaling (green box) mediates bacteriocin expression while XIP signaling induces competence development (red box). The model contains three different temporal layers of signaling: (I) Fast post transcriptional activation of the response regulator ComE and the transcription factor ComR induced by their respective signals within the first 5 min after detection. ComE is activated by phosphorylation, and ComR is activated by binding of XIP. (II) Early transcriptional response of the activated regulators resulting in bacteriocin synthesis (CSP signaling) and transcription of the comRS genes and the alternative sigma-factor sigX (XIP signaling). (III) Late transcriptional response mediated by the alternative sigma-factor SigX. Transcription of the gene encoding the regulator of bacteriocin synthesis, comE, is controlled by SigX. In such a way both QS systems are connected. SigX controls both competence development and—via comE–bacteriocin synthesis. In the competence cascade comE is localized downstream of comRS and sigX and not upstream as previously thought. ComE could thus be viewed as a “late competence gene”, although it is only indirectly involved in competence development. This model explains why co-expression of comS, sigX, comE and the late competence genes is observed and why the modality of sigX expression determines the modality of comE. Our model discriminates between post-transcriptional and transcriptional regulation, which has not been taken into account by previous modeling approaches. Instead of two QS systems operating in parallel connected through a hypothetical regulator as assumed before (Fig 1), we have two QS systems operating independently, triggered by their respective signals CSP and XIP. The competence cascade, however, is linked to bacteriocin synthesis through the alternative sigma-factor SigX which controls comE. Thus, SigX induces competence, and with a time delay, also bacteriocin synthesis. Typically QS networks are organized hierarchically (e. g. in Pseudomonas aeruginosa) [41]. By contrast, in Vibrio harveyi three parallel input channels responding to three different autoinducer signals are integrated by a central response regulator [42]. The type of network architecture found here comprises two different input channels which are connected by the alternative sigma-factor SigX. SigX is stimulated directly by XIP and indirectly by CSP. This type of network architecture has not been found before. A hierarchical cascade mediates competence. XIP is the signal inducing competence, independent from the growth medium. However the medium determines via which way XIP is imported into the cell. We hypothesize that bacteriocin expression is required to induce permeability of the cell for external XIP. CSP signaling is independent from XIP signaling and is mediated via phosphorylation of ComDE. CSP-induced unimodal activation of bacteriocin synthesis is regulated post-transcriptionally and occurs before the transcription of comE which is mediated via XIP signaling. CSP induces competence only indirectly. It is therefore not a competence inducing peptide and should be renamed MIP (mutacin inducing peptide). CSP signaling is also in principle independent from the cultivation medium, but in defined medium bacteriocin excretion is inhibited. The new model simplifies our understanding of QS in S. mutans and resolves the open questions initially formulated. We will subsequently discuss them point by point. Lemme et al. [14] postulated comE being the origin of bimodality upon CSP induced competence development. Here we clearly show that this not the case since strong overexpression of comE independent of its native promoter did not enhance the percentage of competent cells. For comS a positive feedback loop was identified, which is a prerequisite for bimodality to occur [7] and thus a model assuming bistable expression of comS was developed [15]. However, here we show that neither comR nor comS alone can induce unimodal competence development when they are strongly overexpressed. Both genes or rather their gene products are required. Bistable expression of comS is not necessarily required. How then is the bimodality established in the system? Besides a positive feedback loop, noise is the second critical determinant for bimodality [29,43]. We suggest that noise in comR and comS expression accounts for bimodality since the overexpression analysis showed that both are limiting factors for unimodal competence development. Bimodality would then be a feature of the quorum sensing network, and not necessarily the result of bistable expression of one component [44]. XIP is a secreted factor, and thus it is unclear how phenotypic heterogeneity can be stably established if all cells are able to sense XIP, not only the overproducers. Previously it was found that the mature XIP signal of S. thermophilus is not released into the medium, but remains cell associated [13]. Thus cells with higher comS expression levels can re-import more XIP resulting in the observed positive feedback loop. It was previously suggested that an unknown regulator transfers the CSP induced signal from ComDE to ComRS. Here we show that this regulator does not exist. If a regulator would be required for transmitting the CSP induced signal to ComRS, then it would be possible to overcome its effect by overexpression of comRS, which should result in unimodal competence development. However, here we show that overexpression of comRS does not induce competence at all in complex media. We suggest that CSP induces competence only indirectly by allowing XIP to be imported into the cell in complex media. The following findings support this hypothesis: 1. Competence is not induced in CDM by CSP, although externally added CSP is fully active and not degraded as previously thought. 2. A small quantity of CSP is required to induce unimodal competence development in THBY in a strain overexpressing comRS. 3. An increase in the concentration of CSP does not increase the percentage of competent cells in this experiment, suggesting that CSP triggers a process required for competence to occur and is not inducing it directly. 4. The time-lag of 60–90 min between CSP addition and sigX transcription in THBY clearly supports an indirect activation mechanism. 5. Deletion of CipB completely abolished CSP and CSP/XIP induced expression of the late competence gene lytFsm. Our findings are in accordance with a recent study showing that deletion of the bacteriocin encoding gene cipB almost completely abolishes competence development through CSP stimulation [34]. Thus it was suggested that the CipB bacteriocin itself might act as a regulator of competence [39]. This scenario is highly unlikely, because a regulatory role for a bacteriocin has never been demonstrated. Moreover, bacteriocins are secreted into the environment and thus can hardly exert intracellular regulatory roles. Based on our findings we suggest that bacteriocins are required to induce permeability of the producer cell for external XIP. It was shown that deletion of Opp has no influence on CSP induced competence development in complex medium in S. mutans [15]. Thus XIP import upon bacteriocin expression in complex medium is not mediated via Opp and its precise mechanism remains to be elucidated. Pore formation is the mode of action of many bacteriocins [45]. Via these pores the small XIP molecule might enter the cell. Alternatively some of the bacteriocin exporters might allow XIP import. Cell wall and membrane alterations which confer immunity to the bacteriocin producer cells against their own bacteriocin [45] may also account for the altered permeability of the cell for XIP. Presently it is not known how and by which system the XIP precursor is secreted and processed in S. mutans [7,46,47]. Bacteriocin synthesis is induced by post-transcriptional unimodal activation of the ComE response regulator. The CSP signal causes phosphorylation of the ComE protein, resulting in instantaneous unimodal bacteriocin transcription and synthesis. This process is not affected by transport processes across the membrane and therefore it is independent of the cultivation medium. By contrast, the transcription of the comE gene is controlled by SigX, which is regulated via the XIP signaling cascade. Therefore bimodal comE expression occurs much later (90–120 post induction). It was thought that external CSP is degraded by the HtrA protease in CDM, while this protease is inhibited by media components in THBY and thus allows CSP to remain intact [15]. Here we show that external CSP is not degraded in CDM. For the HtrA serine protease of S. pneumoniae it was demonstrated that it indeed cleaves the pneumococcal CSP. The enzyme was not inhibited by bovine serum albumin, but by denatured protein, prompting the authors to suggest that CSP signaling is a means to detect intracellular stress [22]. The CSP of S. mutans is more homologous to the BlpC peptide which induces bacteriocin synthesis than to the CSP autoinducer of S. pneumoniae. Extracellular BlpC is not degraded by HtrA [48] in accordance with the lack of degradation of S. mutans CSP in CDM. These data, too, demonstrate that CSP of S. mutans is functionally related to BlpC of S. pneumoniae. Activated ComR represents the only regulator mediating competence in S. mutans. In a recent review it was proposed that the ComDE two-component system (TCS) of S. mutans should be renamed to Blp since it primarily regulates bacteriocin expression [1]. However, this suggestion was based on the situation in S. pneumoniae, where the BlpRH TCS is exclusively involved in regulating bacteriocin production [48]. In S. mutans, comE is controlled by the master regulator of competence SigX. Thus we would not propose to rename the system Blp. However, since CSP does not stimulate competence directly, and in S. mutans comE is not an early competence gene as the name suggests, we propose to rename CSP to MIP (mutacin inducing peptide) and ComDE to MutDE. Upon XIP induction SigX enhances the ComE level in the cell. To be active, ComE needs to be phosphorylated, yet the histidine kinase ComD is not receiving a signal under these conditions and its deletion still allows cipB expression [34]. Therefore non-cognate kinases must be able to phosphorylate ComE. Recently cross-talk between TCS systems was observed [49]. For the serine/threonine protein kinase PknB of S. mutans a regulatory role in bacteriocin expression was found [50], suggesting that this enzyme might also be able to phosphorylate ComE. Although the transcription of bacteriocin encoding genes was strongly induced by externally added CSP or XIP in CDM, no bactericidal activity was found in the culture supernatants, suggesting that bacteriocin synthesis is regulated post-transcriptionally. This is supported by a study of two phenotypic variants (transparent/opaque) of the S. pneumoniae R6 strain [48]. They showed an identical transcriptional response to the bacteriocin inducer BlpC. However only for cells of the transparent phenotype bacteriocin activity could be detected, while none was found in the opaque phenotype. The authors demonstrated that the HtrA protease abolished bacteriocin synthesis post transcriptionally in these cells [48]. This might be the reason why we detected no CSP and no bactericidal activity in the CDM supernatants. HtrA activity likely prevents CSP and bacteriocin processing and/or secretion in CDM. Multiple environmental factors additionally influence bacteriocin expression in S. mutans, including cell density and nutrient availability [51]. The observation that comE transcription is under the control of SigX represents a novel regulatory role for SigX, which thus not only controls the expression of the transformasome, but also the synthesis of bacteriocins. In S. pneumoniae and B. subtilis the master regulators of competence, SigX and ComK respectively, exclusively regulate transformasome related genes [1,43]. Competence and bacteriocin production is uncoupled in these organisms. Coupling it like in S. mutans makes ecological sense, because it provides the genetic variability which makes competence an adaptive trait, as suggested previously [52]. In its natural niche S. mutans is part of the multispecies oral biofilm which consists of over 600 different species [53]. Thus S. mutans is faced with a strong competitive environment which is at the same time a rich source of genetic variability. In a dual species biofilm of S. mutans and Candida albicans the CSP and XIP triggered QS cascades were induced simultaneously [54]. The ComR regulator has a low stringency with respect to the exact sequence of the XIP peptide required for activation [12], and consequently, Streptococci respond to heterologous peptides [12,55]. Thus S. mutans might sense the presence of other Streptococci via XIP signaling and directly respond to this stimulus by producing bacteriocins and killing the competitor, while exploiting its genetic information at the same time. To conclude, here we deciphered the complex QS signaling network of S. mutans on the single cell level. Competence is exclusively mediated via XIP signaling in a hierarchical network structure. We show that competence development is coupled to bacteriocin synthesis through the alternative sigma-factor SigX, which makes this QS network highly efficient for acquiring new genetic information in the competitive environment of dental biofilms.
All S. mutans strains were routinely propagated in in Todd Hewitt broth medium supplemented with 0. 5% (wt/vol.) yeast extract (THBY; Becton Dickinson, Heidelberg, Germany) in an incubator (5% CO2,37°C) without agitation. When indicated, antibiotics were added to the medium (chloramphenicol 10 μg/ml, tetracycline 12. 5 μg/ml and erythromycin 10μg/ml). For experiments conducted in chemically defined medium (CDM) [56] a 10 ml overnight culture of the appropriate strain grown in THBY was centrifuged for 10 min at 5000 rpm and 4°C. The supernatant was removed and the cell pellet was gently resuspended in fresh CDM medium and again centrifuged as described above. After removal of the supernatant the pellet was finally gently resuspended in 5 ml of CDM. Strains were diluted to an OD of 0. 1 and grown at 37°C and 5% CO2. The plasmids used for the construction of reporter strains (S1 Fig), an analysis of maturation times of various fluorescent proteins in S. mutans (S2 Fig), a scheme showing the integration of the plasmids into the chromosome (S3 Fig), comparison of expression levels of different late competence genes (S4 Fig), comparison of two LytFsm constructs with different fluorophores (S5B Fig), spectral separation of TagBFP2 and GFP+ (S6 Fig), dye swap of comE with the late competence gene lytFsm (S9 Fig), and construction of reporters in various gene deletion backgrounds are described in full detail in S1 Text, which also provides a discussion of the reporter constructs. A list of the constructed strains (S1 Table) plasmids (S2 Table) and primers (S3 Table) can be found in the supplements. Overnight cultures of the strains grown in THBY were diluted to an OD of 0. 1. For the strains intended to grow in CDM the above described washing procedure was applied to remove all traces of the complex media and finally strains were diluted to an OD of 0. 1 and grown at 37°C and 5% CO2. When the bacteria had reached an OD of 0. 15 the culture was divided into three equal fractions. One fraction was treated with 2 μM synthetic CSP, the second with 2 μM synthetic XIP, and the third fraction was used as an uninduced control. Samples (0. 5 ml) were taken after 30,60,90,120 and 180 minutes post induction. Samples derived from strains growing in either CDM or THBY were centrifuged (5 min and 7000 rpm) and washed once with PBS. Subsequently the strains were resuspended in 1 ml of ice-cold PBS and sonicated using a MS72 sonotrode with the Sonoplus HD2200 device (Bandelin, Germany) for at least 20 sec at 10% power. Settings were a 0. 5 sec impulse which was followed by a 0. 5 sec break. Live/Dead staining before and after sonication was performed to exclude that sonication significantly interfered with membrane integrity. For flow cytometry the LSR Fortessa Cell analyser (BD, Germany) was used. 0. 22 μM filtered PBS was applied as sheath fluid. Cytometer settings were chosen as previously reported [14]. 50000 cells were analysed and the resulting data processed with a self-written R-Script. The mutation of the comE gene was accomplished using a PCR-driven overlap extension approach [57]. The triplet encoding the aspartate residue at position 60 (GAT) was changed to GCT (Ala) and GAG (Glu), respectively, using the primers listed in S3 Table. In a first PCR the comE gene and its native promoter region were amplified in two separate parts thereby introducing the desired mutation via the 5`termini of the inner primers (primer pairs PcomE_F1/D60E_R1 and PcomE/D60A_R1 for the first part and primer pairs D60E_F2/ComE_R2 and D60A_F2/ComE_R2 for the second part of the comE gene. The two PCR amplified parts of comE contained homologous flanks of approximately 22 bp to each other and were used as template for a second PCR with only the outer primers (PcomE_F1/ComE_R2), amplifying the entire comE coding sequence including the native promoter. The PCR products were purified and cloned into the vector pIB166 [37] in opposite direction to the constitutive P23 promoter to allow transcription from the native promoter. The plasmid sequence was verified by sequencing and the plasmid transformed into the S. mutans cipB reporter strain in the ΔcomE deletion background. As positive control the native comE sequence was cloned into pIB166 and the transformed empty vector was used as negative control. To construct overexpression strains for ComD, ComE, ComR, ComS and ComR+ComS, independent from their native promoter, plasmid pIB166 containing the strong constitutive P23 promoter from Lactococcus sp. was used [37]. The coding sequences including the native ribosomal binding sites were PCR-amplified (see S3 Table) and cloned blunt end via the SmaI restriction site into the vector. The correct plasmid sequence was verified by sequencing and 100 ng of the plasmid transformed into the LytFsm reporter strain. While this cloning approach was successful for comE, comS, comRS and comR, we were not able to obtain a correct plasmid sequence for comD. Point mutations and frameshift mutations were observed in the ribosomal binding site or the coding region of the gene and strongly suggest that this protein is toxic for E. coli. As it was already successfully cloned in E. coli under the control of the native promoter we assume that the P23 promoter has a significantly higher basal transcription than the native comD promoter in E. coli. Attempts to fuse the GtfB or GtfC promoter to the coding sequence of comD were also not successful, although these constructs could be easily obtained for comE and comR, respectively. We therefore used a PCR-driven overlap extension approach ([57]) to amplify a fusion construct consisting of five parts; the upstream region of gene SMU_1342, the P23 promoter, the comD gene, the chloramphenicol resistance cassette (CAT) and downstream flanking regions of SMU_1342. Homology flanks of the different parts were introduced via the 5`terminus of the PCR primers. This construct allows the integration of the comD gene under the control of the P23 promoter via double homologous recombination at the SMU_1342 locus. In a first PCR all parts of the fusion construct were amplified separately and purified using the PCR Purification Kit (Qiagen, Germany). Subsequently equal amounts (100 ng) of all different PCR products were used as template for a second PCR with primers UP1342_F/D1342_R spanning the entire construct (outer primers). The PCR reaction was directly used as template to transform the LytFsm reporter strain. Chloramphenicol resistant clones were selected on THBY agar plates, picked and cultivated in liquid THBY medium containing 10 μg/ml chloramphenicol. Genomic DNA was isolated from 2 ml of the exponentially growing culture and correct constructs were verified by PCR using the primer pairs UP1342_F/D1342_R and P23_F/ CAT_R). A detailed description of methods for RNA extraction, RNA sequencing and qRT-PCR can be found in the S1 Text. Raw and processed RNAseq data have been deposited in the gene expression omnibus database (http: //www. ncbi. nlm. nih. gov/geo/) under accession number GSE65982. O/N cultures of S. mutans WT, ΔcomS and ΔcomC deletion strains with a starting OD600 of 0. 1 were grown in CDM or THBY until the cultures reached an OD600 = 0. 25. Then the cultures were split and induced with either 2 μM XIP or 2 μM CSP. Uninduced cultures were used as controls. After 2 hours of further growth (37°C, 5% CO2) 2. 5 μl of the culture were spotted on CDM or THBY agar plates. The plates were incubated for further 4 h until the induced cultures were overlaid with a 2. 5 μl drop of a 2 μM CSP (or 2 μM XIP) solution in CDM (THBY). The drops were allowed to soak in the plate. Finally the plate was incubated for 24 h at 37°C and 5% CO2. Overnight cultures of the indicator strains S. sanguis and L. lactis were diluted 1: 200 in fresh CDM Top Agar (0. 7% Agar) placed at 37°C and 5 ml of the mixture was poured over the agar plates containing the spotted S. mutans producer strains. The overlaid plate was incubated for further 16 h in a CO2 incubator. Zones of inhibition were documented using a conventional camera. | Streptococcus mutans is a bacterium of the human dental plaque that contributes to caries development. It controls two important survival mechanisms via a cell-density dependent communication system (quorum sensing): The synthesis of peptide antibiotics, and of a membrane apparatus for genetic competence, i. e. the ability to take up external DNA and integrate it into its own genome. S. mutans synthesizes two different signalling peptides to this end. It has remained elusive, how exactly these signals are propagated within the cell and why only a fraction of the population becomes competent. To actually observe under the microscope which bacterium in the population is activated, and which genes are required for the activation, we constructed strains of S. mutans that reported on the transcription of a gene by starting to fluoresce green. We even constructed strains that reported on two genes simultaneously, by fluorescing either green or blue or both. With these tools, and by additionally knocking out or modifying key genes as needed, we investigated the complete signaling cascade under various conditions. Thus we discovered a central regulatory switch. S. mutans makes sure that external DNA is available when it becomes genetically competent–by killing cells in the environment. | Abstract
Introduction
Results
Discussion
Methods | 2015 | The Alternative Sigma Factor SigX Controls Bacteriocin Synthesis and Competence, the Two Quorum Sensing Regulated Traits in Streptococcus mutans | 14,873 | 287 |
|
Access to laboratory diagnosis can be a challenge for individuals suspected of Buruli Ulcer (BU). Our objective was to develop a clinical score to assist clinicians working in resource-limited settings for BU diagnosis. Between 2011 and 2013, individuals presenting at Akonolinga District Hospital, Cameroon, were enrolled consecutively. Clinical data were collected prospectively. Based on a latent class model using laboratory test results (ZN, PCR, culture), patients were categorized into high, or low BU likelihood. Variables associated with a high BU likelihood in a multivariate logistic model were included in the Buruli score. Score cut-offs were chosen based on calculated predictive values. Of 325 patients with an ulcerative lesion, 51 (15. 7%) had a high BU likelihood. The variables identified for the Buruli score were: characteristic smell (+3 points), yellow color (+2), female gender (+2), undermining (+1), green color (+1), lesion hyposensitivity (+1), pain at rest (-1), size >5cm (-1), locoregional adenopathy (-2), age above 20 up to 40 years (-3), or above 40 (-5). This score had AUC of 0. 86 (95%CI 0. 82–0. 89), indicating good discrimination between infected and non-infected individuals. The cut-off to reasonably exclude BU was set at scores <0 (NPV 96. 5%; 95%CI 93. 0–98. 6). The treatment threshold was set at a cut-off ≥4 (PPV 69. 0%; 95%CI 49. 2–84. 7). Patients with intermediate BU probability needed to be tested by PCR. We developed a decisional algorithm based on a clinical score assessing BU probability. The Buruli score still requires further validation before it can be recommended for wide use.
Buruli ulcer (BU) is a skin infection due to Mycobaterium ulcerans. Classical cases with painless ulcerated plaques and undermined edges are believed to be relatively easy to diagnose clinically in endemic regions [1]. Nevertheless, without laboratory testing to confirm the findings, diagnostic errors are probably highly under-estimated. The most common laboratory technique is the direct examination by microscopy of dry skin swabs stained with Ziehl-Neelsen (ZN) in search of alcohol-resistant bacilli [2]. PCR for insertion sequence IS2404 is very specific and currently the most sensitive diagnostic method [3]. However, it is rarely available on site and samples need to be sent to a reference laboratory and results be sent back to the clinician, resulting in substantial treatment delays and loss-to-follow-up. Cost of the method can be a barrier for the patient outside of projects supported by NGOs or research institutions. Because of the challenges of PCR-based diagnosis of BU, alternative methods should be explored. While rapid point-of-care tests are currently being developed [4], identifying which patients would truly benefit from a diagnostic test can reduce patient expenses and treatment delays for patients with a high probability of BU. Diagnostic scores derived from multivariable prediction models can be useful in clinical decision-making [5]. These scores are based on diagnostic performance of various characteristics of medical history and clinical examination. For BU diagnosis, it has been proposed to combine ZN microscopy and PCR, only performing PCR in ZN-negative specimen, in order to reduce costs and keep diagnosis available at peripheral level for many patients [6]. Therefore, we aimed to identify clinical predictors of BU diagnosis, in order to develop a multivariable prediction model (“Buruli score”), to assist clinicians working in resource-limited settings.
Ethical approval was given by the National Ethics Committee of Cameroon, the Central Commission on Human Subject Research Ethics of the Geneva University Hospital, and the Ethics Committee of Médecins Sans Frontières. The study was further authorized by the Ministry of Health, in the framework of the National Buruli Control Program, as well as from the health authorities of the Akonolinga District and Akonolinga Hospital administration. All patients provided written informed consent.
Between October 2011 and December 2013,367 patients were included in the study, out of 447 screened, and 364 were finally analyzed (3 secondary exclusions due to missing clinical data), corresponding to 422 lesions, of which 381 were ulcerative (90. 3%). Detailed patient flow is presented elsewhere [11]. There were more inclusions during the first half of the study period compared to the second half (215 vs. 110). Because ulcerative and non-ulcerative lesions had different clinical characteristics, the prediction model was based on the 325 patients with 379 ulcerative lesions and available laboratory results (missing for two patients). Median age was 37 years (range 0 to 87), with 28. 9% aged up to 20 years, 26. 5% from 20 to 40 years, and 44. 6% above 40 years (Table 1). Overall 212 (65%) were males and 63 (19. 4%) were HIV-positive, with a median CD4 count of 362 (IQR 210–653; 12 missing CD4 count). In terms of other comorbidities, hypertension was confirmed in 4 cases (1. 2%) and suspected in another 9 (2. 8%); diabetes was confirmed and suspected in 7 (2. 2%) and 22 (6. 8%) cases, respectively. Sickle cell disease was confirmed in 6 (1. 8%) patients. By severity grading according to WHO classification for Buruli ulcer, patients were of category I, II or III in 41. 5%, 30. 5% and 28. 0%, respectively. Demographic and clinical characteristics are shown in Table 1. Proportion of positive laboratory tests for ulcerative lesions (N = 379) varied between 7. 9% for culture and 22. 2% for PCR. ZN was positive for 17. 4% in Akonolinga and 10. 6% in CPC. The estimated BU prevalence from the latent class model was 16. 1% (95%CI 12. 4–20. 7%). Estimated sensitivity went from 46% for culture to 100% for PCR, with intermediate results for ZN (65% and 72% in Yaounde and Akonolinga, respectively; Table 2). Specificity was best for ZN Yaounde (100%) and culture (99%), followed by PCR and ZN Akonolinga (93%). Patients could be clearly discriminated into two groups according to their pattern of laboratory test results: a high BU likelihood (probability >0. 8) and a low BU likelihood (probability <0. 15) Fig 1). In the univariate analysis, the following patient variables were found associated with BU likelihood at a p-value <0. 20: duration of the episode, topical or systemic treatment (oral or parenteral) received previously, history of trauma, type of oedema, age, and gender (Table 1). Variables at lesion level associated with BU likelihood were localization, lesion size, hyposensitivity, induration, locoregional adenopathy, pain at rest, undermining, characteristic smell, green (purulent), yellow (fibrinous), and red (tissue granulation) color. The complete results of the univariate analysis are detailed in the S1 Table. None of the comorbidities (HIV, diabetes, hypertension) was found to be associated with BU. After adjustment with the other variables in the model, the following variables were included in the Buruli score based on an odds ratio >1. 5 or <0. 67 (Table 3): characteristic smell (+3 points), yellow color (+2), female gender (+2), undermining (+1), green color (+1), lesion hyposensitivity (+1), pain at rest (-1), lesion size above 5cm (-1), locoregional adenopathy (-2), age above 20 up to 40 years (-3), and age above 40 years (-5). The Buruli score had an area under the ROC curve (AUC) of 0. 86 (95%CI 0. 82–0. 89) using the outcome of the latent class model as reference, similar to the AUC of the full multivariate model (Fig 2). The cut-off to reasonably exclude BU was set at scores < 0 (NPV 96. 5% 95%CI 93. 0–98. 6). The treatment threshold was set at a cut-off ≥4 points (PPV 69. 0,95%CI 49. 2–84. 7). Patients with scores between 0 and 3 had an intermediate probability of BU and would need to be tested further by PCR. Using the algorithm on the patients included in the study (Fig 3), 56 patients would have been treated for BU, including 12 “false-positives” using the outcome of the latent class model as reference (specificity 95. 5%; 95%CI 92. 3–97. 7). Seven BU cases would have been missed by the algorithm (sensitivity 86. 3%; 95%CI 73. 7–94. 3). Three of them were HIV-positive, one had diabetes and another diabetes suspicion. Overall, PCR would have been performed in 27. 7% (90/320) of the patients. Hosmer-Lemeshow goodness-of-fit showed that our model predicted well the observed data (p = 0. 24). Pregibon’s leverage and delta beta plots did not show major influential observations on model fitting in the dataset. The internal bootstrapping validated c-statistic was 0. 76 (from 0. 86 initially). Excluding variables with smaller coefficients from the score (green color, undermining, hyposensitivity, pain at rest, and lesion size >5cm) tended to decrease the area under the curve (0. 85,95%CI 0. 80–0. 88) and affected patient classification. Study period and mode of recruitment did not affect score performance
We developed a decisional algorithm based on a clinical score to assess the probability of BU infection among suspects. Applying the algorithm to the patients included in the study would have resulted in almost four times less PCR performed. After this first study on calibration, the Buruli score requires external validation before it can be recommended for wide use. | In most Buruli ulcer (BU) endemic areas, laboratory diagnosis is hard to access and comes at a high cost. Clinicians are in need of new tools to assist them in identifying which patients truly require additional work-up and which can be treated directly. We analyzed the clinical data of all patients with ulcerative skin lesions that presented to Akonolinga District Hospital in Cameroon and identified which parameters were associated with BU diagnosis. We attributed a certain number of points to each parameter to build a “Buruli score”. Based on score results, clinicians can be advised either to directly treat BU (score ≥4), to look for another diagnosis (score <0) or to do a PCR test (score between 0 and 3). This algorithm was found to have a good performance. Only one out of four patients still needed an additional laboratory test to be classified between BU and non-BU. However, this score still requires validation in another context before it can be recommended elsewhere. | Abstract
Introduction
Methods
Results
Discussion | smell
medicine and health sciences
pathology and laboratory medicine
tropical diseases
geographical locations
social sciences
neuroscience
bacterial diseases
research design
mathematics
signs and symptoms
forecasting
statistics (mathematics)
neglected tropical diseases
molecular biology techniques
africa
research and analysis methods
infectious diseases
buruli ulcer
cameroon
artificial gene amplification and extension
mathematical and statistical techniques
lesions
molecular biology
laboratory tests
people and places
psychology
diagnostic medicine
biology and life sciences
sensory perception
physical sciences
statistical methods
polymerase chain reaction | 2016 | The “Buruli Score”: Development of a Multivariable Prediction Model for Diagnosis of Mycobacterium ulcerans Infection in Individuals with Ulcerative Skin Lesions, Akonolinga, Cameroon | 2,334 | 224 |
The relationship of HIV tropism with disease progression and the recent development of CCR5-blocking drugs underscore the importance of monitoring virus coreceptor usage. As an alternative to costly phenotypic assays, computational methods aim at predicting virus tropism based on the sequence and structure of the V3 loop of the virus gp120 protein. Here we present a numerical descriptor of the V3 loop encoding its physicochemical and structural properties. The descriptor allows for structure-based prediction of HIV tropism and identification of properties of the V3 loop that are crucial for coreceptor usage. Use of the proposed descriptor for prediction results in a statistically significant improvement over the prediction based solely on V3 sequence with 3 percentage points improvement in AUC and 7 percentage points in sensitivity at the specificity of the 11/25 rule (95%). We additionally assessed the predictive power of the new method on clinically derived ‘bulk’ sequence data and obtained a statistically significant improvement in AUC of 3 percentage points over sequence-based prediction. Furthermore, we demonstrated the capacity of our method to predict therapy outcome by applying it to 53 samples from patients undergoing Maraviroc therapy. The analysis of structural features of the loop informative of tropism indicates the importance of two loop regions and their physicochemical properties. The regions are located on opposite strands of the loop stem and the respective features are predominantly charge-, hydrophobicity- and structure-related. These regions are in close proximity in the bound conformation of the loop potentially forming a site determinant for the coreceptor binding. The method is available via server under http: //structure. bioinf. mpi-inf. mpg. de/.
The entry of the human immunodeficiency virus (HIV) into human cells is initiated by binding of the viral envelope glycoprotein gp120 to the cellular CD4 receptor [1], [2]. This primary interaction induces conformational changes in gp120 [3] that enable viral binding to one of the cell-surface coreceptors CCR5 or CXCR4 [4]. The interaction of gp120 with the coreceptor induces a series of further rearrangements in the envelope glycoproteins that trigger fusion of the virus and cell membranes [1]. The third variable region (V3) of gp120 [5], [6] plays a crucial role in biding to the coreceptor. Whether a virus can bind to CCR5 only (R5 virus), or is capable of binding to CXCR4 (X4 virus) is determined predominantly by the sequence and structure of this region [7]. The phenotype of viral coreceptor usage is termed viral tropism. It has been shown that in the early, asymptomatic stages of infection mainly R5 viruses are observed, whereas progression towards AIDS is often associated with the emergence of X4 viruses [8]. The finding that humans who lack CCR5 expression due to the homozygosity of the Δ32 mutation in the CCR5 gene are resistant to HIV-1 infection [9] stimulated research on CCR5 inhibitors which led to the licensing of Maraviroc (MVC) [10] for clinical use in 2007. Viral tropism is an indicator of disease progression and determining viral tropism is a companion diagnostic obligatory for the application of CCR5 inhibitors. Therefore there is a need for efficient methods for monitoring of coreceptor usage and for a better understanding of its determinants. Computational methods for predicting viral tropism based on the sequence of the V3 loop have been developed [11], [12], [13], [14] as an alternative to costly phenotypic assays for testing of the coreceptor usage [15]. The 11/25 rule was proposed as an initial approach for inferring coreceptor usage, and is based on the observation that a positive charge on either of the 11th or 25th residues in the V3 region is indicative of an X4 virus [5], [6]. Due to its simplicity, the 11/25 rule has been commonly used although it has been shown that for many viral variants, changes at positions 11 or 25 are neither necessary nor sufficient for the tropism switch [11]. More elaborate sequence-based methods for prediction of coreceptor usage rely on a binary encoding of amino acids in the V3 sequence and use statistical approaches to construct predictive models and to infer residues strongly related to the tropism [11], [12], [13], [14]. The geno2pheno[coreceptor] method developed by our lab has been made freely available on the internet and is widely used throughout Europe and beyond for interpreting genotypic data measured as a companion diagnostic to Maraviroc therapy. The method has entered the German/Austrian expert guidelines for HIV-1 tropism testing in 2009 [16] and the respective European guidelines in 2011 [17]. The major drawback of the binary sequence representation is that it only indirectly encodes the physicochemical properties of amino acids and their spatial arrangement in the binding site which ultimately determine viral tropism. Structures of gp120 including the V3 loop have been determined by x-ray crystallography [18], [19]. The V3 loop is an extended structure protruding approximately 30 Å from the CD4-bound core of gp120 [18]. It is composed of a conserved base, a flexible stem that rigidifies upon coreceptor binding and a tip in a β-hairpin conformation. After the first structure of the V3 loop has been resolved [18], new coreceptor prediction methods were developed [20], [21] incorporating structural information on the loop in the prediction process. Sander et al. [21] proposed a distance-based descriptor of the spatial arrangement of physicochemical properties of the loop. They found that the distance information resulting from structural modeling of the side chains of the loop together with a binary encoding of its sequence outperforms prediction methods based on sequence alone. Dybowski et al. [20] developed a two-level classification approach that combines two physicochemical properties of the loop – electrostatic potential and hydrophobicity. This two-level approach resulted in improvement in prediction accuracy over prediction based on sequence alone. Even though including the structural information into the prediction represents a step forward in understanding of the binding mechanism of gp120 to the coreceptor, both methods have limitations. The method by Sander et al. is based on molecular distances that do not offer a direct interpretation of the structural determinants of the phenotype. Dybowski et al. include only two features in their predictor while a systematic analysis of a larger set of physicochemical features of the V3 loop would allow for identifying other features relevant for tropism. Both methods involve costly computational operations such as calculation of the electrostatic potential or modeling of side chains that stand in the way of making the methods available as an online application. Finally, all previously proposed methods except one [14] were developed and tested exclusively on clonal data. Such data are inferred from lab-cloned viruses as opposed to clinically derived data, which are obtained through bulk Sanger sequencing of patient samples and contain viral mixtures. In bulk sequencing data, diversity of virus populations in a patient is represented by a consensus sequence comprising dominant strains. The exact composition of the virus population as well as viral minorities below 10% [22] of the population are not detected by bulk Sanger sequencing which has been shown to pose additional challenge for in silico coreceptor prediction [14]. The work presented here was motivated by the goal of developing a method for genotypic prediction of viral tropism that is at least as accurate as existing structure-based methods, i. e. , more accurate than the widely used sequence-based method [14]. At the same time, the method should allow for a computationally efficient implementation allowing for its general use as an online application. To meet this goal we present a systematic approach to incorporating physicochemical and structural properties of the V3 loop into the prediction of HIV coreceptor usage. We map 54 amino acid indices representing the physicochemical properties of amino acids onto the V3 loop structure and use methods from statistical learning to extract those features that are most informative of coreceptor usage. The extracted set of features represents a small fraction of the initial feature set and models based on this set attain higher prediction accuracy with decreased computational load. Our structural descriptor affords direct interpretation of the features of the V3 loop relevant for viral tropism by pointing to specific physicochemical properties of amino acids in different parts of the loop being predictive of coreceptor usage. We also applied our method to clinically derived (bulk) data and tested its usability for prediction of the MVC therapy outcome.
The structural descriptor of the V3 loop is based on the published structure of the V3 loop [18] and amino acid indices [23] representing physicochemical properties of amino acids in a numerical form. Each residue of the V3 loop sequence is represented by a vector comprising the 56 preselected indices [24]. The residue positions were mapped to spheres centered along the V3 loop backbone (Figure 1). The spheres represent structural proximities along the loop as well as uncertainty in the structural conformation of individual loop variants. The vectors of amino acid indices of the mapped residues were normalized using Gaussian smoothing and summed up within each sphere. Next, the sphere vectors were concatenated into a single V3 loop vector. This vector was used as the V3 loop structural descriptor in the statistical model (full model) for coreceptor usage prediction. A training dataset of 1186 phenotyped V3 sequences from the Los Alamos database [25] (clonal dataset) was used for model development. We investigated the average number of residues covered by each sphere and selected the radius of 8 Å based on predefined criteria (see Text S1, Figure S1). We tested several other radii for their predictive performance (Figure S2). Throughout this study we used area-under-receiver-operating-characteristic (ROC) curve (AUC) and, in line with the common approach to validating genotypic predictions of viral tropism, sensitivity at the specificity of 11/25 rule in a given dataset (ranging between 0. 89 and 0. 97 termed here sensitivity for brevity) calculated in a 10×10-fold cross validation as cutoff-independent measures of the prediction accuracy. The radius of 8 Å yielded the AUC of 0. 847 and a sensitivity of 0. 587. Smaller or larger radii led to significant reduction of prediction performance (p<0. 001 for R = 3 Å and R = 15 Å, paired Wilcoxon test). The performances of models with different parameter values are shown in Figure S2. For comparison we implemented two sequence-based descriptors of the V3 loop. The g2p model represents each amino acid as a binary vector of size 20 in which the position of a single 1 indicates the amino acid it encodes. This representation is used by sequence-based approaches, among others by geno2pheno[coreceptor] [14]. Those approaches are not necessarily based on the same training sets as the one used in our study. Certainly the training set used by the geno2pheno[coreceptor] method is different. The aaindex model encodes each amino acid as a vector of the 56 amino acid indices used in the structural descriptor. In order to reduce the highly redundant feature vector of the structural descriptor of the full model and to investigate which features are informative for coreceptor usage we applied several feature selection procedures: Random Forest (RF) [26], linear support vector machine (SVM) [27] and Lasso regression [28]. We next compared the performance of the full model based on the entire set of features to the descriptor based on separate and combined subsets of features selected by three feature selection methods. SVM was the prediction method used throughout this study independent of the feature set and sequence encoding. Overall, reducing the feature set resulted in improved prediction accuracy over the full set of features and over the g2p model (Table 1). The SVM (1) model based on the top 1% ranking features performed better than the SVM (5) model based on a larger feature set of the top 5% ranked features. The Lasso model based on the most strongly reduced feature set (102 features) selected via Lasso regression resulted in the highest performance with AUC 0. 893 and sensitivity 0. 674. Models based on features selected via RF ranking showed the poorest predictive performance of all models tested (Table 1). The sets of features selected by the three feature selection methods show a limited overlap. The initial feature set contains small subsets of highly correlated features that pertain to highly correlated amino acid properties in overlapping structure regions (spheres). These features convey the same information to the prediction method and can be therefore selected interchangeably by each method (see Text S2 and Figure S3). However, the overall correlation of features in the descriptor is low and a version of the descriptor based on an uncorrelated feature set does not yield improved performance (Text S3, Figure S4). We performed the same feature selection procedures on models based on the structural descriptor with a sphere radius of 10 Å, chosen based on analysis described in Text S1 and Figure S1. The results of models based on this radius showed similar patterns of performance although with lower prediction performance (data not shown). In the rest of this study we used models based on the 8 Å radius. We inspected the predictive performance of models based on combined sets of features selected using different methods (Table 1). Given the performance of the tested models we selected the SVM (1) _Lasso model based on the combined set of the top 1% SVM-ranked features and the Lasso-selected features, termed it clonal model and used as the structural descriptor model in subsequent tests (Figure 2). The performance of the clonal model was not significantly higher than the performance of the Lasso model, however we chose the SVM (1) _Lasso feature set offering higher sensitivity. The AUC and sensitivity of the clonal model were significantly higher than those of the sequence-based g2p and aaindex models (p<0. 01, paired Wilcoxon test). In our approach the features were first selected and then evaluated on the entire sequence set in two subsequent steps involving cross validation. A test involving reselection of the SVM (1) _Lasso features in each cross validation run (nested cross validation) resulted in a decrease of the AUC of only approximately 1. 6 percentage points. Since the analysis of features selected for the clonal model was an additional goal of this study, we refrained from reselection of features within the separate cross validation runs. We regard this difference in performance as a potential uncertainty of our accuracy estimation inherent to the feature selection procedure. The accuracy obtained using nested cross validation is still significantly higher (p∼0. 003) than the accuracy of the g2p prediction suggesting that the selected structural and physicochemical features are more informative of tropism than the sequence alone [14]. In our approach we refrain from modelling of the side chains of the V3 loop. There is certain level of imprecision related to modelling of side chains due to the high flexibility and variability of the loop. We use our approach based on spheres as an approximation of the real structure of the loop that is costly to derive computationally and is unreliable. Such approximation of the structure is robust against indels as we observe no relationship of the model performance to the presence of indels in a sequence (Figure S5). We additionally tested the performance of a model based on a different V3 loop structure (Protein Data Bank (PDB) code 2QAD) [19] and performance of models based on combinations of structure- and sequence-based descriptors. However, these models did not yield an improvement in prediction performance (see Text S4). We compared the performance of our method with the performance of previously published structure-based methods for tropism prediction [20], [21] by testing the descriptor on the datasets used in the study of Sander et al. [21] (Sander dataset) and the study of Dybowski et al. [20] (Dybowski dataset). These datasets have sequence overlap of 19% and 58% with the clonal dataset, respectively. The overlap varies presumably due to different content of the Los Alamos database [25] at a given time point and to the different filtering methods used. In order to avoid overtraining and to test the performance of our model independently of the training dataset, we did not repeat feature selection on the datasets of other studies but based the predictions on the features of the clonal model. Structural descriptors trained on the Sander and Dybowski datasets were constructed based on these features and tested in a 10×10-fold cross validation setting. The clonal model showed performance similar to that of the original method of Sander et al. with a lower AUC 0. 901 (0. 923 reported by Sander et al.) and higher sensitivity 0. 782 (0. 774 reported by Sander et al.), see Table 2. The result of Sander et al. was obtained on a dataset with no insertions or deletions relative to the reference structure and involved costly side-chain modeling steps. In contrast, our result was based on features selected on a different dataset, and the prediction procedure did not involve structural modeling. The clonal model reached better performance on the Dybowski dataset in comparison to the original method with AUC 0. 948 (0. 937 reported by Dybowski et al.) and sensitivity 0. 838 (0. 810 reported by Dybowski et al.), see Table 2. We additionally tested the method on clinically derived patient data from the HOMER cohort [29] (HOMER dataset). We reran the feature selection procedures on this dataset and selected the best performing model (Lasso) as the clinical model. The clinical model showed AUC 0. 774 and the sensitivity 0. 463 [29], a result significantly higher (p<0. 01, paired Wilcoxon test) than the one of the g2p model [14] with AUC 0. 743 and sensitivity of 0. 451 (Table 2, Figure 3). To support this assessment, we performed an additional test on another independent patien-derive sequence set. On this dataset we observed a similar performance advantage of the clinical over the g2p model (see Text S5, Figure S6). Similar to the clonal model, the test of reselecting the Lasso features within the cross validation runs resulted in a decrease of performance of approximately 1. 8% in AUC, which is still significantly higher (p∼0. 002) than the performance of the g2p model. We additionally tested the effect of amino-acid ambiguities on the prediction accuracy of the clinical model and found that the combined information from both types of sequence positions, ambiguous and non-ambiguous is important for tropism prediction (see Text S6 and Table S1). As shown by Sing et al. [14] accuracy of tropism prediction methods applied on clinical data improves upon augmenting the sequence information with clinical correlates, such as VL or CD4+ T cell counts. Accordingly, adding such clinical information as additional features to the input of the clinical model significantly improved the predictive performance (p<0. 001, paired Wilcoxon test) over that of the clinical model to AUC 0. 803 and sensitivity 0. 474 (Table 2, Figure 3). This performance is also significantly higher than that of the g2p model containing clinical correlates in terms of both AUC and sensitivity (p<0. 001, paired Wilcoxon test). This demonstrates the higher prediction accuracy of the new method based on preselected structural and physicochemical features of the V3 loop over the commonly used sequence-based methods such as geno2pheno [14]. Finally, we tested the prediction performance of the clinical model on a dataset of sequences collected at therapy start from a German cohort of patients undergoing MVC therapy (MVC dataset). Among the 53 sequences five originate from patients who experienced therapy failure. With the decision cutoff at the 11/25 rule specificity of the HOMER dataset (specificity 0. 928, score 0. 097) three of these sequences were predicted as X4 viruses in accordance with the patient therapy outcome. The two remaining sequences of patients experiencing therapy failure that were predicted as R5 viruses were also phenotyped as R5 virus, which suggests the presence of undetectable minorities as the potential reason for the classification error. The remaining 48 patients experienced therapy success. 41 of the cases were classified as R5 viruses by the clinical model, which is in accordance with the patient therapy outcome. Out of seven remaining cases that were classified as X4 viruses, two were phenotyped as X4 viruses. For comparison, we predicted tropism of the sequences in this dataset using the g2p model. This sequence-based prediction reported correctly only two therapy failure cases but also 44 therapy success cases which is more than the clinical model predicted. As a measure of the quality of predictions of the MVC dataset we used the Matthews correlation coefficient (MCC), which quantifies the correlation of the observed and predicted binary classification and is suited for datasets with an unbalanced class proportion. Therapy outcome prediction based on structural descriptor showed overall accuracy of MCC = 0. 34 comparing favorably with g2p model yielding MCC = 0. 29. Phenotypic characterization was only available for a subset of 28 sequences from the MVC dataset (Trofile dataset). In this subset the phenotype appeared to be the best predictor of the therapy outcome with one correctly predicted therapy failure case out of three and 23 correctly predicted therapy successes out of 25 (MCC = 0. 25). The clinical model reported the same number of correctly predicted therapy failure cases and lower number of 20 correctly predicted therapy success cases (MCC = 0. 10). The clinical model scored higher than the g2p method that did not report correctly any of the therapy failure cases and predicted correctly 23 therapy successes (MCC = −0. 10). Additionally, the clinical model correctly classified all X4 sequences in the Trofile dataset reaching MCC of 0. 660 and favorably comparing with the g2p showing MCC of 0. 352. Overall, the phenotype as well as the structural descriptor model and the g2p model trained on clonal data showed a generally lower capacity of detecting therapy outcome compared to the models trained on clinical data. Detailed results of the MVC dataset analysis are provided in Tables S2 and S3. In order to facilitate the interpretation of the large number of selected features we clustered the 56 amino acid indices into four groups (Figure 4) using hierarchical clustering. Cluster 1 is composed of two types of indices – related to residue size and volume and to residue occurrence in proteins. Cluster 2 contains the smallest number of indices and is composed of indices related to residue charge. Indices of cluster 3 are related to the secondary and tertiary structure of proteins. Cluster 4 contains indices related to different structural properties e. g. residue occurrence in β-sheet, solvent accessibility, amino acid polarity or hydrophobicity. By combining amino-acid indices with specific positions on the V3 loop, the proposed features can be interpreted in terms of physicochemical properties along the structure of the loop. The features selected for the clonal model are informative about the coreceptor usage. Their analysis can therefore provide insights into the physicochemical and structural factors of viral tropism.
Physicochemical and structural properties of proteins determine their binding affinities. Prediction methods of HIV-1 coreceptor usage based solely on the V3 sequence do not account for this type of properties nor do they provide the information on loop characteristics that are crucial for the interaction. Prediction models incorporating loop structure can provide such information. However, previously reported structure-based prediction models suffer from limitations in terms of (i) runtime and software complexity – which prevents their accessibility via a tool publicly available online – and (ii) interpretation of the prediction result. Here we present a prediction model of HIV coreceptor usage based on V3 sequence and structure [18] that overcomes these limitations. The method is based upon a set of features that was selected from a large initial feature set. The model shows better performance than the one based on the initial feature set, both in terms of prediction accuracy and computational efficiency and shows higher predictive power than the prediction method based uniquely on sequence. In addition, the proposed model affords an interpretable set of physicochemical properties located in specific parts of the loop structure that play a role in determining viral tropism. The approach is generic and can be applied in other supervised learning applications involving the combination of sequence or evolutionary information, physicochemical and structural properties. In particular, computational biology and medical applications involving molecular binding mechanisms are good potential candidates for achieving improved accuracy and interpretability with the proposed approach. Our clonal model was developed on a sequence set comprising different HIV-1 subtypes. The limited number of sequences of each subtype and the high variability of the V3 loop sequence which obviates a clear subtype classification advocate use of a common model for all subtypes, an approach also applied by other prediction methods [11], [14], [20], [21]. The proposed structural descriptor appears to entail information on both structure and sequence, as adding the binary encoding of the sequence to the descriptor does not improve the performance of the clonal model (Text S4). In contrast, the distance-based descriptor of Sander et al. [21] is complementary to the sequence which is demonstrated by the improved performance of the descriptor, when combined with the sequence information. Our method shows a moderately but significantly higher prediction performance of approximately 3 percentage points over the model based on sequence only [14] both on the clonal and clinical datasets with and without patient clinical markers. The model additionally shows a higher or similar prediction performance to that of other structure-based methods [20], [21] without modeling steps that increase the computational cost of the prediction procedure. Our model shows higher prediction accuracy when applied on external datasets from these studies than on the dataset it was trained on, suggesting the selected feature set is not biased towards the used sequence set. Note that our approach implements an approximate representation of the structure of the V3 loop with the goal of predicting coreceptor usage based on an interpretable model and without predicting of an accurate structural model of the V3 loop with respect to insertions or deletions compared to the template structure. Also, unlike method of Sander et al. our approach does not involve modeling of side chains. The effectiveness of a thorough structural modeling and especially of modeling side chains for the purpose of predicting tropism is likely to be limited by the variability of the V3 loop sequence and the structural flexibility of the loop. In our method spheres are used to represent structural proximities over which physicochemical properties are averaged. In this way our approach accounts for the uncertainty of the structural conformation of the loop and avoids costly modeling steps. The accuracy and efficiency of our approach enables its use as a server application. Unlike previously developed structure-based methods [20], [21], our method was tested not only on clonal data but also on clinically derived (bulk) data and showed significantly better performance over the established sequence-based approach. Given the common usage of this type of data in patient diagnosis and the potential difficulties it represents in classification [14], predicting tropism and MVC therapy outcome based on clinical data represents a more realistic scenario for training and assessment of classification methods than prediction of tropism based on clonal data. We also assessed the capacity of our method to predict MVC therapy outcome. For the purpose of this validation, we used a cohort of patients treated with MVC. This analysis is limited due to the low number of cases in the MVC dataset. With increasing use of entry inhibitors, therapy outcome data are expected to become more abundant and the capacity to train models predicting therapy outcome will improve. The higher performance of the clinical over the clonal model in predicting therapy outcome suggests that comprehensive datasets appropriate for specific prediction goals can produce more reliable models. The analysis of features informative of viral tropism points to two critical sites in the loop stem, comprising residues 304,307 and 319–322, respectively and to position 324 located more closely to the base of the stem. The charge of amino acids at these sites is known to play role in coreceptor binding [6], [7]. Additionally, our analysis points to the importance of the propensities of these amino acids for forming specific secondary structures. Residues on both sides of the stem form interactions in the bound conformation of the loop probably contributing to the rigid form of the loop upon binding. The combined effect of charge and propensity for specific local structural conformations might therefore contribute to acquiring the adequate binding site complementarity and local loop conformation required for specific coreceptor binding. The results of other studies of structural features related to HIV tropism are in general accordance with our results. A recently published method [38] predicts coreceptor usage based on a perturbation vector reflecting relative change in compatibility of a given V3 the sequence and structure with the reference structure [39]. Ten most important positions for the coreceptor usage, according to this study are positions 302–304,306–307,309,312,322,324–325. However, no additional interpretation of the characteristics of these positions is provided. Sander et al. point to the residues 298,302,306,308,315,317,319,321,322 and 328 involved in residue pairs important for tropism. The regions found in our analysis are in close proximity or in-between the positions listed by Sander et al. on the V3 structure. However the ranking of Sander et al. is based on the importance of distances among functional atom types in the V3 loop, which is not equivalent to the importance of the residue itself. Findings reported by Dybowski et al. [20], point to electrostatic hulls around positions 306,321 and 322, and between position 301 and 326 as the features of highest importance for the classification which is also in agreement with our results. Additionally, the authors point to hydrophobicity of residues 303 and 307 as important for viral tropism. Given the considerable structural flexibility and sequence variability of the V3 loop, individual features of this region distinguishing between the two virus phenotypes are hard to define. We performed a comprehensive analysis of a large number of physicochemical residue characteristics in various locations on the loop and pointed to those that are the most informative of tropism. The resulting method offers higher performance than the standard sequence-based approach with a comparable efficiency and a direct interpretation of structural and physicochemical determinants of tropism. The method has been implemented as a server application within the geno2pheno framework under http: //structure. bioinf. mpi-inf. mpg. de/.
To construct the clonal dataset we screened the Los Alamos database [25] for all phenotyped V3 loop sequences. In order to avoid bias due to overrepresentation of data from the same patient we filtered the dataset extracting one randomly chosen sequence per patient. The resulting dataset contains 1186 sequences with tropism annotation, 215 of which are annotated as X4 viruses. In the dataset 501 sequences are of subtype B, 286 of subtype C, the remaining sequences are of other and recombinant subtypes. We aligned the sequences in the dataset and the sequence of the V3 loop of the PDB entry 2B4C using ClustalW [40] obtaining an alignment of length 50. In order to assess the robustness of this alignment we aligned each of the sequences in our dataset to the alignment of all the remaining sequences. In this test all the alignments were identically reproduced suggesting that the sequence alignment prior to prediction results in the correct alignment. The clonal dataset is provided as Supplemental Text S7. In our study the positions in the gp120 sequence are numbered relative to the reference as previously described [41]. See Figure S11 for the correspondence between the numbering of the V3 loop positions in the reference sequence HXBc2 [41] and the subtype B consensus sequence. We used the amino-acid indices collected in the AAindex database [23]. In this database various physicochemical and biochemical properties of amino acids are stored in the form of numerical indices. Due to the high number of over 500 indices in the database many of which are redundant we used a representative and interpretable subset of 54 indices, selected using multivariate statistical analysis [24]. This is a minimal fully representative set of indices. Reducing it further would limit the physicochemical information provided by our descriptor. Two of the selected indices named “Normalized frequency of beta-turn” and “Free energy in beta-strand region” were represented by duplicate entries in the AAindex database showing minor differences (AAindex entries: CHOP780101/CHOP780203 and MUNV940104/MUNV940105 respectively). To avoid arbitrary selection between the duplicate entries we used both of the ambiguous indices, which resulted in a set of 56 indices selected for this study. The descriptor of the V3 loop was based on the published structure of the V3 loop with PDB [42] code 2B4C [18]. To construct the descriptor for each V3 loop sequence we used spheres defining structural neighbors inside the loop structure within which the physicochemical properties of residues are averaged as detailed below. The spheres are positioned along the reference loop backbone and centered at its residues. Specifically, positions of residues were defined as the position of the representative atom of each residue in the structure – the Cα atom for Glycine and the Cβ atom for all other amino acid types. Positions of insertions in the alignment relative to the reference structure were inferred based on the positions of representative atoms of the residues at both ends of the insertions (flanking atoms). First, a line connecting the flanking atoms was calculated. Then the inserted residues were placed along the line at equidistant positions. This way we approximate the location of atoms on the loop structure without precise modeling of the structure which is likely to be inaccurate given the flexibility of the V3 loop structure and which would considerably slow down the prediction process. The resulting coordinates of the residues of the V3 loop sequences were used as centers of the spheres defining the structural neighborhoods in the loop structure. In addition to the set of spheres corresponding to alignment positions additional spheres were positioned at the midpoints of lines connecting centers of each pair of consecutive alignment spheres. This way we obtained a set of 99 spheres – 50 corresponding to alignment positions and 49 positioned in-between consecutive alignment positions. Example spheres are illustrated in Figure 1. Each V3 sequence position was mapped to a sphere if the corresponding representative atom was located within the given sphere. The details of the selection of the sphere radius and Gaussian smoothing parameter within the spheres are described and illustrated in Text S1 and Figure S1. The model based on the structural descriptor classifying viruses as R5 or X4 was constructed using a linear SVM [43] implemented in the R package e1071 [44]. For model evaluation we used the ROC curve that illustrates the trade-off between specificity and sensitivity. The AUC and the specificity at the sensitivity of the 11/25 rule were used as measures of model performance. We used the R package ROCR [45] for visualization and evaluated the models with ten times ten (10×10) fold cross validation. Each descriptor feature was normalized to [0,1] within the training dataset. We used two classification methods performing feature ranking: Random Forests (RF) [26], with the mean decrease in Gini index and linear SVMs with the feature weights as two measures of feature importance [27]. We also used Lasso regression [28] which performs feature selection by assigning zero coefficients to the less important features. For the methods producing feature ranking (RF and linear SVM) we tested two cutoffs for the selected features: top 1% and top 5% of a gamma distribution fitted to the ranking of all the features using maximum likelihood. We used all features selected by the Lasso regression method. The feature ranking of the SVM and Lasso regression methods was obtained via an average of a 10×10-fold cross validation. The RF method performs internal randomization, its feature ranking was therefore inferred from a single run of the method. We tested the performance of models based on subsets of features selected by each method and combined feature sets selected by different methods. Models based on subsets of selected features were named after the feature selection method with the percent cutoff indicated in parentheses (e. g. SVM (1) ). Names of models based on combinations of feature sets selected using several feature selection methods were composed of the corresponding feature selection methods separated by an underscore (e. g. SVM (1) _Lasso). As the analysis of the features selected for the clonal model was a goal of our study, feature selection was performed on the entire clonal dataset. To assess how the choice of the set of sequences on which the features are selected impacts the model' s prediction accuracy, we performed two different types of tests. In the first test, features of the model were reselected on the training set in each cross validation run on the clonal set (nested cross validation). In the second test we applied the features of the clonal model to other sequence sets – Sander and Dybowski datasets. The HOMER dataset was filtered to contain one randomly chosen sequence per patient, which resulted in a set of 954 sequences out of which 167 comprised X4 viruses. Each sequence in the clinical dataset represents a population of variants genotyped and phenotyped in bulk, an approach used in the routine clinical practice. These sequences contain ambiguous positions with alternative amino acids representing different variants in the population. The ambiguous positions were represented by a balanced average of vectors of indices of all alternative amino acids at a given position. Due to these differences between the clinically and clonally derived data, we repeated the feature selection on this dataset and constructed the clinical model. The MVC dataset comprises 53 patient cases under MVC therapy whose therapy outcome can be assessed based on the viral load (VL). We define as therapy success an observed 2log decrease in VL with respect to the level immediately before the therapy start or a VL drop below 50 copies/ml measured three months after the therapy start [46]. We classified the viruses sequenced at therapy start with respect to their tropism in order to investigate the capacity of the structural descriptor to predict the therapy outcome. Since the MVC dataset was derived in clinical bulk sequencing we used the clinical model to predict the phenotype of the sequences in this dataset. We used the prediction score at the specificity of 11/25 rule which corresponds to a false discovery rate (FDR) of 6. 28% in the HOMER dataset as a classification cutoff between the R5 and X4 viruses. The FDR is an estimate of the expected proportion of sequences incorrectly classified as X4 viruses with a given cutoff and is calculated as the fraction of R5 viruses in the training set scored above the cutoff among all sequences scored above the cutoff in 10×10-fold cross validation. In the MVC dataset we additionally distinguish 28 sequences that were phenotyped using the Trofile assay (Trofile dataset). Summary statistics for all datasets used are presented in Table 3. Clustering of the 56 amino acid indices was performed in order to facilitate the interpretation of the large number of selected features. As a similarity score among the indices we used the absolute value of their correlation. This way, indices that express the same affinities among amino acids are considered similar. We performed hierarchical clustering of the 56 amino acid indices and computed silhouette values [47] in order to select the best set of clusters. The highest silhouette value was obtained for a partitioning of indices into 12 clusters. The highest silhouette value of a partitioning of indices into fewer than 12 clusters was obtained for four clusters. We selected four as the number of clusters for further analysis as it represents a small and interpretable number of groups of indices. The silhouette values as well as the 12 clusters are shown in Figures S12 and S13. | Human Immunodeficiency Virus (HIV) requires one of the chemokine coreceptors CCR5 or CXCR4 for entry into the host cell. The capacity of the virus to use one or both of these coreceptors is termed tropism. Monitoring HIV tropism is of high importance due to the relationship of the emergence of CXCR4-tropic virus with the progression of immunodeficiency and for patient treatment with the recently developed CCR5 antagonists. Computational methods for predicting HIV tropism are based on sequence and on structure of the third variable region (V3 loop) of the viral gp120 protein — the major determinant of the HIV tropism. Limitations of the existing methods include the limited insights they provide into the biochemical determinants of coreceptor usage, high computational load of the structure-based methods and low prediction accuracy on clinically derived patient samples. Here we propose a numerical descriptor of the V3 loop encoding the physicochemical and structural properties of the loop. The new descriptor allows for server-based prediction of viral tropism with accuracy comparable to that of established sequence-based methods both on clonal and clinically derived patient data as well as for the interpretation of the properties of the loop relevant for tropism. The server is available under http: //structure. bioinf. mpi-inf. mpg. de/. | Abstract
Introduction
Results
Discussion
Materials and Methods | sequence analysis
protein structure
biology
computational biology
macromolecular structure analysis | 2013 | Analysis of Physicochemical and Structural Properties Determining HIV-1 Coreceptor Usage | 9,411 | 325 |
Influenza viruses elude immune responses and antiviral chemotherapeutics through genetic drift and reassortment. As a result, the development of new strategies that attack a highly conserved viral function to prevent and/or treat influenza infection is being pursued. Such novel broadly acting antiviral therapies would be less susceptible to virus escape and provide a long lasting solution to the evolving virus challenge. Here we report the in vitro and in vivo activity of a human monoclonal antibody (A06) against two isolates of the 2009 H1N1 pandemic influenza virus. This antibody, which was obtained from a combinatorial library derived from a survivor of highly pathogenic H5N1 infection, neutralizes H5N1, seasonal H1N1 and 2009 “Swine” H1N1 pandemic influenza in vitro with similar potency and is capable of preventing and treating 2009 H1N1 influenza infection in murine models of disease. These results demonstrate broad activity of the A06 antibody and its utility as an anti-influenza treatment option, even against newly evolved influenza strains to which there is limited immunity in the general population.
Controlling the spread of influenza remains a major challenge due to the unpredictable nature of the virus. Recently, a novel human adapted H1N1 virus has emerged and progressed globally such that the World Health Organization (WHO) has declared the first influenza pandemic in 40 years [1], [2]. Globally, efforts have been undertaken to produce vaccines and stockpile small molecule antiviral reserves to prevent and treat widespread influenza disease. While these strategies are effective, they are not without limitations. Vaccines have not provided lasting immunity against influenza because of viral mutation (“antigenic drift”) and reassortment (“antigenic shift”) [3], [4], [5], [6]. Popular small molecule antiviral treatments (oseltamivir) have recently lost effectiveness due to the rapid proliferation of seasonal H1N1 strain resistanc, demonstrating the urgent need to develop novel treatments for influenza infection and disease. Such new treatment options would ideally be both broadly protective and provide a novel mechanism of attack against the virus. Antibodies have very desirable properties as prophylactic and therapeutic agents: long serum half-life, low immunogenicity and high specificity for antigens. In addition, antibodies are currently being used against infectious disease. For example, antibody clinical prophylaxis against RSV is a standard of care and antibody therapy is in development for treatment of anthrax [7], [8], [9], [10]. A related passive immunity strategy against influenza was used in the past during times of crisis, and retrospective studies have quantified the benefits of such strategies [11]. Furthermore, it would be beneficial for this agent to act on a highly conserved site to increase its therapeutic lifespan. Recently, work by us and others have described novel human monoclonal antibodies capable of very broad heterotypic protection that could be used in the treatment and prevention of influenza virus infections [12], [13], [14]. Here we report in vitro neutralization and in vivo efficacy in prophylactic and therapeutic mouse models of the novel 2009 H1N1 pandemic influenza virus infection by one such broadly protective antibody derived from an H5N1 avian influenza survivor.
Human IgG1 antibody was expressed and purified essentially as previously described [12]. The A/California/04/2009 virus used in the microneutralization studies is a recombinogenic virus composed of the hemagglutinin (HA) and neuraminidase (NA) gene segments from A/California/04/2009 and the remaining influenza viral gene segments are from A/PR/8/34 [15]. The recombinant virus was propagated in MDCK cell culture. All other strains were amplified in 10–11 day old embryonated hens' eggs. Microneutralization assays were performed as previously described [12]. Briefly, two-fold dilutions of mAb were incubated with 100 TCID50 of virus for 1 h at 37°C prior to addition to monolayers of MDCK cells. Cell monolayers were incubated for 72 h, and the presence of virus in supernatant was determined by HA assay of duplicate samples. The neutralizing titer was defined as the minimum inhibitory concentration at which the infectivity of 100 TCID50 of the appropriate virus for MDCK cells was completely neutralized in duplicate wells. Animal experiments were performed in accordance with the guidelines of the Mount Sinai School of Medicine and St. Jude Children' s Research Hospital Institutional Animal Care and Use Committees (IACUC). Female 6–8 weeks old Balb/C (Jackson Laboratories) or DBA/2 (Charles River) mice were housed 5–6 per cage in ABSL3+ containment. Food and water were provided ad libitum. For the prophylactic studies, mice (5–6 per group, except where noted) received 1,2. 5,10, or 25 mg antibody A06 per kg of bodyweight in approximately 200–300 µL of sterile phosphate-buffered saline (PBS) by intraperitoneal (IP) injection. The control groups received 200–300 µL of 25 mg/kg non-immune human serum IgG (Sigma) (n = 3) or PBS by IP injection. Antibody was administered either 1 hour (Balb/C) or 24 hours (DBA/2) before being challenged with A/California/04/09, which had been previously mouse-adapted by 9 sequential lung passages, or wild-type A/Netherlands/602 virus. Mice were inoculated by intranasal administration of 3. 3,25, or 33 MLD50 (50% mouse lethal dose) influenza virus in 30–50 µL of PBS. 2000 PFU (25 MLD50) of mouse-adapted A/California/04/09 was used for infection of Balb/C mice in the prophylactic and therapeutic studies, while 10 PFU (3. 3 MLD50) and 100 PFU (33 MLD50) was used for the A/Netherlands/602 strain in DBA. 2 mice. Symptoms preceding death are weight loss >30% and general inactivity. Morbidity and mortality were monitored either daily or at days 0,3, 7,10, and 14. For therapeutic studies, Balb/C mice (10 per group) were given a lethal virus dose of 25 MLD50 A/California/04/09 (2000 PFU) followed by a single 15 mg/kg dose of antibody 24,48,72,96,120, or 144 hours post infection. Morbidity and mortality were monitored for 17 days and the mice were weighed on days 0,3, 7,10,14, and 17 following virus challenge. For dose escalation studies, mice were given a viral dose of 3. 3 MLD50 (10 PFU) A/Netherlands/602/209 followed either 1 day or 2 days post infection with a single IP injection of 2. 5,10 or 25 mg/kg dose of antibody A06, vehicle (PBS), or non-immune IgG (25 mg/kg dose). Morbidity and mortality were monitored for 14 days and the mice were weighed daily following virus challenge. All data for both the prophylactic and therapeutic studies was plotted for days 3,7, 10,14, and 17 (where appropriate). Survival data were plotted (Kaplan-Meier) and analyzed using the logrank test to determine statistical significance (P<0. 05). Mean weight data were also plotted. All data were plotted and analyzed using GraphPad Prism v. 5. 02 software. One thousand non-redundant 2009 novel H1N1 hemagglutinin amino acid sequences deposited to the Influenza Sequence Database [16] were aligned using MUSCLE v4 multiple sequence alignment function accessible through the Influenza Sequence Database website (http: //www. ncbi. nlm. nih. gov/genomes/FLU/FLU. html). Sequences were visually inspected for amino acid changes within the predicted antibody binding site in the hemagglutinin HA2 region described in [13], [17]. Variants in both the predicted contacting and non-contacting positions were noted along with the frequency of occurrence.
We previously reported the discovery of broadly neutralizing antibodies from avian influenza survivor antibody libraries, capable of mechanistically novel, heterosubtypic neutralization against numerous H1N1 and H5N1 viruses [12]. Subsequent to our publication, others have reported highly related and broadly neutralizing human antibodies [13], [14]. The novel unifying mechanism these anti-hemagglutinin neutralizing antibodies exhibit is that they do not inhibit virus-induced hemagglutination. Structural analysis by both Sui et al. and Ekiert et al. have shown the antibodies bind to the highly conserved stem of hemagglutinin (HA) that prevents a conformational change required for viral host cell fusion [13], [17]. The reason these antibodies are broadly neutralizing is attributed to the high sequence conservation of the antibody epitope between H1, H5 and H9 type hemagglutinins, which is coincidentally maintained in the newly emergent 2009 pandemic H1N1 strain (Table 1). From these collective observations, we predicted the 2009 pandemic H1N1 influenza would be susceptible to neutralization by the previously described antibody isolated from the Turkish avian influenza survivor libraries. As a first step to test our prediction the A06 antibody (previously referred to as mAb1 [12]) was tested in in vitro viral microneutralization assays against a recombinogenic virus containing the 2009 H1N1 pandemic reference isolate A/California/04/2009 influenza virus (CA04) HA and neuraminidase (NA) proteins upon a A/PR/8/34 based viral background, (hereafter referred to as A/California/04/2009 6: 2) [18]. In these assays the A06 antibody demonstrated complete viral neutralization of the 2009 H1N1 virus at final concentrations as low as 10 µg/ml (Table 1), which is in good agreement with neutralization results against other H5N1 and H1N1 strains that we have tested (Table 1). To assess whether the A06 antibody could prevent or decrease the severity of influenza infection in vivo, we performed a dose-escalating study in a murine prophylactic model of disease. Briefly, Balb/C mice were given a single IP dose of A06, infected intranasally 1 hour later with 25MLD50 of a mouse-adapted CA04 H1N1 virus (see methods), and then monitored for survival and body weight over the following 14 days (Figure 1A). In this study, vehicle treated mice either died or were euthanized 6 to 10 days post-infection and displayed pronounced progressive weight loss during the course of infection. In sharp contrast, mice treated with either 25 mg/kg or 10 mg/kg of A06 survived the lethal challenge and regained lost weight by day 7. Survival in the 2. 5 mg/kg treatment group was 83%, with the mice losing more weight than the higher dosed groups in the first 3 days post-infection, but still regaining their pre-study weights by day 14. Survival in the 1 mg/kg group was observed, but was the least prominent of all treatment groups (33%) with the surviving mice losing body weight through 10 days post-infection. Logrank test analysis of the survival curves demonstrated statistical significance (P<0. 0001). These studies demonstrate antibody A06 is able to protect mice from the lethality and weight loss associated with influenza virus infection in a dose dependent manner. To further support the previous results, we wanted to show efficacy on a non-mouse- adapted novel human H1N1 strain as well as assess efficacy against two different levels of viral challenge. In the subsequent prophylaxis study, we used a novel H1N1 influenza strain A/Netherlands/602/2009 (Netherlands602) in the more sensitive and susceptible DBA/2 mouse strain at both 3. 3MLD50 (Figure 1B) and 33MLD50 (Figure 1C). In the 3. 3MLD50 challenged study, mice treated with vehicle died or were euthanized between 7 and 9 days post infection and displayed pronounced progressive weight loss during the course of infection. Antibody A06 administration provided significant survival (P<0. 0001) and considerable body weight maintenance benefits. Specifically, mice challenged with 3. 3MLD50 Netherlands602 in both the 25 mg/kg and 10 mg/kg dose groups survived and lost some weight through day 7 that was rapidly regained to their pre-study levels by day 10. Survival in the 2. 5 mg/kg treatment group was 80% with greater weight loss observed compared to the higher dosed groups. In the subsequent DBA/2 study where mice were challenged with 10 times more virus (33MLD50) than the previous A06 antibody treated groups they also displayed significant survival (P<0. 0001) and substantial body weight maintenance benefits compared to controls. Specifically, the mice manifested disease and mortality more rapidly than those in the 3. 3MLD50 challenge study, as both groups treated with PBS or non-immune IgG died or were euthanized between 6 and 7 days post infection. In contrast, all mice treated with 25 mg/kg of antibody A06 survived, whereas those treated with 10 mg/kg or 2. 5 mg/kg of A06 had an 80% survival rate. The average body weight of all the treated groups declined through the first 7 days post-infection, but was restored to pre-study levels by day 10. In summary, prophylactic administration of antibody A06 appeared beneficial in abrogating influenza-mediated weight loss, allowing faster recovery of infected animals. Passive immunity may provide both prophylactic and therapeutic benefits against influenza infection. To address whether the A06 antibody is therapeutically effective following infection, we treated groups of CA04-infected Balb/C mice (25MLD50 infectious titer) with a single 15 mg/kg dose of antibody A06 at 1,2, 3,4, 5, or 6 days post-infection. All mice dosed 1 day after infection survived, 90% of mice dosed 2 days after infection survived, and 50% of the mice dosed 3 days after infection survived. All mice dosed 4 days post infection and later either died or were euthanized between days 7 and 10. Weight loss in the therapeutic study was more severe than seen in the mice treated prophylactically and similar to the vehicle-treated mice in the prophylaxis study. All mice in this study lost ∼30% of their body weight due to the established novel H1N1 infection (Figure 2B) and earlier treatment appeared linked to higher study end weights and survival (Figure 2A and B). These results demonstrate the utility of a therapeutic passive immunity approach against an emergent strain of influenza and extend previous findings that heterotypic neutralizing antibodies are beneficial in in vivo models of influenza prophylaxis and therapy. As a significant benefit in overcoming influenza infection was seen in treatment groups administered A06 antibody at 1 or 2 days post infection, our next study expanded the analysis at these times through a dose escalation study. Specifically, we administered the A06 antibody at 2. 5,10, and 25 mg/kg either 1 day (Figure 3A) or 2 days (Figure 3B) after infection with 3. 3MLD50 of the Netherlands602 strain of the 2009 pandemic H1N1 virus in DBA/2 mice. As seen previously, PBS vehicle treated mice died or were euthanized by 9 days post infection and displayed pronounced progressive weight loss during the course of infection. However, mice receiving 25 mg/kg or 10 mg/kg doses of antibody A06, either at 1 day and 2 days post infection, survived the Netherland602 virus infection, corroborating results found with the CA04 viral challenge (Figure 3A and 3B, left panels). Importantly, the lowest antibody dose (2. 5 mg/kg) was sufficient to overcome infection in all except one animal. As a benefit the treated mice also gained weight after treatment with A06, arriving at their pre-study weight by day 14 (Figure 3A and 3B, right panels). In summary, our results demonstrate that antibody A06 is a very effective treatment following novel H1N1 infection, even at doses of 2. 5 mg/kg in two different mouse models of influenza infection and treatment. Broadly active anti-influenza agents need to target essential sites that are minimally prone to mutation. As a predictive assessment of the potential efficacy of the A06 antibody against current H1N1 pandemic isolates, we analyzed a large number of novel influenza hemagglutinin protein sequences within the proposed A-helix epitope [13], [17] for genetic drift. From the analysis of 1000 full length hemagglutinin protein sequences (NCBI Influenza Virus Resource November 11,2009) [16], we found only 5 isolates that varied from CA04 reference strain at three positions in the proposed A-helix antibody epitope on the HA2 subunit (Table 2). Only one of the five isolates (Canada-NS/RV1535/2009) contained a mutation to a proposed contact point, which was a conservative substitution of valine for isoleucine at residue 56. It is significant to point out that isoleucine is found at an analogous position in the H9N2 Hong Kong/1073/99 which is recognized by the A06 antibody (Table 1 and unpublished data), suggesting the isolate would still be susceptible to the A06 antibody. Three of the remaining four mutations occurred at the non-contacting residue 43, where lysine, serine, or aspartic acid was found instead of asparagine. The final mutation was a proline replacement for alanine non-contacting position 44. In summary, the analysis suggests the isolates display limited allowances for genetic drift within this region that may maintain susceptibility to A06.
We previously demonstrated, in vitro, that the A06 antibody neutralizes a broad range of seasonal H1N1 and avian H5N1 influenza viruses causing human disease. In this study, we extend these results and demonstrate that the A06 antibody is able to protect from and treat the antigenically distinct 2009 pandemic H1N1 virus infection in mouse models and also neutralize the current seasonal H1N1 Brisbane/59/2007 strain in vitro. These results provide further evidence for the use of passive immunity as a weapon against influenza infection. Passive immunity offers several benefits in comparison to current chemotherapeutic anti-viral treatment options. First, passive immunity provides the opportunity to protect at-risk individuals from infection. At-risk segments of the population include those who do not mount an immune response to vaccine, the immunocompromised, those in poor health, pregnant women, and those in critical care. The potential for long-lasting protection arising from a single injection of antibodies such as A06 is appealing. In addition, while orally available drugs are desirable to reach a larger patient population and increase patient compliance in courses of treatment, their use in critical care settings involving the later stages of disease is limited by the route of administration. Quite simply, injectable therapies are needed for patients that are unable to receive orally administered anti-influenza treatment. Current anti-viral treatments provide ease of use and therapeutic benefit early in the course of infection. However, they suffer from several limitations, namely high rates of resistance, as exhibited recently in the seasonal H1N1 virus [19]. The unexpected speed at which the H274Y mutation conferring oseltamivir resistance took over as the dominant strain in the 2007–2008 influenza season demonstrates the challenges facing widespread use of anti-viral agents targeting the neuraminidase protein [3], [19], [20]. Antibodies such as A06 that attack a highly conserved region of the hemagglutinin protein and not the mutagenic hot spots near the receptor binding domain or the neuraminidase protein may face fewer challenges arising from mutation. Using the method of Caton, et al [21], we have not been able to generate escape mutants after multiple attempts using the A06 antibody on both H1N1 and H5N1 influenza strains, suggesting that A06 is attacking a conserved, susceptible portion of the virus (JS, unpublished results). Furthermore, conservation of the predicted antibody binding site in the 2009 pandemic strain isolates demonstrates the epitope has not changed significantly from the time of its emergence in March 2009. It is likely the ability to tolerate change in the hemagglutinin A-helix/fusion peptide region may be highly restricted due to functional constraints, as evidenced by the maintenance of this epitope across numerous influenza sub-types. However, an alternative interpretation to this observation is that the region has not been sufficiently pressured to change and may mutate when subjected to greater selective pressure, even though we have not seen it yet in escape mutant analysis. Nevertheless, even if escape were possible, passive immunization would likely be highly effective when used in conjunction with other established therapies to reduce the prospect of viral escape or resistance. Here we have presented A06 antibody in vitro neutralization results with numerous H1N1 and H5N1 strains from each sub-type. Though recent H1N1 strains were neutralized with similar antibody concentrations, two older strains, A/PR/8/34 and A/Texas/1991, required substantially higher amounts of antibody to be effective. Upon further sequence examination of these two recent strains we have observed potential N-linked glycosylation sites proximal to the predicted epitope in A/PR/8/34 (amino acids 285–287) and A/Texas/1991 (amino acids 286–288). It is possible that glycosylation at theses sites sterically hinders the antibody and reduces its efficacy in this in vitro system. Further testing will need to be performed both in vitro and in vivo to evaluate the relevance of such a potential glycosylation site. Still, all H5N1 strains tested, representing all major clades of highly pathogenic avian influenza, were effectively neutralized by antibody A06. Considering the ability of the antibody to neutralize the novel H1N1 virus, multiple seasonal H1N1 isolates, isolates from all clades of human H5N1, and that the proposed epitope is highly conserved amongst the initial sampling of one thousand reported novel H1N1 hemagglutinin isolates, we predict that A06 will be active against influenza strains bearing this epitope. Additional testing is required to determine the efficacy and utility of passive immunity in man. However, the profile of the A06 antibody and other broadly protective antibodies warrants their testing in man. Success of such antibodies would justify their use in cases of local, national, and global crisis. In addition, injectable administration of antibodies such as A06 could protect critical care patients unable to receive orally- administered anti-viral therapy. Use of passive immune therapy in an integrative approach with anti-viral chemotherapeutics could even decrease the frequency and speed at which resistance to either agent is generated. Furthermore, as these types of antibodies were found in large survivor, vaccinee, and naïve donor combinatorial antibody libraries, it suggests the mode of activity is immunologically relevant. As a result, these broadly reactive anti-fusion antibodies and their protective mechanisms should also be used as an additional guide in the production and assessment of all future influenza vaccines. | Influenza viruses constantly challenge our ability to prevent and treat their resulting infection. From a survivor of the H5N1 influenza we have discovered an antibody that is effective against both H5N1 and seasonal H1N1 influenza viruses. Here we show the antibody is effective against 2009 pandemic influenza in a cell culture assay and also in mouse models of disease when given before and even after lethal influenza infection. The present work demonstrates the viability of this particular antibody and the general approach of using antibodies against viral pathogens as opposed to traditional treatments that are losing their efficacy for the prevention and treatment of influenza infection. We conclude the efficacy of this antibody warrants further experimental testing as an alternative therapy for treatment in man. | Abstract
Introduction
Methods
Results
Discussion | biotechnology
immunology/antigen processing and recognition
infectious diseases
immunology/immune response
infectious diseases/viral infections
infectious diseases/respiratory infections
immunology/immunity to infections
biochemistry/drug discovery | 2010 | Protection from the 2009 H1N1 Pandemic Influenza by an Antibody from Combinatorial Survivor-Based Libraries | 5,429 | 159 |
Bacterial biodiversity at the species level, in terms of gene acquisition or loss, is so immense that it raises the question of how essential chromosomal regions are spared from uncontrolled rearrangements. Protection of the genome likely depends on specific DNA motifs that impose limits on the regions that undergo recombination. Although most such motifs remain unidentified, they are theoretically predictable based on their genomic distribution properties. We examined the distribution of the “crossover hotspot instigator, ” or Chi, in Escherichia coli, and found that its exceptional distribution is restricted to the core genome common to three strains. We then formulated a set of criteria that were incorporated in a statistical model to search core genomes for motifs potentially involved in genome stability in other species. Our strategy led us to identify and biologically validate two distinct heptamers that possess Chi properties, one in Staphylococcus aureus, and the other in several streptococci. This strategy paves the way for wide-scale discovery of other important functional noncoding motifs that distinguish core genomes from the strain-variable regions.
Analyses of bacterial pan genomes reveal a high level of genetic diversity, even within a short evolutionary scale [1,2]. This raises the question of how bacterial chromosomes remain organized yet allow recombination events to occur. Noncoding functional DNA motifs have been implicated in bacterial genome maintenance, although their identification is rare and limited to studies in very few organisms. Where examined, they tend to be species or genus specific. Some examples are the highly frequent DNA uptake sequences, involved in discriminating self from foreign entering DNA during competence in Haemophilus influenzae and Neisseria meningitidis [3], the crossover hotspot instigator (Chi), first identified in E. coli, involved in recombinational repair [4,5], and a recently characterized chromosome dimer resolution motif named KOPS (FtsK orienting polar sequences) in E. coli [6,7]. Where studied, motifs with a biological function appear to occur nonrandomly in the genome. However, prediction of genomic motifs with biological function based on their distribution remains a challenging question and, to our knowledge, few approaches exist to answer it. Most currently developed motif-finding methods are targeted towards discovery of sequences involved in gene expression, which presupposes a constrained position of motifs with respect to genes. The approaches employed are not suitable for identification of genome organization motifs, which tend to be scattered around the chromosome. In this work, we formulate the following hypotheses upon which we devise a strategy to predict such motifs: (i) Natural selection of such motifs guarantees their overall distribution, rather than their occurrence at specific positions [8]. Prediction criteria for such sequences should therefore be based on motif distribution properties on complete genomes. (ii) Functional motifs should be conserved across strains of the same species, which can be assessed through comparative genome analysis. As recently shown, bacterial genomes can be segmented into a core genome (“backbone”) conserved among all strains of the same species, and strain-variable regions (“loops”) [1,9]. Motifs related to general cellular processes would be expected to be enriched on the backbone. A predictive approach for motif identification was undertaken, and the parameters for this approach were based on a well-documented bacterial motif, the Chi site, an essential component of double-strand break repair [4]. Chi was first identified as the sequence 5′-GCTGGTGG-3′ in E. coli [5], where it is a key modulator of enzymatic activities of the RecBCD complex [10,11]. Chi orientation-dependent recognition by RecBCD [12] leads to repair of genomic breaks incurred during DNA replication [13]. Most sequenced bacterial genomes encode functional RecBCD analogs [14,15]. However, Chi motifs are not conserved between species and only a few have been identified [16]. We successfully applied this approach to identify two sequences having Chi activity, one in S. aureus and another in three streptococcal species.
A set of criteria was generated to describe Chi distribution characteristics in the E. coli genome. A backbone/loop segmentation of the E. coli genome was performed by complete genome alignment of commensal K-12 [17], enterohemorragic O157: H7Sakai [1], and uropathogenic CFT073 [18] strains. The backbone is 3. 7 Mb long and each strain carries several hundred loops (Table 1; data accessible on the MOSAIC database Web site http: //genome. jouy. inra. fr/mosaic/). Chi motif frequency and overrepresentation were assessed separately on backbone and loop segments (the loop segment correspond to all loops concatenated). Chi is highly frequent on the backbone (its frequency is 1 every 6 kb, noted 1/6kb), but not on loops (Table 1). Statistical significance of these frequencies was checked by comparing observed and expected Chi frequencies based on composition in mono- to hepta-nucleotides of the backbone or of the loops. This was done by calculating an overrepresentation p-value based on a Gaussian approximation of short motif counts under a Markov model of order 6 [19]. On the backbone, Chi is the most overrepresented octamer among 65,536 possible sequences. In contrast, Chi occurrence on E. coli loops is not exceptional and is explained by loop DNA composition (Table 1). Moreover, comparison of all octamers distributed on the complete genome versus the backbone showed that other octamers appeared to be more overrepresented than Chi on the complete genome (its overrepresentation rank is 5 compared to a highest overrepresentation rank of 1). These were filtered out when analyses were restricted to the backbone (see motifs in red circle, Figure S1). This suggests that the analysis of backbone helps reduce the “noise” produced by motifs that are overrepresented due to the loops, and thus eliminates the candidates that are unlikely to have Chi activity. Interestingly, among these motifs, we identified octamers that are part of the BIME elements (a family of repeated elements that is significantly enriched on the loops [9]). This result supports the assumption that restricting the motif search to backbone facilitates identification of biologically important genomic motifs such as Chi. In E. coli, 75% of Chi sites are skewed towards the replicative leading strand [16,20,21], in keeping with their function in stimulating double-strand break repair upon replication fork collapse [13,22]. As the E. coli genome shows a strong GC skew, and Chi is G-rich, this bias might simply reflect the underlying GC skew. This is not the case with Chi skew; when compared to that expected from mono-nucleotide composition, it was found to be significant on the E. coli backbone, beyond what might be expected due to leading-strand bias (Figure 1). The above results validate our initial hypotheses concerning motif distribution for the E. coli Chi site. We therefore used its distribution properties as prediction criteria for Chi identification by an in silico approach in species where its sequence was unknown. We first chose the human pathogen S. aureus, for which numerous available genome sequences reveal a high level of intraspecies diversity [23]. Identification of Chi and characterization of its properties could provide insight into genome plasticity in this problematic pathogen. The S. aureus backbone was defined from an alignment of six S. aureus genomes (Mu50, Mw2, N315, COL, MRSA252, and MSSA476). The 2. 44-Mb backbone corresponds to 86 % of the mean genome length. It was used to analyze oligomer distribution on the replicative leading strand. We looked for motifs that were (i) significantly overrepresented; (ii) frequent, with average frequencies higher than 1 in 15 kb; and (iii) skewed (≥60%) with a significant score. Chi sites previously identified in different bacteria ranged in length from 8 to 5 nucleotides [24–26]. We started by analyzing octamers, as longer motifs can be more readily reduced to find the minimal active motif than the converse. None of the overrepresented and skewed octamers were frequent enough to be retained as potential candidates. We thus focused on heptamers (Figure 2A). Four motifs were strikingly overrepresented, of which motif C (5′-GAAGCGG-3′) and D (5′-GAATTAG-3′) matched the skew score criteria. As motifs A (5′-GAAAATG-3′) and B (5′-GGATTAG-3′) had skew scores only slightly lower than the threshold (1. 13 and 1. 54, respectively) they were also retained as potential candidates. The selected motifs were tested for S. aureus Chi activity using a biological screening assay previously used to identify Chi in Lactococcus lactis, H. influenzae, and Bacillus subtilis [24–26]. Briefly, a rolling circle (RC) plasmid may generate a double-strand DNA extremity during replication, which provides an entry point for RecBCD-like enzymes [27]. This extremity is degraded unless the plasmid contains a correctly oriented Chi site, in which case degradation is aborted, and high molecular weight multimers (HMW) accumulate. Motifs A and D were present in both orientations in the initial RC plasmid, which does not accumulate HMW in S. aureus (Figure 2B, lane 1), thus ruling them out as S. aureus Chi site candidates. Motif B was also ruled out, as it did not provoke HMW when cloned in the RC plasmid (unpublished data). However, insertion of an oligonucleotide containing motif C into the RC plasmid vector resulted in a strong HMW signal (Figure 2B, lane 2). The RC plasmid containing the inverse orientation of motif C did not generate HMW (Figure 2B, lane 3), in keeping with the orientation dependency of the E. coli Chi [12]. Chi sites identified in other bacteria vary in length and degeneracy [24–26]. It was thus possible that the S. aureus Chi motif necessitated only a subsequence of the identified seven-nucleotide motif. Substitutions at any of the seven defined nucleotide positions tested negative for HMW when present on the RC plasmid. Specifically, alterations in the first or last nucleotide of the motif were also negative (Table S3; Figure 2B, lanes 4–6). We conclude that the seven-nucleotide motif 5′-GAAGCGG-3′ is necessary and sufficient to confer Chi activity in S. aureus. The S. aureus Chi site is very frequent and overrepresented on the backbone (1/9 kb; p-value 10−28, rank 5), and 88% of Chi motifs are on the leading strand. As in E. coli, overrepresentation of the S. aureus Chi site did not apply to loops (Table 2). Chi site occurrence in S. aureus loops ranged from 1/15 kb to 1/20 kb, and in only Mu50 loops were they slightly overrepresented (p-value 8 × 10−6, rank 42), although levels were below our threshold values. As observed in E. coli, comparison of heptamer distribution on strain N315 complete genome versus backbone shows that some motifs appear overrepresented on the complete genome mostly due to their presence on loops, and that these motifs are filtered out by analysis on the backbone (motifs in red circle, Figure S2). The staphylococcal Chi sequence comprises a nonactive submotif, which was previously identified as the five-nucleotide Chi site of B. subtilis, 5′-AGCGG-3′ [26] (Table S3). Analyses on the B. subtilis complete genome show that its Chi motif is very frequent (1/0. 6 kb) and significantly overrepresented (p-value 6 × 10−19, rank 54) but does not fully conform to criteria described for Chi in E. coli and S. aureus, with respect to skew significance (its skew is only 58% and is not statistically significant). We considered the possibility that distribution of the S. aureus Chi site on the B. subtilis genome would meet the criteria we described for Chi. This is not the case; the motif, although frequent (1/4. 4 kb), is not overrepresented (p-value not significant, rank 2,959). Analysis of two other longer motifs (5′-AAGCGGC-3′ and 5′-AGCGGCGC-3′), reported in [26] as having very high Chi activity, shows that they are likewise not overrepresented (respective ranks 11,872 and 4,258, both with nonsignificant p-values). A more thorough analysis will require additional B. subtilis genome sequences to extract backbone DNA, but both the shortness of the sequence and its genome features in this species leads us to suggest that Chi properties may not be universal, and that Chi evolution in this bacterial branch may undergo different selective pressures leading to a role for Chi on both DNA strands. The generality of the Chi prediction method was challenged by examination of four species of the Streptococcus genus: pathogens Streptococcus pyogenes and Streptococcus pneumoniae, the opportunist pathogen Streptococcus agalactiae, and the nonpathogen Streptococcus thermophilus. These species belong to the same family as the nonpathogenic food bacterium, L. lactis. A backbone was constructed for each of the species (resulting from an alignment of two to six genomes; see Table 3 for details) and heptamer distribution was analyzed. The best candidate fulfilling the prediction criteria is the same in each of the four species (Table 3), and corresponds exactly to the L. lactis Chi site [24]. This prediction was confirmed by experimental validation in S. agalactiae, S. thermophilus, and S. pneumoniae (Figure S3; Table S4). As observed in E. coli and S. aureus, Chi is mostly not overrepresented on loops of streptococci (Table 3). These results indicate that five species of the Streptococcaceae family seem to share the Chi site independent of their pathogenicity status, which leads us to speculate that DNA protection by Chi evolved prior to functional differentiation of these bacteria. More generally, a comparison of all known Chi motifs in the context of the phylogenetic relationships of the corresponding species (Figure 3) indicates at least partial Chi conservation among closely related bacteria, regardless of their niche or pathogenic status. Chi has been considered as a key element in preventing chromosomal fork collapse during replication [13], thus ensuring faithful DNA break repair. The observed enrichment of Chi on backbone DNA suggests that Chi plays an important role in maintaining integrity of backbone regions that presumably encode essential functions. The much lower occurrence of Chi motifs on loops (of which the largest are up to 100 kb), likely reflects their exogenous origin. This might result in less efficient repair and hence lower stability of these regions. Conversely, moderate Chi overrepresentation in certain loop sets (e. g. , in S. aureus strain Mu50) might suggest that these loops are older and have persisted by conferring a selective advantage to the bacterium. Thus, Chi enrichment on backbone, which is common to all species we examined, probably reflects the evolutionary balance between genome stability and diversification through horizontal transfer. The approach reported here to predict genome organization motifs is based on a combination of comparative genomics and statistical analysis of motif distribution properties. It may prove useful in the discovery of other genomic motifs with known or novel functions via their specific characteristics in the complete genome, or in defined genomic regions. In addition to Chi, analysis of the S. aureus backbone also identified that heptamer motifs B (5′-GGATTAG-3′) and D (5′-GAATTAG-3′) were overrepresented on the backbone and skewed. Like Chi, these motifs are essentially not remarkable on the loops (Table S5). Interestingly, motif B is also highly overrepresented and skewed in two species related to S. aureus: Bacillus cereus and Bacillus anthracis. These results suggest that this motif (or the degenerated motif 5′-G (G/A) ATTAG-3′) might have a single biological function in several related species, which remains to be discovered. We recently used statistical analysis to characterize the KOPS family of octamers in E. coli, which orient the chromosome for segregation during cell division [6]. Like Chi, KOPS are overrepresented, frequent, and skewed on the backbone (Figure 1). Characteristics of the non-Chi motif identified here make it a tempting potential candidate to test as the KOPS homolog in S. aureus. Using the strategy presented here, KOPS and other biologically characterized repeated sequences could be modeled to discover motifs with similar roles in very distant bacteria.
Sequence data. We recently described a method for genome segmentation into backbone and loops based on multiple genome alignments [9]. In E. coli, the backbone/loops segmentation resulted from comparison of strains K-12, O157: H7 Sakai, and CFT073. Note that these analyses could not be performed in other species where Chi is known, due to the absence of appropriate data that would allow construction of a reliable backbone. For S. aureus segmentation, we used strains Mu50, Mw2, N315, COL, MRSA252, and MSSA476. Data are available at http: //genome. jouy. inra. fr/mosaic/. Leading strands were defined as the DNA strand reported in Genbank files downstream of the replication origin up to the terminus and the reverse complement strand from the replication terminus to the origin. In E. coli, the position of the replication origin is annotated for the K-12 strain, and was determined in the other strains by a sequence similarity search. The origin position was previously mapped and assigned position 1 in all the other species. The terminus position was chosen as the first nucleotide of the chromosome dimer resolution dif site as described for E. coli and B. subtilis [28,29]. The S. aureus, B. cereus and, B. anthracis dif-like sites were identified as sequences similar to the B. subtilis dif site 5′-ACTTCCTAGAATATATATTATGTAAACT-3′, (allowing two mismatches, one insertion, and one deletion), by performing a PATSCAN analysis [30] on the Micado database (http: //genome. jouy. inra. fr/micado). For the four streptococci, we used the same strategy using the streptococcus-specific dif site identified by P. Le Bourgeois [31]. The terminus positions thus determined coincided with those reported by other methods [8,32,33]. Motif overrepresentation score. Motif count analyses were performed on the leading strand of the backbone common to the aligned genomes for a given species, and on the leading strand of all loops of each strain. To assess overrepresentation of short motifs, the observed count of each motif was compared to the count expected in random sequences showing the same oligonucleotide composition. The significance of the difference between the counts was evaluated by calculating the associated p-value, which is the probability that the count of a given motif “u” in a random sequence (under a chosen stationary Markov model, see below) is greater than the observed count for this motif. The p-value is obtained using a Gaussian approximation of motif counts [34], which has been shown to be reliable for short motifs in comparatively long sequences [35]. For a motif “u, ” the p-value denoted pF (u) is given by: where X is distributed according to the standard Gaussian distribution N (0,1), Nobs (u) the observed count of u, EN (u) is the estimated expected count of u in the chosen model, and γ is an explicit normalizing factor [34]. Note that [Nobs (u) − EN (u) ]/γ can be directly interpreted as an overrepresentation score: the higher its positive value, the more overrepresented is the motif u. The random sequence models are Markov chain models that take into account the nucleotide monomer, dimer, trimer, etc. , (m + 1) -mer composition of the sequence, depending on the order m that is used; When examining Chi among E. coli octamers we chose the maximal model M6 (which takes into account the monomer to heptamer nucleotide composition). Analyses on heptamers in S. aureus and streptococci were performed using the M5 model (based on monomer to hexamer composition). Parameters of those models were estimated separately on the backbone and the loops to take into account the difference in oligomer composition of loops compared to backbone. We checked that in all cases the sequences were long enough to reliably estimate the parameters of the model. Note that p-values obtained on the backbone and loops cannot be directly compared because of differences in the sequence lengths [36]. We therefore present the overrepresentation ranks (in which the overrepresentation score of a given motif is compared to all possible motifs of the same length), which are directly comparable. Calculations were made using R' MES software (http: //genome. jouy. inra. fr/ssb/rmes/), which provides scores for all motifs of a given length, and ranks them according to their overrepresentation score. The Bonferroni correction was used to choose the significance threshold; for a given significance level (10−3 in our case), the motif p-values have to be divided by the number of tests (16,384 for heptamers and 65,536 for octamers, respectively). Thus, a p-value is significant when it is ≤6 × 10−8 (corresponding score 5. 2) for a given heptamer, and ≤1. 5 × 10−8 (corresponding score 5. 6) for a given octamer. Motif skew score. The significance of the skew for a motif “u” was also determined by calculating a p-value, namely the probability that its skew in a random sequence is greater than the observed skew Sobs. Since the skew is the ratio between the counts of u and that of ū (ū is the reverse complement of u), the p-value denoted pS (u) was obtained using a Gaussian approximation of motif counts as above. More precisely, we have: where X is distributed according to the standard Gaussian distribution N (0,1), We chose a Markov model of order 0 to take into account the richness in G of the leading strand, which is due to G/C skew. Under such a model, the expectation, variance, and covariance of motif counts are easily derived [34]. As above, the quantity –μ/σ can be directly interpreted as a score of skew exceptionality: The higher the positive value, the higher the exceptionality of the skew. These calculations can also be obtained with the third version of the R' MES software (http: //genome. jouy. inra. fr/ssb/rmes/). Because skew is only a secondary criterion for Chi activity, we used a much less stringent significance level (p < 0. 05, corresponding score 1. 6) for both heptamers and octamers. Criteria for motif selection. We selected the most overrepresented motifs on the backbone leading strand that were also frequent (frequency higher than 1/15kb), skewed (skew higher than 60%), and with a skew p-value lower than 0. 05. This allowed us to construct a list of potential candidates ordered by decreasing overrepresentation p-value. The strains and plasmids used in this study are listed in Table S1. S. aureus strain RN4220 (an avirulent strain that accepts foreign DNA by transformation of competent cells) was grown in Brain Heart Infusion medium with aeration at 37 °C. S. pneumoniae and S. agalactiae were grown in Todd Hewitt yeast extract at 37 °C. S. thermophilus strains were grown in M17 medium supplemented with 1% lactose at 42 °C. E. coli strain TG1 was used for plasmid constructions. Chloramphenicol (10 μg/ml in S. aureus, 20 μg/ml in E. coli, 3 μg/ml in S. thermophilus, 5 μg/ml in S. pneumoniae, and 7 μg/ml in S. agalactiae) was used for plasmid selection. Plasmid DNA preparation, PCR amplifications, and DNA modifications were carried out according to suppliers' instructions. S. aureus cell suspensions were lysed using 100 μg/ml lysostaphin (Sigma) for 30 min at 37 °C as described [37]. DNA transformation was performed by electroporation [38]. Oligonucleotide cloning. Oligonucleotides (Table S2) were inserted at the HincII/EcoRI sites of the multiple cloning site of pRC2 (Table S1) using standard methods. Plasmid constructions were carried out in E. coli and confirmed by DNA sequencing. The pACYC184 replication origin present on pRC2 was then deleted by NciI/PvuII digest, giving rise to a pRC plasmid carrying the designed oligonucleotide. The resulting plasmid was then transferred to S. aureus. Detection of Chi activity by HMW accumulation. The RC plasmid used for Chi activity detection in S. aureus is based on the pVS41 (pC194) replicon [39]. It was derived from pRC2 deleted for its pACYC184 replication origin by NciI/PvuII digest, Klenow fill-in, and self ligation, giving rise to pRC. pRC carrying oligonucleotides containing putative Chi motifs (see above for construction) was used for HMW detection. Total DNA was extracted from strains carrying different constructions and HMW detected by Southern blot hybridization as described [40]. To identify mutations leading to abolition of Chi activity, we used pRC that carried oligonucleotides differing from Chi by a single point mutation (see above for construction and Table S2). We also made use of pRC constructs containing random 1–2-kb NciI/PvuII DNA fragments derived from Pseudomonas aeruginosa PAO1 that were screened for HMW accumulation; DNA inserts of plasmids giving no HMW signal were sequenced and checked against the complete PAO1 genome sequence. Sequences were screened for motifs that differed by single mutations from the putative Chi site. The RC plasmid used for Chi activity detection in S. agalactiae and S. thermophilus is pRC2. Its derivative containing an oligonucleotide with the putative streptococcal Chi site (5′-GCGCGTG-3′) is called pRC2Chi2 [40]. HMW accumulation was detected by Southern blot hybridization as above. Detection of Chi activity by measurement of transformation efficiency. Attempts to visualize HMW in S. pneumoniae by Southern blot hybridization were not reproducible, despite numerous efforts and modifications. We therefore used a test previously described for Chi identification in B. subtilis [26]. Briefly, in B. subtilis, multimeric plasmid molecules transform cells better than monomeric or dimeric species. Therefore, HMWs, being composed of multiple repeats of plasmid molecules, confer increased transformation ability to plasmid extracts. We monitored the amount of HMW by measuring transformation efficiencies in B. subtilis (strain 168) of total DNA extracts from different S. pneumoniae strains relative to that of a wild-type strain carrying the pRC2 vector. Total DNA extracts were quantified by spectrophotometric dosage at 260 nm and calibration against known quantities of λ BstEII on agarose gels. Total DNA extracts were performed as described and standard transformation techniques were used [26].
The National Center for Biotechnology Information (NCBI) Genbank (http: //www. ncbi. nlm. nih. gov/sites/gquery) accession number of the genomes used in this study are as follows: E. coli strains K-12, O157: H7 Sakai, and CFT073: U000096, BA000007, and AE14075, respectively; S. aureus strains Mu50, Mw2, N315, COL, MRSA252, and MSSA476: BA000017, BA000033, BA000018, CP000046, BX571856, and BX571857, respectively; S. agalactiae strains NEM316, A909, and 2603VR: AL732656, CP000114, and AE009948, respectively; S. thermophilus strains CNRZ1066 and LMG18311: CP000024 and CP000023, respectively; S. pneumoniae strains R6 and TIGR4: AE007317 and AE005672, respectively; S. pyogenes strains M1GAS, MGAS10394, MGAS315, MGAS5005, MGAS6180, and MGAS8232: AE004092, CP000003, AE014074, CP000017, CP000056, and AE009949, respectively. | Availability of bacterial “pan genomes, ” based on multiple genome sequences of a given species, has revealed the existence of core genomes, but also of high levels of variable, nonconserved regions. The nature of bacterial strategies that assure genome organization while permitting biodiversity remains an intriguing question. A first clue in addressing this question comes from growing evidence for the existence of noncoding functional DNA motifs with specific distribution properties along the chromosome, which are implicated in DNA integrity. In this work, we addressed the challenging problem of predicting such motifs from pan-genome information. We analyzed characteristics of the “crossover hotspot instigator” motif, Chi, which, in E. coli, is an 8-base-pair sequence involved in chromosome maintenance. Our results show that Chi has specific distribution properties that are restricted to core genomes common to related E. coli strains. Using statistical modeling combined with comparative genomics, we then predicted the identity of Chi in core genomes of S. aureus and several streptococci, and confirmed them in vivo. The strategy developed in this study may be extended to reveal and characterize novel functional motifs, and should be instrumental in analyzing genome biodiversity mechanisms. | Abstract
Introduction
Results/Discussion
Materials and Methods
Supporting Information | microbiology
computational biology
molecular biology
genetics and genomics
eubacteria | 2007 | Identification of DNA Motifs Implicated in Maintenance of Bacterial Core Genomes by Predictive Modeling | 7,235 | 276 |
Dengue is endemic in the Western Pacific and Oceania and the region reports more than 200,000 cases annually. Outbreaks of dengue and severe dengue occur regularly and movement of virus throughout the region has been reported. Disease surveillance systems, however, in many areas are not fully established and dengue incidence is underreported. Dengue epidemiology is likely least understood in Papua New Guinea (PNG), where the prototype DENV-2 strain New Guinea C was first isolated by Sabin in 1944 but where routine surveillance is not undertaken and little incidence and prevalence data is available. Serum samples from individuals with recent acute febrile illness or with non-febrile conditions collected between 2007–2010 were tested for anti-DENV neutralizing antibody. Responses were predominantly multitypic and seroprevalence increased with age, a pattern indicative of endemic dengue. DENV-1, DENV-2 and DENV-3 genomes were detected by RT-PCR within a nine-month period and in several instances, two serotypes were identified in individuals sampled within a period of 10 days. Phylogenetic analysis of whole genome sequences identified a DENV-3 Genotype 1 lineage which had evolved on the northern coast of PNG which was likely exported to the western Pacific five years later, in addition to a DENV-2 Cosmopolitan Genotype lineage which had previously circulated in the region. We show that dengue is hyperendemic in PNG and identify an endemic, locally evolved lineage of DENV-3 that was associated with an outbreak of severe dengue in Pacific countries in subsequent years, although severe disease was not identified in PNG. Additional studies need to be undertaken to understand dengue epidemiology and burden of disease in PNG.
The dengue viruses (DENV) are the most important arboviral pathogens of humans causing an estimated 390 million infections annually, of which approximately one quarter are symptomatic [1]. Infection with DENV causes a spectrum of clinical outcomes ranging from self-limiting febrile illness (dengue fever, DF) to potentially fatal severe dengue, characterized by plasma leakage, thrombocytopenia, and hypovolemic shock. Dengue is endemic in more than 100 tropical and subtropical countries, where the principal mosquito vectors Aedes aegypti and Aedes albopictus are found [2,3]. DENV is a single-stranded positive-sense RNA virus of the Flaviviridae family. Like other RNA viruses, DENV displays considerable genetic diversity and is grouped into four antigenically distinct serotypes (DENV-1-DENV-4) which may be distinguished on the basis of serum neutralization tests. The four serotypes are more precisely classified, using phylogenetic approaches, into distinct genotypes which have been defined as clusters with nucleotide sequence divergence of not more than 6% [4]; lineages within the genotypes may represent strains with similar geographic origins [5]. Certain genotypes have been associated with more [6,7] or less [8] virulent disease, and there is some evidence for humoral and cellular immune selection focused on viral B- and T-cell epitopes [9,10]. DENV genetic diversity thus appears to impact host mechanisms shown to mediate pathogenesis [11] and ultimately, disease severity. Dengue was first identified in Papua New Guinea (PNG) when Sabin isolated the prototype DENV-2New Guinea C strain from febrile soldiers deployed on the northern coast of New Guinea in 1944 [12]. A DENV strain with similar biologic features as the prototype DENV-1Hawaii that Sabin had recently isolated from febrile soldiers in Hawaii, was also isolated from soldiers in the same area of New Guinea in 1944 [13], suggesting that at least two serotypes may have circulated in the north of PNG in the 1940s. More recently, DENV-1-3 genetic data derived from viremic travellers returning to northern Australia from PNG between 1999–2010 [14] indicate that dengue is endemic in PNG, supporting an earlier serological study demonstrating a seroprevalence rate of 8% among patients presenting to clinics in Madang with acute febrile illness [15]. Despite the likely transmission of multiple DENV serotypes and the potential for introduction of DENV from endemic neighbouring countries which experience large-scale epidemics of severe dengue [16], little is known about the epidemiology and transmission dynamics of dengue in PNG, where dengue surveillance is not undertaken, individuals with acute febrile illness are not routinely tested for DENV infection, and where severe disease is rarely reported. We sought to determine DENV serotype and genotype prevalence in local populations presenting with febrile illness, or with a range of non-febrile conditions. We sequenced whole genomes of DENV and conducted a phylogenetic analysis to determine the evolutionary origin of PNG DENV. In addition we determined anti-DENV-1-4 neutralizing antibody profiles in adults and children in order to assess prevalence and serotype diversity.
Samples analysed in this study were collected from Madang, on the northern coast of PNG, and from Lihir Island in New Ireland Province. Madang is a town of about 30,000 people with a sea port that is a major hub for domestic and international shipping, and an airport where domestic flights from throughout PNG arrive several times each day. Lihir Island is 800 kilometres northeast of Madang in the Bismarck Archipelago, in the western Pacific Ocean. The island’s population doubled to more than 12,000 people after establishment of a gold mine in 1997 and although many residents still live a predominantly traditional subsistence lifestyle, in recent years there has been an influx of PNG-national and expatriate mine workers and development of an international airport and sea port. Madang sera were collected from febrile patients presenting to the outpatient clinic of Yagaum rural hospital or to Jomba town clinic from September 2007 through June 2008, and who were enrolled in a malaria study [15]. Sera excluded for malaria antigens were tested for anti-DENV IgG and IgM, and NS1 antigen; 8% (46/578) were identified as probable acute DENV infection ie. NS1 antigen-positive and/or anti-DENV IgM-positive. A total of 55 acute phase sera (46 sera identified serologically as probable acute DENV infection plus 9 additional febrile sera that were not tested for DENV), were assessed in the present study for the presence of DENV by DENV E gene RT-PCR and virus isolation was attempted on RT-PCR positive samples. Whole genomes of isolated viruses were sequenced using Illumina. Convalescent sera from 119 patients excluded for acute DENV infection and who presented for recollection were collected an average of 29. 6 days (range 5–159 days) after the first patient visit and tested for the presence of anti-DENV neutralizing antibody (NAb) to all four serotypes simultaneously using a microneutralization assay optimized for small sample volumes [17]. Lihir sera (55 in total) were collected from patients presenting to the outpatient clinic of Lihir rural health centre from May through November 2010 during pre-employment medical visits, at antenatal screening visits, or from patients presenting for a range of conditions including joint pain, diabetes and fever. The sera were also assessed for DENV genomes (11/55 sera from febrile patients) and for anti-DENV NAb (44/55 sea from patients with non-febrile conditions). Patient data are summarized in Table 1. Ethics approval for this study was granted by the Medical Research Advisory Committee, Ministry of Health, Government of Papua New Guinea (2010) and the Human Research Ethics Committee, University of Western Australia (2010). All data analysed were anonymized. DENV was isolated from serum by inoculation onto monolayers of Vero cells [5]. Briefly, 100μl of acute phase serum was inoculated onto a Vero cell monolayer in minimal media in a 2. 5 ml culture tube, and incubated overnight. On the following day the inoculum was removed and 3ml of DMEM with 2% FBS (supplemented with L-glutamine and antibiotics) was added to the cells, and the culture was maintained at 37°C with 5% CO2 for 7 days or until cytopathic effect (CPE) was observed; for most samples a blind passage into a second 2. 5 ml culture tube was required in sorder to isolate virus. Successful virus isolation was identified by NS1 antigen ELISA (Platelia Dengue NS1 Antigen ELISA; Bio-Rad, Australia). Viral RNA was extracted from 140 μl of culture supernatant using QIAmp viral RNA Mini kits (Qiagen), according to the manufacturer’s instructions. cDNA was synthesized from extracted RNA using SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen) as per the manufacturer’s instructions. DENV serotype was identified by RT-PCR using serotype-specific primers [5], and the LongRange PCR Kit (Qiagen) (thermocycling conditions are available on request). A serum microneutralization (MN) assay was used to measure serum anti-DENV antibodies [17]. This approach was selected to allow simultaneous assessment of antibody to all four DENV serotypes in samples with limited volumes. Standard anti-DENV-1-4 sera NIBSC 05/248 (National Institute for Biological Standards and Control [NIBSC], Potter’s Bar, Hertfordshire, United Kingdom) were assayed against the homologous DENV prototype strains DENV-1Hawaii2001; DENV-2NGC; DENV-3H-87 and DENV-4H-241 and consistently produced MN titres of 10–20 and thus, the cut-off value for a positive test result was a reciprocal serum dilution of 10. Subject results were summarized and presented as geometric mean titres (GMT). Cross-neutralization experiments in which Standard DENV-1-4 sera were each tested against heterologous prototype DENV consistently produced negative results. Standard anti-JE serum (NIBSC 02/182) and serum samples from individuals with diagnosed other flavivirus infection (JEV, MVEV, and KUNJV) were tested against DENV-1-4 and were always negative. All sequences have been deposited in GenBank and assigned accession numbers KY794785-KY794790.
Three circulating DENV serotypes: DENV-1, DENV-2 and DENV-3 were identified by full length E gene PCR of acute phase serum samples from 17 febrile patients sampled in Madang between 2007–2008 and Lihir in 2010 (Table 2). DENV-1, DENV-2, and DENV-3 infections were identified in adults and children from Madang town or from villages and rural settlements around the town. In several instances, two serotypes were identified in individuals sampled within a period of 10 days. Two DENV cases originated in Lihir in May and October 2010 –a 47 year-old male Australian traveller infected with DENV-2 and 32 year-old female resident of Lihir infected with DENV-3. These data clearly identify hyperendemic DENV transmission on the northern coast of PNG, in Madang, and on Lihir Island in the Bismarck Archipaleago. Whole genomes were sequenced from one DENV-2 and five DENV-3 isolates. Phylogenetic analysis of the five DENV-3 isolates indicated that the four from Madang group together to form a previously unidentified lineage within Genotype I (Fig 1). The Madang lineage included other DENV-3 from the region, including a group exported to Solomon Islands and Fiji and collected in 2013 [21] and 2014, respectively. The DENV-3 strain originating in Lihir in 2010 also grouped within Genotype I and clustered with a lineage formed by DENV-3 first found in travellers between PNG and northern Australia [14]. Collectively, these data identify a lineage of DENV-3 endemic to the northern coast of PNG, closely related to DENV-3 circulating in neighbouring Indonesia, and the subsequent introduction of this lineage into the western and south Pacific. The entire lineage is well supported and distinguished from other viruses within Genotype I by seven amino acid substitutions spanning several genes [prM, E, NS2A, NS3, NS5] (Table 3) and are indicated by the highlighted branch in Figure I. Several of these substitutions were non-conservative, involving the replacement of one encoded amino acid with another of very different properties. At least two of these substitutions were fixed through the process of positive selection according to at least one statistical test implemented in HyPhy [22] and the online version Data Monkey (http: //www. datamonkey. org/ [23,24]. The Lihir 2010 DENV-2 virus grouped with the Cosmopolitan genotype (Fig 2) and clustered with DENV-2 originating in Makassar, Indonesia in 2007 [25], and identified in travellers between PNG and northern Australia [14]. Cosmopolitan DENV-2 circulates widely in South and Southeast Asia, Africa, the Middle East, and northern Australia, and these data illustrate introduction of DENV-2 into the Western Pacific region from Southeast Asia. Overall dengue seroprevalence at the two study sites was 85. 3%. Assessment of convalescent sera from residents of Madang identified previous DENV infection in the great majority of samples tested (101/119 sera; 84. 9%), most of which showed multitypic NAb responses to all four DENV serotypes (Table 4). DENV-2 and DENV-3 GMTs were of greatest magnitude. Lihir sera showed a similar profile (Table 5) to that seen in Madang where the majority of sera were seropositive (38/44; 86. 4%), multitypic DENV-1-DENV-4 NAb responses predominated and a majority neutralized all four serotypes. In both locations, monotypic NAb responses were identified in 11% of sera and overall were directed against each of the four serotypes. Seroprevalence increased with age (Table 6, Fig 3), a pattern that is typically seen in dengue-endemic areas.
Although dengue was first described in Papua New Guinea more than 70 years ago when Sabin isolated the prototype DENV-2 strain New Guinea-C from US soldiers deployed along the northern PNG coast during the Second World War [12], the lack of reported case data and DENV transmission data since that time has meant that the distribution of dengue in PNG is not understood. In this study we identified infection with three DENV serotypes among febrile patients presenting to health clinics on the northern coast of PNG, in Madang, over a nine month period in 2007–2008, and infection with two serotypes within 6 months in 2010, in Lihir Island. These findings are consistent with hyperendemic DENV transmission between 2007–2010 and confirm that dengue is endemic in this country. Our understanding of DENV transmission in PNG has previously been largely limited to detection of DENV in febrile travellers to northern Australia [14]. The Madang DENV-3 Genotype 1 lineage evolved locally and circulated over the year the study was conducted. Our phylogenetic analysis showed that it was most closely related to DENV-3 detected 3. 5 years later in the Solomon Islands in January 2103 and then in the Fiji Islands in 2014, likely exported by travellers and having moved over distances of thousands of kilometres. PNG shares geographic borders with Indonesia, where dengue epidemics occur regularly and endemic, locally evolved DENV-3 lineages have been described [5,25]. The Madang lineage is most closely related to DENV-3 originating in Indonesia, as is the DENV-3 originating in Lihir Island which clustered with DENV-3 identified in travellers to northern Australia. Similarly, the DENV-2 virus identified in this study in a local resident of Lihir Island in 2010 is representative of strains known to circulate in the region, clustering with a highly similar strain originating in Makassar, Indonesia, in 2007; both viruses belong to a DENV-2 (Cosmopolitan genotype) clade also circulating in neighbouring Singapore in 2008 and identified in travellers to northern Australia in 2004 and 2006. Our results provide further evidence of DENV movement between endemic countries in the Asia Pacific region that has been described by ourselves and others [26–29], and also illustrate the value of sequence data as a means of understanding virus dispersal. PNG may be a source, or at least a place of transit, for dengue to enter the Pacific and further disseminate to other Pacific Island nations. DENV epidemic virulence has been linked to introduction and transmission of specific serotypes and lineages [6,7]. The DENV-3 lineage we identified to be circulating in Madang in 2007–2008 and which our analysis showed to have evolved locally was subsequently associated with hospitalizations and deaths when it was introduced into the Solomon Islands in 2013 [30], although severe disease was not identified among patients in Madang. This entire lineage was distinguished by several amino acid substitutions that may influence virus phenotypic characteristics and ultimately, disease outcome. Severe dengue has also been linked to genetic polymorphisms including HLA type [31]; PNG and the Solomon Islands both belong to the Melanesian subgroup of Pacific Islanders and although relatively few HLA data are available for this population group, certain allele frequencies are known to be shared. Studies to assess the immunopathogenesis of dengue in PNG would be informative and should be undertaken in future. Dengue seroprevalence was very high with an overall rate of 85. 3%, and increased with age. More than half of seropositive individuals from Madang were greater than 11 years of age and about 40% were older than 20 years. This age distribution reflects opportunities for multiple exposures over the lifetime of the individual in a setting where dengue is endemic or is regularly introduced and indeed, anti-DENV neutralising antibody profiles were predominantly multitypic. A majority of individuals in Madang and in Lihir demonstrated responses to two or more DENV serotypes and most sera neutralized all four DENV serotypes. Serum samples included in this analysis were from two main groups–convalescent patients excluded for acute DENV infection in Madang, and clinic attendees in Lihir (predominantly women attending antenatal clinics) therefore these neutralization data likely reflect multiple DENV infections in the years prior to sampling, and corroborate the genetic evidence for circulation of multiple serotypes. Studies in rural Haiti and Nicaragua [32,33] have shown that in endemic populations anti-DENV antibody prevalence increases with age and begins to plateau in adolescence reflecting long-term exposure to, and infection with, endemically circulating DENV. In the present study further support for endemic DENV transmission in PNG is indicated by the young age (median 2 years) of individuals in Madang with monotypic anti-DENV NAb responses to DENV-1, DENV-2, DENV-3 or DENV-4 whereas the majority of multitypic infections were in adults. Three individuals with monotypic NAb profiles were babies less than 1 year old and anti-DENV NAb were likely passively transferred maternal antibody. We do not know the age of the mothers but it is possible they were relatively young and/or had experienced a single DENV infection in the past, or that they had experienced multiple infections but that only antibodies to a single serotype were present at sufficiently high levels in the infant to be detectable in our assay. Hyperendemic DENV transmission is associated with symptomatic dengue infection and with greater incidence of severe dengue [3]. No febrile patients were diagnosed with severe dengue in Madang [15] or Lihir and indeed, severe dengue is rarely identified in PNG despite the hyperendemic transmission we have identified and which has likely been occurring for a significant period of time. In previous studies we demonstrated high seroprevalence and predominantly multitypic NAb responses to all four serotypes in sera collected in New Guinea between 1959–1963 [17] indicating DENV transmission in the decades prior to sampling. The reasons for the rarity of severe disease are unclear but may be related to poor recognition of dengue and the lack of routine dengue surveillance. It is also possible that undefined DENV-specific immune mechanisms may contribute to the apparent rarity of severe dengue disease. In summary we have identified hyperendemic dengue transmission in PNG in the period up to 2010, which has likely been occurring for many years. We also demonstrate circulation of DENV which has evolved locally, and shown by others to have been introduced into the western and south Pacific in subsequent years. The absence of regular sampling for DENV in PNG and the potential for misdiagnosis of febrile individuals has meant that the evidence consensus for dengue presence is low [2] and the true burden of disease is unknown. A dengue outbreak in Port Moresby was confirmed by the PNG Department of Health in 2016 [34] highlighting the need for additional studies to be undertaken to understand the epidemiology and impact of dengue in this country. An important aspect of this is to understand the origin and transmission patterns of PNG DENV, define endemic and introduced genotypes and lineages, and to characterize epidemic virulence associated with circulation of these viruses among the PNG population and within the Asia Pacific region. | Dengue virus (DENV) was first identified in Papua New Guinea (PNG) in 1944. Dengue is currently assumed to be an endemic disease in PNG although there is little incidence or prevalence data, and the evidence consensus for dengue presence is low. Routine surveillance is not undertaken and dengue is not a notifiable disease. Severe dengue is rarely identified by local clinicians and the reasons for this are unclear but may be related to poor recognition of dengue and a low index of suspicion, despite high incidence and prevalence rates in neighbouring countries. For example, Indonesia shares borders with PNG and regularly reports outbreaks of severe dengue and transmission of multiple DENV serotypes. DENV infection is identified in travellers from PNG however there are no data on locally circulating strains and how they may compare to viruses associated with severe dengue epidemics in other countries in the Asia Pacific region. We identified evidence for previous infection with all four DENV serotypes among people living on the northern coast of PNG, in Madang, and on Lihir Island in the Bismarck Archipelago off the northeastern coast. We also detected DENV-1, DENV-2, and DENV-3 virus in febrile patients, and we describe the first whole genome sequences of endemically circulating DENV since the prototype 1944 DENV-2New Guinea C strain was characterized. Of note, severe dengue was not diagnosed in any patient infected with these viruses in PNG although introduction of the PNG DENV-3 strain into the Solomon Islands five years later resulted in a large outbreak of severe dengue with hospitalizations and deaths in that country. Dengue epidemiology and burden of disease should be investigated in PNG. | Abstract
Introduction
Materials and methods
Results
Discussion | reverse transcriptase-polymerase chain reaction
geomorphology
taxonomy
dengue virus
medicine and health sciences
pathology and laboratory medicine
landforms
engineering and technology
pathogens
topography
geographical locations
microbiology
australia
rna extraction
viruses
phylogenetics
data management
rna viruses
phylogenetic analysis
molecular biology techniques
extraction techniques
islands
prototypes
research and analysis methods
computer and information sciences
artificial gene amplification and extension
technology development
medical microbiology
papua new guinea
microbial pathogens
evolutionary systematics
molecular biology
people and places
flaviviruses
polymerase chain reaction
oceania
viral pathogens
earth sciences
biology and life sciences
evolutionary biology
organisms | 2018 | Hyperendemic dengue transmission and identification of a locally evolved DENV-3 lineage, Papua New Guinea 2007-2010 | 4,855 | 382 |
Symbionts can have mutualistic effects that increase their host’s fitness and/or parasitic effects that reduce it. Which of these strategies evolves depends in part on the balance of their costs and benefits to the symbiont. We have examined these questions in Wolbachia, a vertically transmitted endosymbiont of insects that can provide protection against viral infection and/or parasitically manipulate its hosts’ reproduction. Across multiple symbiont strains we find that the parasitic phenotype of cytoplasmic incompatibility and antiviral protection are uncorrelated. Strong antiviral protection is associated with substantial reductions in other fitness-related traits, whereas no such trade-off was detected for cytoplasmic incompatibility. The reason for this difference is likely that antiviral protection requires high symbiont densities but cytoplasmic incompatibility does not. These results are important for the use of Wolbachia to block dengue virus transmission by mosquitoes, as natural selection to reduce these costs may lead to reduced symbiont density and the loss of antiviral protection.
Heritable symbionts are frequent in insects and their evolutionary success relies on various strategies. By sharing a common route of transmission with their host’s genes, they benefit from increasing host fitness. Consequently, numerous endosymbiotic bacteria evolved towards mutualism, for example by complementing their host diet [1,2], increasing tolerance to environmental stresses [3] or protecting against natural enemies [4–9]. However, because most of these heritable bacteria are maternally-transmitted, the evolutionary interests of host and symbiont are not perfectly aligned since only females transmit the symbiont. This has led to many symbionts evolving selfish strategies that consist of parasitic manipulation of their host’s reproduction by inducing female-biased sex-ratios or cytoplasmic incompatibility (CI) [10]. CI is a sperm modification that results in embryonic mortality in crosses between uninfected females and males harboring the symbiont, thus giving a competitive advantage to infected females that can rescue the sperm modification. Mutualism and reproductive manipulation are not mutually exclusive, and some symbionts display both [11]. However, the balance between the benefits and costs of these extended phenotypes to the symbiont’s fitness, as well as the genetic correlations between them, will determine which of these strategies is favoured by natural selection. Wolbachia, which are common maternally-transmitted bacterial symbionts of arthropods, can be both parasites and mutualists. Wolbachia has been shown to protect Drosophila and mosquitoes against several RNA viruses—including Dengue and Chikungunya viruses [7,9, 12–15]. Some strains also protect insects against filarial nematodes [16], Plasmodium parasites [12,17,18] and pathogenic bacteria [19]. Although it is unclear how important antiviral protection is in nature and whether it is under strong selection, some protective Wolbachia strains are able to invade host populations while inducing no other known phenotypes [20,21]. In addition, Wolbachia has the ability to spread rapidly through insect populations by parasitically manipulating reproduction, in particular by CI [22]. This combination of traits makes Wolbachia an attractive tool for blocking disease transmission by mosquitoes, as CI allows it to spread through vector populations while its antiviral effects can prevent them from transmitting arboviruses [23,24]. Levels of both antiviral protection and CI may evolve rapidly. During the 20th century in natural populations of D. melanogaster the Wolbachia strain wMelCS, which provides strong antiviral protection, was partially replaced by wMel [25,26], which provides weaker protection [27]. In North American populations of D. simulans, field and experimental data suggest that the strain wRi has evolved to produce weaker levels of CI within a few decades [28]. Efforts to use Wolbachia to block the transmission of viruses have focused largely on the mosquito Aedes aegypti, which is the primary vector of dengue virus. Wolbachia has been successfully introduced into two Australian populations of Aedes aegypti [29], and three years post-release it had reached a stable and high prevalence in the field despite having a negative effect on the fecundity of mosquitoes [30]. Both antiviral protection and levels of CI were maintained over time [30,31]. In the long-term, the presence of fitness costs is expected to select for both host genes and bacterial genes that reduce these costs [32]. In accordance with this prediction, the Wolbachia strain wRi evolved from reducing the fecundity of the flies to increasing it within two decades in North American populations of D. simulans [33]. It is possible that the evolution of lower costs could be achieved by a decrease in bacterial densities, as costly Wolbachia tend to have high bacterial densities [27,34,35]. Since a high Wolbachia density may be required for the expression of both antiviral protection [14,27,34,36–38] and CI [35,39–42], the evolution of reduced Wolbachia density might translate into a correlated decrease in the ability to block arbovirus transmission and invade insect populations. To investigate these questions, we used sixteen Wolbachia strains in a common host genetic background to measure the level of CI induced and effects on other fitness-related traits, and have tested for correlations between these traits and antiviral protection. Our results demonstrate that antiviral protection is independent of CI but that it is associated with reduction on other fitness components. Furthermore, this trade-off can be explained by the density of the bacteria in the somatic tissues of the insect. Overall, our study suggests that newly introduced Wolbachia infections may evolve towards weaker protection in the field.
Cytoplasmic incompatibility causes an excess of embryonic mortality in crosses between symbiont-infected males and uninfected females. Therefore, in order to measure levels of CI induced by different Wolbachia strains, we crossed infected males of each strain with uninfected females and counted the number of eggs that hatched (9,432 eggs from 380 females). There was a significant effect of Wolbachia (Deviance = 681. 81; df = 16; P < 0. 0001) with a clear division between 10 strains that induce CI and six that do not (Fig 1B). The strength of CI also varied among the 10 CI strains, ranging from just 0. 5% of the eggs hatching in incompatible crosses involving the wMel strain, to 38. 7% of the eggs hatching with wStv. We have previously shown that these strains provide varying levels of protection against the viruses DCV and FHV [14], and using this data we found that there was no correlation between CI and the antiviral effects of Wolbachia. This was the case regardless of which virus the flies are infected with or whether antiviral protection is measured in terms of increased survival (black line in Fig 2A and 2B) or reduced viral titre (black line in S1A and S1B Fig). This conclusion also holds if we only analyse the 10 strains that induce significant CI (red line in Fig 2A and 2B; S1A and S1B Fig). Since a decrease in hatch rate in incompatible crosses can be due not only to CI but also to an induced cost on male fertility, we also analysed the correlation between protection and levels of CI corrected for differences in male fertility (the hatch rates of infected females mated with infected males relative to hatch rates when mated with uninfected males). Similar to the uncorrected estimate, these corrected levels of CI did not show any significant correlation with antiviral protection, whether measured as survival after infection (Pearson’s correlation test: All strains: DCV: P = 0. 28 and FHV: P = 0. 86; CI-inducing strains: DCV: P = 0. 67 and FHV: P = 0. 71) or as viral titre (Pearson’s correlation test: All strains: DCV: P = 0. 58 and FHV: P = 0. 95; CI-inducing strains: DCV: P = 0. 87 and FHV: P = 0. 75). As Wolbachia is vertically transmitted, reductions in the survival or fecundity of Wolbachia-infected females will reduce the fitness of both the host and the symbiont. To estimate these costs, we measured egg hatch rates (in parallel to the CI crosses, 16,469 eggs from 555 females), early-life fecundity (280,260 eggs from 1,548 females) and female lifespan (913 females) of flies infected with the 16 different Wolbachia strains. We found significant variation in egg hatch rates between fly lines infected with different Wolbachia strains (Fig 1C; Deviance = 340,97; df = 16; P < 0. 0001). When the father was uninfected, four strains caused a significant reduction in hatch rates, with three of them resulting in less than 40% of the eggs hatching (Fig 1C, grey bars). Additionally, when both the mother and father were infected, there was a trend towards even lower hatch rates, with two more strains becoming significant (Fig 1C, blue bars). This suggests that male fertility is also being reduced by Wolbachia or that rescue of CI is not perfect for some of the strains (ie the modification of sperm in males that is required for CI still causes embryonic mortality when the egg is infected). Fecundity and lifespan are also affected by Wolbachia. For fecundity, two strains increased and two strains reduced the number of eggs laid (Deviance = 250. 55; df = 16; P < 0. 0001; Fig 1D). Wolbachia also affected female survival (Deviance = 52. 37; df = 16; P < 0. 0001), with five of the sixteen strains significantly shortening lifespan (Fig 1E). The strains that provide the greatest protection against viruses (measured as survival) tended to cause the greatest reductions in the other life-history traits of the flies. Hatch rates of Wolbachia-infected females were significantly reduced in flies carrying the symbionts providing the highest levels of protection against both DCV and FHV, whatever the Wolbachia-infection status of males (Fig 3A and 3B; S2A and S2B Fig). Because the tested traits are not phylogenetically independent, we reanalyzed these correlations using phylogenetic independent contrasts (see methods). The correlations between hatch rates and level of protection were robust to the phylogenetic non-independence of the data (S1 Table). Higher levels of antiviral protection were also associated with reduced male fertility (Fig 3C and 3D) and lower fecundity (Fig 3E and 3F), but these correlations were only significant in case of DCV. Phylogenetic independent contrasts analyses also showed that correlations with male fertility and fecundity were significant but it strongly depended on the branch length used in the linear models (S1 Table). No correlation with the level of protection and female lifespan was detected (S2C and S2D Fig; note the smaller sample sizes for this trait). Interestingly, wAu, which is a native strain of D. simulans, provides high antiviral protection yet induced little reduction in hatch rates or fecundity. If the antiviral effects of Wolbachia were measured as changes in viral titres rather than survival, most of the correlations became non-significant or marginally-significant, but the direction of the relationships remained the same, with low viral titres associated with stronger costs (S3A–S3J Fig). Again, costs induced by wAu on hatch rates were generally lower than expected by the correlations with viral titres. Similar to antiviral protection, we tested for correlations between levels of CI and other components of host fitness. There was no significant correlation between the level of CI and male fertility, female fecundity, lifespan or the hatch rate of eggs from crosses between Wolbachia-infected females and uninfected males (Fig 4A–4C; S4B Fig). In crosses where both parents were Wolbachia-infected, the level of CI was negatively correlated with hatch rates (S4A Fig). This was only the case when both CI inducing and non-CI inducing strains were analyzed, and it may reflect incomplete rescue of cytoplasmic incompatibility. We hypothesized that Wolbachia must infect the germline to induce CI and somatic tissues to provide antiviral protection, so differences in tissue tropism between symbiont strains may partly explain why they have different phenotypic effects on their hosts. To examine this, we measured Wolbachia density in somatic tissues (head and thorax of females), testes and freshly laid eggs (as a proxy for the female germline). There were large between-strain differences in density (Fig 5A–5C). For example, in somatic tissues the Wolbachia copy number varies over a 19-fold range. Furthermore, the strains have different tissue tropisms, with a significant strain-by-tissue interaction (Fig 5A–5C). The density in the testes and head + thorax tended to be tightly correlated (Pearson’s correlation test: r = 0. 89; P < 0. 0001), and frequently differed from the density in eggs (Pearson’s correlation test: head + thorax–eggs: r = 0. 63; P = 0. 01; testes–eggs: r = 0. 61; P = 0. 013). Variation in Wolbachia density can explain between-strain differences in antiviral protection but not differences in CI. Protection against DCV and FHV was positively correlated with Wolbachia density in head and thorax, whether measured as survival (Fig 6A and 6B) or viral titres (S5A and S5B Fig), even after removing potential phylogenetic effects (S1 Table). This holds when both the density in the soma and eggs are included as predictive variables: protection shows a significant partial correlation with density in the soma but not with density in the eggs (only marginally significant for FHV titre; S2 Table). On the contrary, there is no correlation between levels of CI and density in the somatic tissues (Fig 6C), in the testes or in the eggs (S6A–S6B Fig). The only exception to this was when only analyzing CI-inducing strains, levels of CI were positively correlated to the Wolbachia density in eggs (red line in S6B Fig; note eggs are uninfected in the CI cross). The negative effects of Wolbachia on host life-history traits are related to the symbiont density, with hatch rates, male fertility and fecundity all negatively correlated to the Wolbachia density in the somatic tissues (Fig 6D–6F) but not with the density in the eggs (Pearson’s correlation test: Hatch rate with uninfected father: P = 0. 08; hatch rate with infected father: P = 0. 06; male fertility: P = 0. 58; fecundity: P = 0. 27). The same conclusion holds when controlling for the Wolbachia phylogeny (S1 Table), although for male fertility and fecundity significance depends on the branch length used for the linear model. When these traits are analyzed with a multiple regression, they show significant partial correlations with density in the soma but not with density in the eggs (S2 Table). There was no correlation between female lifespan and Wolbachia density in any of the tissues (Pearson’s correlation test: head + thorax: P = 0. 73; testes: P = 0. 32; eggs: P = 0. 13).
Heritable bacterial symbionts have successfully colonized a wide range of arthropods by using a diversity of strategies ranging from mutualism to parasitism. Typically the evolution of these symbiont strategies has been considered in isolation, but this can be misleading if there are trade-offs between these traits and other components of host or symbiont fitness. Identifying these trade-offs is not only a prerequisite to understand the evolution of symbiosis, but will also inform the use of symbionts in applied programs. Using a set of Wolbachia strains that provide varying levels of protection against viral pathogens, we found that this mutualistic effect was independent of the ability to parasitically manipulate host reproduction. Antiviral protection relies on the bacteria reaching high densities in somatic tissues and is associated with strong reductions in several host life-history traits, while reproductive parasitism is not linked to symbiont density in somatic tissues and not costly to infected females. While some symbionts are mutualists that spread through populations by increasing host fitness and others are parasites that manipulate host reproduction, others simultaneously have both effects [11]. It is already well known that in Wolbachia antiviral protection and CI are highly genetically variable traits [14,27,37,44]. However, to our knowledge, our study is the first to assess both traits in a wide array of strains in a common host genetic background. We found no correlation between the expressions of these phenotypes, with four strains only providing protection, two strains only inducing CI, eight strains inducing both protection and CI, and two strains showing neither phenotype. Therefore, these traits have independent evolutionary trajectories. Some strains may also rely on alternative strategies to be maintained in populations, such as enhancing the host fecundity or other fitness components [45]. For instance, two of the tested strains in our study were associated with increased fecundity. Besides antiviral protection and reproductive manipulation, Wolbachia infections can induce fitness costs, with important life-history traits being affected such as lifespan, fecundity, egg viability or larval development and competitiveness [30,46–53]. In accordance with previous studies, we found Wolbachia-induced costs on several traits that should reduce both the fitness of the host and of Wolbachia. In some cases these costs could be very large–for example some strains result in the majority of infected eggs never hatching, suggesting that those strains might not be able to invade natural host populations. We found that antiviral protection trade-offs with egg hatch rates, female fecundity and male fertility. In many cases highly protective strains induced substantial reductions in these fitness components. Because Wolbachia relies on host reproduction for its transmission, these trade-offs will affect both the host and symbiont, as both partners benefit from antiviral protection and both will suffer from reduced female reproduction. Further evidence that antiviral protection is costly comes from a comparison of the two main Wolbachia genotypes in D. melanogaster populations, which showed that the genotype that provided the greatest antiviral protection also shortened the lifespan of infected flies (Chrostek et al. 2013). Similarly, when wAu is transferred into D. melanogaster it reaches high densities, provides strong protection against viruses and shortens the lifespan of flies (Chrostek et al. 2014). Interestingly, using a similar experimental design to ours, another study showed that high levels of protection conferred by the symbiont Hamiltonella defensa against parasitoids in aphids are associated with less costly symbiont strains contrary to what we found [54]. While the mechanisms of protection in Wolbachia remain to be elucidated, in H. defensa it is known that protection relies on the presence of a bacteriophage encoding a toxin [55,56]. It is likely that different mechanisms of protection lead to different trade-offs with host life-history traits. The reason that Wolbachia-mediated antiviral protection is so costly appears to be that it requires high symbiont densities. The density of Wolbachia in host tissues have been repeatedly shown to be involved in the ability of the bacteria to protect against viruses [12,14,27,36–38,57,58], and this was also the case in the present study with high protection being associated with higher densities in the somatic tissues of the flies. Using our sixteen Wolbachia strains we were able to test for a correlation between density and costs, and found that high densities of the bacteria in somatic tissues correlate with lower egg hatch rates, male fertility and fecundity. Harboring high loads of Wolbachia might be harming flies due to a re-allocation of resources from host to symbiont or pathological effects of the symbiont infection. Accordingly, wMelPop, a mutant strain that over-replicates causes a severe life-shortening effect [38,48,59] and other high density Wolbachia genotypes in D. melanogaster are associated with reduced lifespan [27,34]. The correlations between antiviral protection, costs on life-history traits and Wolbachia density remained when controlling for phylogenetic effects, which supports the hypothesis that there is a causal link between antiviral protection and costs that is mediated by symbiont density. Contrary to antiviral protection, we did not observe any trade-off between the expression of CI and the other host fitness components. The explanation for this is likely that CI levels were not correlated to the density of Wolbachia in somatic tissues (note that our sample size is limited if considering just CI inducing strains). CI is thought to be the result of a sperm modification causing improper segregation of the paternal chromosomes after fertilization of the egg [60]. Rather than the overall density of Wolbachia in the somatic tissues, it is the ability of the bacteria to specifically colonize sperm cysts that is thought to allow the expression of CI [39,42]. For this reason we investigated whether differences in tissue tropism between strains might affect whether they cause CI. While tissue tropism did vary, there was no correlation between density in testes and CI, but it may be that this is a poor proxy for the number of sperm cysts that are infected. However, we found that, among CI-inducing strains, levels of CI were positively correlated with the bacterial density in eggs (our measure of female germline density), similar to what was found in another study [40]. It is possible that higher density in the eggs might correlate with bacteria targeting the germ line in developing male embryos. Alternatively, the egg is the site of the rescue activity that prevents the expression of CI in Wolbachia-infected embryos [60], so strains inducing high levels of CI may have evolved towards higher density in the egg to overcome the effect of the sperm modification. Our findings have important implications regarding the evolution of Wolbachia symbioses, as trade-offs will act as a constraint on the evolution of mutualism (protection) but not reproductive parasitism (CI). Selection will act on both host and parasite genes to reduce the cost of Wolbachia infection, and alone this is likely to lead to the evolution of lower bacterial densities and therefore reduced antiviral protection. Thus, unless antiviral protection is sufficiently strongly selected for, it may reach lower levels or even disappear as the two partners coevolved towards less harmful Wolbachia infection. This prediction is supported by the partial replacement of the highly protective strain wMelCS by wMel, a lower density strain inducing lower protection, in populations of D. melanogaster [25–27]. Strikingly, in the pathogenic strain wMelPop, the symbiont density and the associated level of protection and costs on other life-history traits have been shown to evolve quickly, over a few host generations, suggesting that such changes may rapidly occur in nature [38]. As the most protective strains are very costly, they may only be favoured when there is very strong selection by viruses. Over the long term, selection may sometimes be able to break a trade-off [61] and lead to the evolution of Wolbachia strains that provide the benefits of antiviral protection but without the associated costs. Because we transferred most of the symbiont strains from other species into D. simulans, the control of the bacterial density and associated costs is expected to be inefficient due to a lack of coevolution between the two partners. This situation therefore reflects new associations that have arisen by horizontal transmission (as frequently occurs during the evolution of Wolbachia). We had one highly protective strain that naturally occurs in D. simulans, and this strain induced little cost on egg hatch rates despite showing rather high bacterial density and strong protection. This strain does not induce CI and yet shows rapid spread in natural populations [20,21] suggesting that protection might be the selective force driving the evolution of this strain. While this suggests that natural selection may be able to break the association between antiviral protection and cost, this may not be inevitable as naturally occurring protective Wolbachia strains in D. melanogaster still reduce the lifespan of flies [27]. Whether CI or antiviral protection is favored by selection will depend not only on the costs of these traits but also the strength of selection favouring the trait. Selection on the symbiont to evolve CI may often be very weak–there is no selection for the phenotype in males in panmictic populations [32,62], and its evolution relies on population structure generating local relatedness [63] [64] (see [65] for an alternative explanation). Our observation that CI is not associated with costly changes in the phenotype of infected females (the transmitting sex) means there may often be little selection on the symbiont to reduce the strength of CI, making it stable over evolutionary time even when population structure is weak. Finally, our results have implications for the control of vector-borne viral diseases by the introduction of Wolbachia into mosquito populations, as such efforts may fail if selection to reduce the cost of infection leads to reduced symbiont density and therefore the loss of antiviral protection [66]. This is even more likely if viruses cause little harm to its vector or are rare in the vector population, thus inducing little selective pressure on protection [23]. This is the case for the main target of these control efforts, dengue virus, which is thought to only decrease the fitness of mosquitoes by a few percent [67] and its prevalence in mosquito populations is low [68]. Therefore, the long-term maintenance of protection may rely on selection by the wider community of viruses favouring protection. The first releases of Wolbachia infected Aedes aegypti mosquitoes took place in 2011 [29], and one year later the Wolbachia strain still protected against dengue virus infection [31]. Only further monitoring over future years will determine whether this is truly an ‘evolution proof’ method of disease control.
All Wolbachia strains were in the D. simulans STCP line that was generated by six generations of sib matings [69]. Wolbachia was previously backcrossed or microinjected into the STCP line [14,44,69,70]. Flies were maintained on a cornmeal diet at 25°C, 12 hours light/dark and 70% relative humidity. To minimize inbreeding effects, before each experiment STCP females were crossed to males of a different Wolbachia-free isofemale line (14021–0251. 175, Dsim\wild-type, San Diego Drosophila Species Stock Center). Groups of 30 first instar F1 larvae were then transferred to new vials to ensure a constant larval density. Measurements of fitness traits were carried out on emerging F1 adults. Except for the fecundity measurements, F1 larvae were raised on a standard cornmeal diet (agar: 1%, dextrose: 8. 75%, maize: 8. 75%, yeast: 2%, nipagin: 3%) with 100 μl of 15% liquid yeast on the top of the food. For the fecundity experiment (see below), F1 larvae developed on a diet depleted in maize (4. 4%) and dextrose (4. 4%) with no added yeast to create less favorable conditions. Two generations before the experiments, Wolbachia infection statuses were checked by PCR using primers wsp81F and wsp691R [71]. Virgin F1 male and female flies were collected and aged for 3 and 5 days respectively. Because multiple male matings can decrease the strength of CI [72,73], a male and female were placed in a vial for 4–8 hours. In D. simulans, remating does not occur within 8 hours after the first copulation (Nina Wedell, personal communication). Females were then placed individually on a 50 mm diameter Petri dish with standard cornmeal diet containing food coloring with 15 μl of 15% liquid yeast on the top of the food. Around 20 hours later, females were removed and eggs were counted. Females that laid five or less eggs were discarded. Hatch rates were estimated by counting unhatched eggs about 35 hours later. The compatible crosses between uninfected males and uninfected females showed a mean egg hatch rate of 98%, thus suggesting that most females in this experiment were mated. Moreover, for Wolbachia-infected lines, discarding potentially non-mated females for which none of the eggs hatched did not change the significance of correlations with the other traits as mean hatch rates with or without those females were strongly correlated (Pearson’s correlation test: r ≈ 0. 99, df = 14, P < 0. 0001). F1 larvae were raised on our poor diet, and 0 to 2-day-old flies were placed on standard cornmeal food with live yeast on the surface to stimulate egg maturation. After 2 days (2- to 4-day-old), 3 males and 3 females were placed in petri dishes of colored poor diet. Over 6 days, flies were anaesthetized with CO2 and transferred onto a new dish every 24 hours. The number of eggs was recorded by photographing the Petri dish and counting eggs using a multi-point counter tool in ImageJ [74]. As for the hatch rate experiment, F1 larvae were raised on our standard cornmeal diet. Five male and 5 female freshly emerged flies were placed per vial on poor diet. Flies were tipped onto fresh food every 3 days and the number of dead female flies recorded daily for 72 days until all flies died. To investigate Wolbachia tissue tropism, F1 larvae were reared on standard diet, and virgin males and females aged to 3 and 5-day-old respectively. Males were then anaesthetized on ice and dissected in Ringer’s solution [75]. For each Wolbachia strain, 10 pools of 5 pairs of testes were collected. Five-day-old females were allowed to mate with 2- to 4-day-old virgin Wolbachia-free STCP males for 24 hours. Females were then isolated and 10 replicates of 3 females per strain were placed in Petri dishes onto grape agar food with 15 μl of 15% liquid yeast on the top. After 6 to 8 hours, 20 eggs were harvested from each Petri dish and transferred into a microcentrifuge tube. In parallel, the head and thorax was separated from the abdomen of 6-day-old females. For each Wolbachia strain, 10 replicates, each consisting of a pool of head and thorax collected from 10 females were transferred into microcentrifuge tubes. All tissues were frozen at -80°C for DNA extraction. DNA was extracted from the tissue samples using EconoSpin All-In-One Silica Membrane Mini Spin Columns (Epoch Biolabs) and the QIAamp DNA Micro kit (Qiagen). Using the extracted DNA, quantitative PCR (qPCR) was used to determine the Wolbachia density in the carcasses (head and thorax), testes and eggs. For carcasses and testes, the amount of the Wolbachia gene atpD (atpDQALL_F: 5’-CCTTATCTTAAAGGAGGAAA-3’; atpDQALL_R: 5’-AATCCTTTATGAGCTTTTGC-3’) relative to the endogenous control gene actin 5C (Forward primer: 5’-GACGAAGAAGTTGCTGCTCTGGTTG-3’; Reverse primer: 5’-TGAGGATACCACGCTTGCTCTGC-3’) was quantified using the SensiFAST SYBR & Fluorescein kit (Bioline). The Wolbachia density was estimated as: 2ΔCt, where Ct is the cycle threshold and ΔC t = Ctactin5C-CtatpD. The PCR cycle was 95°C for 2 min, followed by 40 cycles of 95°C for 5 s, 55°C for 10 s, 72°C for 5 s. Since embryo mortality due to Wolbachia was observed in our experiment on hatch rate, the Wolbachia density in eggs was estimated as the amount of the gene atpD in a sample relative to the amount of the same gene in a positive control placed on every qPCR plate as follow: 2ΔCt, where Ct is the cycle threshold and ΔC t = Ctpositive control-CtatpD. For each sample, two qPCR reactions (technical replicates) were carried out and a linear model was used to correct for plate effects. Statistical analyses were performed using R [76]. Hatch rates were analyzed using mixed effect generalized linear models with a logit link function and the effect of individual mothers treated as random (package lme4). Fecundity was analyzed using a linear model with the total number of eggs laid over 6 days as a response and a random temporal block effect. Female lifespan was analyzed with a generalized linear model with the Wolbachia infection status as a fixed effect and vial as a random effect. To test the effects of Wolbachia strain individually on these traits we performed multiple comparisons with the control cross (uninfected flies) using Dunnett’s test (package multcomp). Wolbachia densities within tissues were log2-transformed and analyzed with a linear model including the effect of the Wolbachia strain, tissue, and their interaction. Between-strain differences in density were tested using multiple comparisons (Tukey’s HSD test, package multcomp). Between-trait correlations were tested with Pearson’s correlation tests unless the assumptions of normality and homoscedasticity were not reached, in which case Spearman’s tests were used. In order to take into account the phylogenetic non-independence of the data, significant correlations were further analyzed using independent contrasts [77] with the function crunch (R package caper) [78] (See S1 Table). | Arthropods are commonly infected with heritable bacteria, and some of these symbionts can protect their hosts against infection and/or be reproductive parasites. Which of these traits evolves will depend on whether the trait is costly to the symbiont and the host. Using a panel of strains of the symbiont Wolbachia in the fruit fly Drosophila simulans, we found that the beneficial effect of antiviral protection and the parasitic phenotype of cytoplasmic incompatibility occur independently across the strains. We found that high antiviral protection is associated with high symbiont densities and strong reductions in other life-history traits affecting the fitness of both the symbiont and the host. In contrast cytoplasmic incompatibility did not induce costs on these traits. This trade-off between antiviral protection and other fitness components may select for reduced antiviral protection, which would endanger the long-term success of programs using Wolbachia to block the transmission of mosquito-borne viruses. | Abstract
Introduction
Results
Discussion
Methods | 2015 | Should Symbionts Be Nice or Selfish? Antiviral Effects of Wolbachia Are Costly but Reproductive Parasitism Is Not | 8,482 | 248 |
|
Fatty-acid metabolism plays a key role in acquired and inborn metabolic diseases. To obtain insight into the network dynamics of fatty-acid β-oxidation, we constructed a detailed computational model of the pathway and subjected it to a fat overload condition. The model contains reversible and saturable enzyme-kinetic equations and experimentally determined parameters for rat-liver enzymes. It was validated by adding palmitoyl CoA or palmitoyl carnitine to isolated rat-liver mitochondria: without refitting of measured parameters, the model correctly predicted the β-oxidation flux as well as the time profiles of most acyl-carnitine concentrations. Subsequently, we simulated the condition of obesity by increasing the palmitoyl-CoA concentration. At a high concentration of palmitoyl CoA the β-oxidation became overloaded: the flux dropped and metabolites accumulated. This behavior originated from the competition between acyl CoAs of different chain lengths for a set of acyl-CoA dehydrogenases with overlapping substrate specificity. This effectively induced competitive feedforward inhibition and thereby led to accumulation of CoA-ester intermediates and depletion of free CoA (CoASH). The mitochondrial [NAD+]/[NADH] ratio modulated the sensitivity to substrate overload, revealing a tight interplay between regulation of β-oxidation and mitochondrial respiration.
Pathophysiological mechanisms underlying acquired and inborn metabolic diseases, such as type-2 diabetes and deficiencies in the fatty-acid oxidation, are largely elusive. Although we know many important molecular factors, yet the complexity of the metabolic and regulatory network hampers elucidating the relation between the primary disease factors and their systemic effects [1], [2]. Moreover, the experimental accessibility of large parts of the metabolic networks is limited. Computational kinetic models yield insight into the dynamics of metabolic networks and make predictions about the parts that are experimentally inaccessible. Fatty-acid (FA) β-oxidation is a prime example of a pathway involved in many diseases, but for which it is difficult to acquire a complete and quantitative view on the relation between metabolite concentrations and fluxes. Insulin resistance, one of the hallmarks of metabolic syndrome, is strongly associated with elevated levels of free FAs [3]. It has been argued that an imbalance between cellular FA uptake and oxidation leads to accumulation of FAs and other lipid molecules in the cytosol, which in turn causes insulin resistance [4], [5]. Others showed that a functioning acyl-CoA uptake into mitochondria is needed to develop insulin resistance, leading to the hypothesis that intermediates of FA β-oxidation are part of the problem [6]. Since the acyl-CoA intermediates are difficult to measure, conclusions are often based on acyl-carnitine levels in the blood [7], [8], which are interpreted as a reflection of acyl-CoA concentrations in the mitochondria. Similar limitations hamper the understanding of systemic effects of enzyme deficiencies in the FA β-oxidation and their impact on global energy and glucose regulation [9], [10]. Clearly, a more direct view on the dynamics of β-oxidation intermediates is urgently needed. A careful look at the basic biochemistry of the FA β-oxidation reveals complex interactions, of which the implications have never been investigated (Figure 1). First, it is a cyclic pathway. In each cycle the acyl-CoA substrate is shortened by two carbon atoms and the product is a substrate for the next cycle. Second, the shortened acyl-CoA product competes with the substrate for a set of enzymes with overlapping chain-length specificity. The complete breakdown of palmitoyl CoA, an acyl CoA with 16 carbon atoms, to 8 molecules of acetyl CoA (C2) requires seven reaction cycles. There are, however, only four acyl-CoA dehydrogenases and two parallel sets of enzymes for the further conversion of enoyl CoA (Figure 1). This results in a competition in two ways: substrates of different chain lengths compete for common enzymes, while enzymes with overlapping specificity compete for common substrates. The competition between substrates generates a feedforward inhibition in the network: the more palmitoyl CoA molecules enters the pathway, the more enzyme molecules they occupy, which are then not available for shorter acyl CoAs downstream in the pathway. Third, the two parallel pathways for the breakdown of enoyl-CoA molecules are of a different nature. Enoyl CoAs are either converted by a sequence of three enzymes (crotonase, medium/short-chain hydroxyacyl-CoA dehydrogenase (M/SCHAD) and medium-chain ketoacyl-CoA thiolase (MCKAT) ) or via the mitochondrial trifunctional protein (MTP), which catalyzes the entire sequence of reactions (Figure 1). At present it is unclear how these properties affect the pathway behavior. In contrast to the myriad of models of carbohydrate metabolism, we know of only two kinetic models of FA β-oxidation and each of these ignores the above biochemical interactions. The first model [11] contains only a single cycle of β-oxidation and thereby lacks the competition between the pathway substrate and its downstream products. The second model does comprise the complete conversion of acyl CoAs to acetyl CoA, but the authors overlooked that this should lead to competition and modeled it as if distinct pools of each enzyme existed for each substrate [12]. In this study we present and validate a quantitative kinetic model of the mitochondrial FA β-oxidation, which explicitly includes molecular competition. This model reveals how the complex biochemical wiring of the network gives rise to non-intuitive or ‘emergent’ behavior [13]. We will show i) how the competition renders the pathway vulnerable to substrate overload; ii) how the robustness of the system can be modulated; and iii) what the role is of the experimentally inaccessible CoA esters in this process.
We constructed a model for mitochondrial FA β-oxidation with all enzyme reactions and transporters (Figure 1). The reactions correspond to the mitochondrial β-oxidation module in the updated human metabolic reconstruction [14], with the difference that we tailored them to rat liver. Thus, we added the long-chain acyl-CoA dehydrogenase (LCAD), which is involved in rodent β-oxidation [15]–[18]. The model is limited to the β-oxidation of saturated FAs containing an even number of carbon atoms. It starts from the transport of palmitoyl CoA across the mitochondrial inner membrane by the carnitine shuttle. The shuttle is reversible and also catalyzes the export of acyl-CoA esters of other chain lengths, which can be released as acyl carnitines in the extramitochondrial space. For simplicity we assumed that CPT1 activity is restricted to palmitoyl CoA (C16). Once inside the mitochondrial matrix, acyl-CoA molecules are oxidized to enoyl CoA by a set of chain-length-specific, FAD-dependent acyl-CoA dehydrogenases (SCAD, MCAD, LCAD and VLCAD in Figure 1) [15], [17]. The resulting enoyl CoAs undergo a sequential hydratase, dehydrogenase and thiolase reaction, leading to the formation of an NADH, an acetyl CoA and an acyl CoA which is two C-atoms shorter than the substrate of the cycle. This sequence of reactions is either catalyzed by the MTP (the right-hand branch in Figure 1) [19] or by three different enzymes (the left hand ‘crotonase’ branch in Figure 1) [20]–[25]. We assume that MTP does not release the hydroxyacyl-CoA and ketoacyl-CoA intermediates in the matrix, but channels them from one active site to another. In contrast, the intermediates in the ‘crotonase’ branch can diffuse freely through the matrix. There is overlap between the substrates that can be converted by the two pathways. However, MTP has a preference for longer chain lengths, while C4 and C6 substrates are uniquely converted by the crotonase branch. The reaction scheme in Figure 1 is translated into a set of 45 ODEs describing how the time derivatives of the concentrations of the variable metabolites depend on the enzyme rates. For instance: (1) denotes that the rate of change of the enoyl-CoA concentration of chain length C14 in the mitochondrial matrix equals the rate v of formation by VLCAD and LCAD minus the rate of consumption by crotonase and MTP. VMAT denotes the volume of the mitochondrial matrix. In equation 1 each reaction rate is specific for a certain combination of enzyme and substrate, since most enzymes catalyze multiple reactions. The competing substrates inhibit each other' s conversion competitively. This leads for instance to the following set of rate equation for crotonase: (2) In the above ODE (equation 1) n = 14 for conversion of C14 and therefore the parameters for C14 are filled in, except in the Σ-term which sums over all substrates and products that are converted by crotonase and thus bind the enzyme reversibly. All of the reactions are modeled as reversible, mostly of Michaelis-Menten type as above, and most parameters were taken from the literature and mainly based on biochemical characterization of purified rat-liver enzymes. Some affinity and equilibrium constants were taken from rat-heart datasets. However, for enzymes of which only one isoenzyme exists, the affinity constants may be considered to be tissue-independent. Equilibrium constants do not depend at all on the enzyme that catalyzes the reaction. Three parameters, Vcpt1, Vvlcad and Vlcad, were estimated, since we could not find a reliable value in literature. Three conserved moieties follow from the pathway stoichiometry: the total of FAD, NAD+ and CoA species. The model fulfills the criterion of microscopic reversibility, i. e. alternative reaction paths have the same overall equilibrium constant. For instance, the overall equilibrium constant of the MTP reaction equals the product of the equilibrium constants in the crotonase branch. We note that most equilibrium constants are not far from 1; hence the pathway lacks a strong thermodynamic driving force and depends on the delivery of substrate upstream and the further conversion of products downstream. A notable exception is the couple M/SCHAD – MCKAT, which has equilibrium constants of respectively 2. 2·10−4 and 1. 1·103. This implies that M/SCHAD favors the reverse direction, which is compensated by the high equilibrium constant of MCKAT. MTP dampens this thermodynamic hurdle by keeping the intermediate ketoacyl CoA bound. Based on the above principles, we constructed a model of 45 variable metabolite concentrations, 56 reactions and 234 parameters (for detailed description see Text S1), which predicts fluxes and metabolites in time and at steady state. The predicted model outcome was validated with experiments on isolated rat-liver mitochondria. We measured the oxygen-consumption flux and the acyl-carnitine concentrations (C4–C16) in time upon addition of palmitoyl CoA or palmitoyl carnitine. Mitochondria were incubated with an excess amount of malate. In this way CoASH could be regenerated from acetyl CoA by malate dehydrogenase and citrate synthase, allowing the β-oxidation to proceed. Each cycle of β-oxidation in the presence of excess malate gives rise to 1 FADH2 by acyl-CoA dehydrogenase and 2 NADH by enoyl-CoA dehydrogenase and malate dehydrogenase. The oxygen flux due to the β-oxidation per se was obtained by correcting the measured oxygen consumption for the oxygen consumption due to malate dehydrogenase. That malate to citrate was the main source of NADH beyond the β-oxidation was confirmed by measuring citrate, which indeed accumulated over time. Over the entire time course of 24 min, there was no significant gap between the carbon consumed as palmitoyl carnitine and the recovered products: carbon consumed = −16 • ([acyl carnitineC16]t24 - [acyl carnitineC16]t0) = 328±89 µM (SD) and carbon produced = (SD); paired t-test p>0. 05. Note that we did not include the carnitine moiety in the balance as it plays only a catalytic role. The same holds for the four carbons in citrate that are derived from malate. α-Ketoglutarate was not detectable. The uncertainty of the carbon balance leaves, however, room for other intermediates to contribute. We determined the time course of palmitoyl carnitine and used this as an input function for the substrate concentration in the model. We assumed that the time course was similar for palmitoyl CoA when the latter was given as substrate (see Text S1). Within the experimental error the experimental and modeled flux corresponded well on both substrates (Figure 2A and Figure S1A). With palmitoyl carnitine, but not with palmitoyl CoA, a sharp initial overshoot was observed in the oxygen consumption flux, which was reproduced qualitatively by the model (Figure S1D–G). For the experiment with palmitoyl CoA as substrate, the dynamics of palmitoyl carnitine and the downstream intermediates myristoyl, octanoyl and hexanoyl carnitine (C14, C8 and C6) were quantitatively reproduced by the model. The remaining acyl carnitines (C10–C12 and C4) showed a qualitative correspondence between the experiment and model (Figure 2 and Figure S2A): the dynamics of the acyl carnitines were in all cases in the same direction, but the concentrations and/or timescale differed. For the experiment with palmitoyl carnitine as substrate, a qualitative correspondence was found, but the timescale and also concentrations differed (Figure S1B–C and S2B). We emphasize that the model parameters were not fitted to the experiments hence the correspondence between model and experiment is remarkably good. FA levels in plasma from obese people are often elevated [3], [8]. This means a constantly high availability of FAs for the β-oxidation. We have simulated this by calculating the steady-state flux and metabolite concentrations as a function of the concentration of palmitoyl CoA, the substrate in the model (Figure 3). The flux increased with an increasing concentration of palmitoyl CoA, but above 50 µM the flux decreased steeply, yet smoothly (Figure 3A, standard model). This was preceded by an accumulation of CoA esters (Figure 3B and Figure S3E) and consequently a decrease of CoASH (Figure 3B, standard model). Most often a metabolic flux reaches a maximum at saturating substrate concentration, hence we considered the steep decline at high palmitoyl CoA an emergent, non-intuitive property of the β-oxidation. To prove that this property resulted from the competition between metabolites for a common set of enzymes, we adapted the model such that a percentage of each enzyme was dedicated to a particular chain length (‘No competition’ in Figure 3). This percentage was attributed such that at 25 µM of palmitoyl CoA the flux distribution among parallel enzymes was similar in both models (Figure S4). Indeed, in the model without competition the flux increased until saturation (Figure 3A), while the accumulation of CoA esters and the depletion of CoASH were prevented (Figure S3F and 3B). A priori either the accumulation of the CoA esters or the depletion of CoASH could cause the decline of flux in the standard model. To distinguish between these possibilities we fixed the CoASH concentration in the standard model and left the CoA esters free to accumulate (Figure 3B), i. e. as if an external supply of CoASH had broken the moiety conservation. In this scenario the CoA esters accumulated extremely (Figure 3B and Figure S3G), but the flux nevertheless increased until saturation (Figure 3A, fixed CoASH). Accordingly, the drop in flux observed in the standard model is a result of CoASH depletion rather than of the high CoA-ester concentration. Sensitivity analysis revealed that the influx of FAs into the FA β-oxidation by CPT1 and the consumption of the products of the enzyme M/SCHAD (NADH and ketoacyl CoAs in the mitochondrial matrix) had the largest effect on the CoASH concentration (Table S1). To explore the role of CPT1, we decreased its activity by adding the endogenous CPT1 inhibitor malonyl CoA or by lowering its Vmax. At low concentrations of palmitoyl CoA the flux was lower than in the standard model, but it rose to the same maximum and the drop in flux occurred at a higher palmitoyl-CoA concentration (Figure 3A and S3A). Accordingly, the accumulation of CoA esters and the depletion of CoASH still occurred, but at higher concentrations of palmitoyl CoA (Figure 3B and S3B, S3H–I). The influence of M/SCHAD is due to its very small equilibrium constant (2 · 10−4 [11], [26]) implying that the reaction can only work in the direction of the β-oxidation when its product concentrations are extremely low as compared to its substrate concentrations. To shift the M/SCHAD reaction in the forward direction, we increased the NAD/NADH ratio in the mitochondrial matrix from 15 to 40 without altering the sum of NAD and NADH. This prevented the flux decline at higher palmitoyl-CoA concentrations (Figure 3C and S3C). Accordingly, the concentrations of the CoA esters stayed low (Figure S3J) and a high CoASH concentration was maintained (Figure S3D).
In this paper we present the first dynamic model of the FA β-oxidation that appreciates the complex biochemical interactions in the network. Notably, we included the extensive competition in the system, as well as the qualitatively different pathways for conversion of enoyl CoAs. The parameters were based on biochemical analysis of individual enzymes and not fitted to obtain the desired metabolite and flux profiles. In this light, the correspondence between model predictions and experimental observations (Figure 2) was remarkably good. This allowed us to further explore the properties of the pathway. We found that the unique pathway structure makes the FA β-oxidation vulnerable to substrate overload: at high palmitoyl-CoA concentrations the shorter CoA esters accumulate to outcompete the palmitoyl CoA. Above a critical palmitoyl-CoA concentration this results in depletion of CoASH and a steep decline in flux. This is an example of an ‘emergent’ property: it would not have occurred in a linear pathway without competition. Hence, it could not be predicted from the properties of the individual enzymes, but resulted from the wiring of the entire network. It is tempting to speculate that this overload phenotype is at the basis of various diseases and may be one of the mechanisms of lipotoxicity. We emphasize, however, that we modeled the FA β-oxidation in isolation. In reality, surrounding pathways may protect the pathway from overload. In addition, the model revealed two possible protection mechanisms at the pathway boundaries. In the following we will discuss these protective mechanisms as well the possible role of fatty-acid overload in insulin resistance. First, according to the model the flux decline could be prevented by decreasing the activity of CPT1. Accordingly, increased concentrations of the CPT1 inhibitor malonyl CoA as well as a decreased catalytic capacity of CPT1 have been observed experimentally [27]–[32]. The most convincing data are for malonyl CoA, which is increased in skeletal muscle of obese humans and rodents [27], [28], [31], [32] as well as in liver tissue from obese mice [27]. Instead of being a cause of obesity via decreased FA oxidation and increased synthesis, malonyl CoA may rather confer protection against overload of the β-oxidation. A second mechanism to prevent overload was to keep the products of the thermodynamically unfavorable M/SCHAD reaction low. In the parallel MTP pathway the long-chain hydroxyacyl-CoA (LCHAD) dehydrogenase is linked to the preceding long-chain enoyl-CoA hydratase and the following long-chain ketoacyl-CoA thiolase at the inner mitochondrial membrane [19], [23], [33]. Since the intermediate CoA esters are not detectable, it has been proposed that they are directly channeled from one active site to another [19], [23], [33]–[37]. The overall equilibrium constant of the lumped MTP reaction is 0. 7, dampening the thermodynamic hurdle at LCHAD. Channeling has not been described for the short-chain intermediates (C4 and C6) and therefore we modeled their conversion by a sequence of enzymes (the ‘crotonase branch’) including the M/SCHAD reaction. The impact of the thermodynamic hurdle at M/SCHAD is demonstrated by the fact that the long-chain intermediates have a strong preference for the MTP branch, even though they could in principle be converted by the crotonase branch (Figure S4). Only C4 and C6 substrates, which are not recognized by MTP, take the crotonase route. Due to the low equilibrium constant of M/SCHAD (Keq = 2×10−4), the C6 and C4 intermediates accumulate in the model as well as in the experiment (Figure 2). A high mitochondrial NAD+/NADH ratio shifts the M/SCHAD equilibrium in the forward direction, prevents accumulation of the short-chain intermediates and eventually protects against the overload phenotype (Figure 3). This provides a functional explanation for the co-existence of M/SCHAD and Complex I in a respiratory supercomplex [38], [39]: locally, the NADH produced by M/SCHAD may be kept low by channeling it directly to complex I. In agreement with this hypothesis, the respiration-linked β-oxidation rate in gently-disrupted mitochondria (assuming an intact complex between M/SCHAD and Complex I) was much higher than in completely-disrupted mitochondria [40]. Besides the protection mechanisms found in the model, alternative mechanisms might be provided by surrounding pathways. Limited formation of palmitoyl CoA, either by inhibition of the synthetase reaction or due to a low cytosolic CoA concentration, would be very effective. Little is known however, about regulation of palmitoyl CoA synthesis. Another option is upregulation of the pathways that consume the product acetyl CoA, such as the Krebs cycle or the formation of ketone bodies. Indeed, the liver can produce high amounts of ketone bodies when confronted with a high fat load [41]. According to the model, a decreased acetyl-CoA concentration (K1acesink in the parameter list) should increase the concentration of CoASH (Table S1), although not as strongly as a decrease of the mitochondrial NADH concentration. So far, direct experimental evidence for the overload phenotype is lacking. It is not unlikely, however, that it is at the basis of the well-known association of high FA levels to insulin resistance. Chronic exposure of muscle to elevated lipid levels results in an increased expression of FA β-oxidation genes, but this is not accompanied by an upregulation of downstream metabolic pathways, such as the TCA cycle and electron transport chain. In line with our model, it has been reported that this results in incomplete oxidation of fatty acids and accumulation of acyl carnitines and ketone bodies (reviewed in [41]). The fact that CPT1 is required to confer insulin resistance, suggests that the accumulated intermediate metabolites of the FA β-oxidation and/or ketone bodies may be involved in insulin resistance [6]. It is unclear, however, if the CoA esters or the carnitine esters are responsible for this effect, or both. Administration of l-carnitine, which is the ‘scavenger’ of CoA esters, sometimes restores glucose tolerance, in rats as well as humans [7], [42]. It is tempting to speculate that carnitine protects by trans-esterification of CoA esters to carnitine esters, which liberates CoASH and prevents accumulation of intermediates. In vivo it is not clear, however, if carnitine prevents overload, since it plays a dual role: in the mitochondrial matrix it scavenges intermediate CoA esters, but in the cytosol it drives the uptake of acyl CoAs into the mitochondria. Since glucose and FA oxidation share the mitochondrial cofactors NAD+/NADH and CoASH, it may be expected that the presence of glucose will make the β-oxidation even more susceptible to overload. Vice versa a severe drop in CoASH will also compromise glucose oxidation via the TCA cycle. To further understand the interplay between glucose and FA metabolism and the quantitative role of various protective pathways, it will be of key interest to link the new model with (partially) existing models of glucose metabolism, TCA cycle, respiration, ketone-body synthesis and FA synthesis [43]–[46]. This should be the next step in elucidating the mechanisms behind acquired and inborn diseases of glucose and FA metabolism.
Experimental procedures were approved by the Ethics Committees for Animal Experiments of the University of Groningen. The computational model was built and analyzed in Mathematica Wolfram. It consists of a set of Ordinary Differential Equations (ODEs). Time simulations were done with the algorithm NDSolve. Steady states were calculated by setting all time derivatives to zero and solving the resulting set of non-linear equations with the algorithm FindRoot. FindRoot is a root-finding algorithm combining damped Newton' s method, the secant method and Brent' s method. The solutions fulfilled the criterion that all time derivatives of metabolite concentrations approached zero (<10−11). We have no indications for alternative steady-state solutions, since different initial conditions led to identical steady states. As an input for the steady-state root-finding algorithm we used the endpoint of a time simulation. A Mathematica script and the corresponding pdf is added to this paper (Protocol S1 (steady state) and Protocol S2 (time simulation), as well as a detailed model description in Text S1). The models will become publically available at JWS Online Cellular Systems Modeling (jjj. bio. vu. nl). Mitochondria were isolated from liver tissue of adult female Wistar rats (250–300 gram) according to Mildaziene [47]. The oxygen consumption rate of uncoupled mitochondria was measured with either palmitoyl CoA or palmitoyl carnitine as substrate, in the presence of ADP at 37°C in a stirred, two-channel high-resolution Oroboros oxygraph-2 k (Oroboros, Innsbruck, Austria). All measurements were done in 2 ml of MiR05 medium [48] to which 0. 5 mg/ml mitochondrial protein, 0. 2 µM FCCP (uncoupler), 2 mM malic acid, 500 µM l-carnitine, 1 mM ADP and 25 µM of either palmitoyl CoA or palmitoyl carnitine were added. In parallel, 15–20 ml of the same reaction mixture was analyzed in a stirred, open vessel from which samples were taken in time for acyl-carnitine concentrations and TCA-cycle intermediates. Samples of 1. 5 ml each were quenched by adding HCl to a final concentration of 0. 4 M. The determination of the acyl-carnitine concentrations was done according to Derks et al. [49]. The TCA-cycle intermediates are measured according to [50]. | Lipid metabolism plays an important role in the development of metabolic syndrome, a major risk factor for cardiovascular disease and diabetes. Furthermore, inborn errors in lipid oxidation cause rare, but severe diseases in children. To obtain more insight into the response of lipid oxidation to dietary and medical interventions, we constructed a computational model. The model correctly simulated the rate of lipid oxidation and the time courses of most acyl carnitines. The latter are used as diagnostic markers in blood. Subsequently, we subjected the model to an increased supply of lipids, as often happens in obese people. We discovered that the lipid-oxidation machinery easily becomes overloaded, very much like a highway during rush hours: the more cars enter the road, the slower they proceed and the more they clog the road. Analogously, an overload of lipids slowed down the lipid oxidation and led to an accumulation of intermediate metabolites in the pathway. Potential protection mechanisms of cells consist of restricted entry of lipids into the oxidation pathway or efficient downstream processing of reaction products. In future research we will use the model to test dietary or medical interventions in silico and thereby guide the development of new treatment and prevention strategies. | Abstract
Introduction
Results
Discussion
Materials and Methods | systems biology
biochemistry
enzymes
metabolic networks
biology
computational biology | 2013 | Biochemical Competition Makes Fatty-Acid β-Oxidation Vulnerable to Substrate Overload | 6,724 | 277 |
Rhesus macaques are unusual among schistosome hosts, self-curing from an established infection and thereafter manifesting solid immunity against a challenge, an ideal model for vaccine development. Previously, the immunological basis of self-cure was confirmed; surviving worms had ceased feeding but how immunological pressure achieved this was unclear. The schistosome esophagus is not simply a conduit for blood but plays a central role in its processing. Secretions from the anterior and posterior esophageal glands mix with incoming blood causing erythrocyte lysis and tethering and killing of leucocytes. We have analysed the self-cure process in rhesus macaques infected with Schistosoma japonicum. Faecal egg output and circulating antigen levels were used to chart the establishment of a mature worm population and its subsequent demise. The physiological stress of surviving females at perfusion was especially evident from their pale, shrunken appearance, while changes in the structure and function of the esophagus were observed in both sexes. In the anterior region electron microscopy revealed that the vesicle secretory process was disrupted, the tips of lining corrugations being swollen by greatly enlarged vesicles and the putative sites of vesicle release obscured by intense deposits of IgG. The lumen of the posterior esophagus in starving worms was occluded by cellular debris and the lining cytoplasmic plates were closely adherent, also potentially preventing secretion. Seven proteins secreted by the posterior gland were identified and IgG responses were detected to some or all of them. Intrinsic rhesus IgG colocalized with secreted SjMEGs 4. 1,8. 2,9, 11 and VAL-7 on cryosections, suggesting they are potential targets for disruption of function. Our data suggest that rhesus macaques self-cure by blocking esophagus function with antibody; the protein products of the glands provide a new class of potential vaccine targets.
Schistosomiasis is one of the most important parasitic diseases in tropical and sub-tropical regions of the world, with about 800 million people at risk and more than 200 million infected [1,2]. With intense efforts over six decades, great progress has been made in combating this disease in China [3], nevertheless zoonotic schistosomiasis japonica remains a major public health problem. Although comprehensive measures, including mass treatment, snail control and environmental modifications have proved effective in reducing the prevalence and morbidity in endemic areas, none of them can prevent re-infection. Furthermore, unlike other schistosome species, Schistosoma japonicum has a wide range of reservoir hosts in China [4]; 42 species in 28 genera within seven orders of mammals can harbor the infection naturally, adding considerably to the complexity of disease control and prevention. A vaccine with long-term efficacy would augment efforts to control and ultimately eradicate the disease and hence has received wide attention but so far has proved to be an elusive goal. After rhesus macaques are exposed to cercariae of S. japonicum, an infection becomes patent and the animals manifest all the characteristics of a permissive host. However, unusually among all the natural and laboratory hosts, after a lag period of up to 12 weeks, faecal egg output drops sharply and eventually declines to zero [5,6]. The females recovered after this time were reduced in size with smaller ovaries and fewer eggs in the uterus [7]. These observations on the fate of a schistosome infection in rhesus macaques are common to both Japanese and Chinese mainland strains of S. japonicum [6,7] as well as two other human species, S. mansoni [8,9] and S. haematobium [10]. That they might represent a ‘self-cure’ from the schistosome infection was first proposed by Cheever et al. [6], although a clear definition was not articulated. Here we contend that a phenomenon where a patent infection establishes but worms are subsequently eliminated without any external intervention, has indeed all the hallmarks of a self-cure process. Although a number of studies on rhesus monkeys have demonstrated complete or partial self-cure, they provide few clues about the nature of the mechanism that drives the process. One reason is that most were carried out between the 1940s and 1970s when knowledge of immune effector mechanisms was quite limited. A frequent observation was that successful self-cure was related to initial cercarial dose, with only higher intensity infections bringing about almost complete worm elimination [9]. Most importantly, animals that self-cured showed a strong resistance to a challenge infection with normal cercariae [10,11,12]. Such resistance was only complete when a challenge exposure was applied to animals whose self-cure was well underway [10,13], indicating that resistance developed slowly. However, once protective immunity was established, it did not require any surviving adult worms for its maintenance because when they were killed by drug treatment, it still persisted [13]. More convincing evidence for an immunological basis to self-cure finally came decades later [14], where the timing and intensity of IgG production was shown to correlate inversely with the number of S. mansoni worms recovered by perfusion at 18 weeks. No evidence was found for an acute antibody-mediated lethal hit. Instead the mechanism appeared to involve cessation of blood feeding, starvation and ultimately organ failure. These conclusions were reinforced by the retarded growth of blood-feeding worms during in vitro culture with rapid-responder serum versus slow-responder or naive rhesus serum. Schistosomes feed avidly on blood, with male and female adult S. mansoni ingesting some 39,000 and 330,000 erythrocytes per hour, equating to daily intakes of 105 nl and 880 nl of whole blood, respectively [15]. We recently demonstrated that the worm esophagus does not act simply as a conduit but plays a central role in blood processing [16,17]. Erythrocytes are lysed there whilst leucocytes are tethered and damaged, forming a stationary plug in the posterior esophageal lumen around which blood flows. The posterior region is surrounded by a gland and, in S. japonicum at least, the anterior esophageal region is also a morphologically distinct secretory organ [18]. Very little is known about the products of these two glands, in part due to their minute size. Expression of four genes, Micro Exon Genes (MEGs) 4. 1,4. 2,14 and Venom Allergen Like (VAL) -7 [17,19,20] exclusively in the posterior gland of S. mansoni was revealed by whole mount in-situ hybridization (WISH). MEG-4. 1 protein expression was also demonstrated in the posterior gland of S. japonicum [17]. So far only one single gene MEG-12, has been identified from this newly discovered anterior esophageal gland, whose protein product was shown to destabilize erythrocyte membranes (R DeMarco, personal communication). Here we have made an in-depth exploration of the self-cure process in rhesus macaques infected with S. japonicum (China mainland strain). We examined the morphological and physiological aspects of the parasites, as well as host responses to them. We have also explored the possible mechanisms of self-cure and potential immune targets of the rhesus macaque host, against the background of new findings on schistosome feeding [21]. An ideal model on which vaccine development can be based should satisfy the dual criteria that high protection (up to 100%) is induced and persistent immunity is maintained. In this context, we believe our findings with this ‘old’ model open a new route to an effective vaccine.
The six adult male rhesus macaques used in the study had a mean age 9. 67±0. 82 years and a mean weight 7. 98±0. 85 kg at the outset. Cercariae of S. japonicum were shed from patent snails (Oncomelania hupensis) provided by the Jiangsu Institute of Parasitic Diseases (Wuxi, China), collected from the water surface using a bacteriological loop and placed on glass cover slips for infection. Rhesus macaques anaesthesized with ketamine hydrochloride (6 mg/kg body weight, Gutian Pharmaceutical Corporation, Fujian, China), were infected percutaneously with 600 cercariae, via the shaved abdominal skin for 30 minutes. Blood was obtained by intravenous sampling prior to infection and at 2-week intervals throughout the experiment up to perfusion (week 22), stood at room temperature for 1hour to clot, and kept overnight at 4°C to facilitate clot retraction before serum was recovered for storage at -20°C. The body weight of each monkey was also determined monthly. All animals were individually inspected daily. Those showing signs of diarrhoea were given oral dehydration therapy as required. Faecal samples were collected overnight at 2 week intervals from week 2. The number of eggs per gram of faeces (EPG) was determined from three individual samples/animal/time point, using both the Percoll technique [22] and the Kato-Katz method [23]. For Percoll, 250 mg fresh faeces were suspended in 3 ml phosphate buffer saline (PBS) and layered on top of 3 ml of 0. 9% NaCl/ 60% Percoll (Sigma, Germany) solution. After centrifugation, the pellet was re-suspended in a small volume and suspension passed through a mesh sieve to exclude larger particles; the eggs in the flow-through were then counted under a microscope. For the Kato-Katz method, nine slides, each containing 41 mg of sieved stool, were prepared from three individual stool samples and examined by qualified technicians in a blinded manner. Soluble circulating anodic antigen (CAA), released into the bloodstream from the parasite’s gut, was measured using the up-converting phosphor lateral flow (UCP-LF) technology as previously described [24,25], with modifications. Briefly, serum was treated with 4% (wt/vol) trichloroacetic acid to remove proteins and antibody complexes. After centrifugation at 10,000xg for three minutes, the supernatant was mixed with an assay buffer containing an anti-CAA mouse monoclonal antibody conjugated to UCP reporter particles and incubated for 1 hour at 37°C. A lateral flow strip was placed into the mixture and chromatography was permitted to continue until strips were dry. Strips were read using a modified Packard Fluorocount meter, and the test line was normalized to the control line for each test strip; serum CAA levels were then expressed as the ratio of fluorescence counts of the test line/control line. Perfusions to recover adult worms were performed 22 weeks after infection. In order to collect blood samples animals were first anaesthetized as described above. Heparin (Yi Cheng Bio technique Co. Ltd. , Shanghai, China) (5000 units) was injected via the femoral vein afterwards and allowed to circulate for 5 minutes. The animals were then sacrificed by a further injection of sodium pentobarbitone. Portal perfusion was performed as described for the olive baboon with modifications [26]. Briefly, the thorax and abdominal cavity were opened and the aorta clamped above the bifurcation into the iliac arteries. The vena cava was clamped just before entry to the heart. A Foley balloon catheter was then introduced via an incision in the aorta and the hepatic portal vein was slit. RPMI-1640 medium (minus phenol red, Gibco, Life technologies) buffered with 10 mM HEPES (Invitrogen) was then infused into the aorta via a peristaltic pump, and the perfusate collected at the portal vein outlet. The worms were concentrated by gravity and washed in fresh RPMI-1640 medium plus 10 mM HEPES. After counting under a stereomicroscope, the worms from each macaque were divided into four groups and fixed for later processing. These four groups were used for morphological observations, immunocytochemistry, ultrastructure, and cryotomy, respectively. After fixation, worms were photographed in the different fixatives using a digital camera (Nikon D70 with manual macrolens and extension tubes to give x1. 5 magnification) to record general morphological information and visual appearance. Body length as an index of well-being was measured from the resulting images of all worms using the ‘Analyzing digital images’ package (Lawrence Hall of Science). The amount of hemozoin pigment in the gut was also used as a second index but for females only. Based on their gut pigment, they were classified as: black, gut completely full from ovary to tail, plus traces in the anterior bifurcated section; intermediate, pigment incomplete in the posterior region; white, virtually no pigment anywhere. Reference normal worms were recovered from C57BL/6 mice and rabbits (National Institute of Parasitic Diseases, Chinese Center for Disease Control and Prevention) six weeks after exposure to 40 and 1000 cercariae, respectively, fixed and photographed as above. Random samples of worms were fixed and stored in AFA (ethanol/40% formaldehyde/glacial acetic acid, in the ratio 85: 10: 5). Their morphology was visualised by staining for 30 minutes in Langeron’s Carmine [27], differentiation in 70% acid ethanol until background stain had disappeared, clearing in Histoclear (National Diagnostics UK, Hessle, UK), and mounting in DPX (VWR International Ltd, Lutterworth, UK). High magnification pictures of the ovaries of carmine-stained worms were obtained using a MZ16F stereo microscope (Leica Microsystems, Milton Keynes, UK) fitted with a Spot RT3cooled CCD camera (Diagnostic Instruments Inc, Sterling Heights, MI). The surface area in cross section of ovaries was calculated using length and width measurements from digital images, and the formula π a b, where a and b are the minor and major axes of an ellipse. Total eggs in the uterus of 34 females from rhesus macaques were independently counted directly under a compound microscope by two people. The uterine egg complement of females from rabbits and mice was determined after length measurement, by digesting singly in 10% KOH at 37°C for 1 hour to dissolve the body tissue before counting. For immunocytochemistry, intact adult worms were fixed and permeabilized using the protocol of Mair et al. [28]. The scope of this investigation was limited by the number of worms recovered from each monkey. Briefly, they were fixed for four hours in 4% formaldehyde and then incubated in permeablising fluid (PBS containing 0. 1% Triton X-100,0. 1% BSA and 0. 02% sodium azide; antibody diluent solution, AbD) overnight at 4°C. Subsequent steps were carried out, with shaking, at 4°C in AbD. To detect the presence of host antibodies, permeabilized worms recovered from macaques were incubated with FITC-labelled rabbit anti-monkey IgG (H+L; Sigma-Aldrich, Poole, UK for two days at 4°C and given extensive washes in AbD before viewing. To explore details of intrinsic bound antibody in the esophageal lumen, worm head cryosections were prepared as described by Li et al [18] and reacted with 1: 100 dilution of FITC- labelled rabbit-anti monkey IgG. The musculature and nuclei were visualized by staining of f-actin with a 1: 100 dilution of AF555-conjugated phalloidin (Invitrogen, Molecular Probes) plus 4' , 6-diamidino-2-phenylindole (DAPI; diluted in 1: 600 in 10% Normal Goat serum in PBS; Sigma-Aldrich, Poole, Dorset, UK), respectively, for 30 minutes. Although only one MEG has so far been described in S. japonicum, an analysis of genes highly up-regulated in the Day 3 skin schistosomulum of S. mansoni has provided pointers to other genes potentially expressed in the esophageal region [29]. We were thus able to find predicted sequences for SjMEGs 4. 2,8. 2,9, 11 and14, and SjVAL-7, by searching the NCBInr database using the homologues of known or suspected esophageal gland proteins from S. mansoni [17,29]. To investigate the localization of these six proteins, antibodies were raised by vaccinating rats either with the purified recombinant expressed in a modified pET28a system (SjVAL-7 and SjMEG-8. 2) or a synthetic peptide (S2 Table) derived from the parent protein sequence and coupled via a C-terminal cysteine to carrier ovalbumin. The two recombinant proteins were purified from E. coli lysates after induced expression, using nickel-column affinity chromatography. The homogeneity of the preparations was confirmed by 1D SDS-PAGE electrophoresis and the amino acid composition by tandem mass spectrometry. A 100 μg sample in 0. 1 ml phosphate buffered saline (PBS), emulsified in 0. 1ml complete Freund’s adjuvant (Sigma-Aldrich), was administered subcutaneously to rats on the back of the neck, with two subsequent boosts at 3-week interval with conjugates emulsified in incomplete Freund’s adjuvant, before a terminal bleed at week 8. The antibody titres of individual rats were determined by ELISA using plates coated with bovine serum albumin-conjugated synthetic peptides or purified recombinant proteins. Permeabilized S. japonicum worms were reacted with the six rat antibodies at 1: 1000 dilution in AbD containing 10% normal goat serum and washed extensively in AbD before localization using Alexa Fluor (AF) 488-labeled goat anti-rat antibody at 1: 100 dilution (Invitrogen). To explore the correlation between intrinsic IgG distribution and localisation of known secreted proteins from esophageal glands, worms previously probed with anti-monkey IgG were reacted with seven rat primary antibodies (the above six plus anti-MEG-4. 1 Ab) before localization using AF647-labeled goat anti-rat antibody (Invitrogen) at 1: 100 dilution. The musculature and nuclei were visualised as described above for cryosections, except that permeabilised worms were incubated overnight with a 1: 100 dilution of phalloidin. Optical slices were obtained using a LSM-710 confocal microscope (Zeiss, Cambridge, UK). Imaging conditions were as follows: DAPI: 405 nm diode laser with 405 main beam splitter (MBS); FITC/AF488: 3 mW argon laser with 488/561/633 MBS; Phalloidin AF555: 561 nm diode laser with 488/561 MBS; AF647: Helium/Neon laser with 488/561/633 MBS; Langeron’s carmine: 561 nm diode laser with 488/561 MBS. The structure of the esophageal region in representative worms from self-cured rhesus macaques, showing little or no gut pigment, was examined by TEM and SEM exactly as described by Li et al [18]. Levels of IgG antibodies against soluble adult worm proteins (SWAP) and soluble egg antigens (SEA) were determined by ELISA, as described previously [14]. ELISA plates were coated at an Ag concentration of 1μg/well and all serum samples diluted 1: 1600 to allow between-time and-sample comparisons. Wells were probed with peroxidase-labelled rabbit anti-monkey IgG (Sigma-Aldrich), diluted 1: 2000, colour developed using TMB substrate (Sigma-Aldrich) and absorbance read at 450 nm. To detect the presence of antibodies against SjMEGs 4. 1,4. 2,8. 2,9, 11and 14, and SjVAL-7 antigen, we coated ELISA plates with each synthetic peptide conjugated to a carrier protein BSA or with recombinant proteins (for SjMEG-8. 2 and VAL-7) at 1μg/well and performed the ELISAs as above. Differences between normally distributed variables were tested for significance using Student’s t test. The significance of trends in antibody titre against SEA and SWAP antigen preparations over selected sections of the time-course was assessed by linear regression using the r2 value as the indicator. Fluctuations in monkey body weight after infection were assessed using both Student’s t test and the Wilcoxon Signed-Rank Test on paired data sets.
The status of the worm population in the rhesus macaques over the 22 week time course was evaluated by two surrogate measures of worm burden: egg number in the faeces, and the level of soluble circulating anodic antigen (CAA) derived from the worm gut in the bloodstream. Faecal eggs were detected as early as 5 weeks after infection with cercariae. The mean value for egg output peaked at week 8 at 900 eggs per gram of faeces (EPG, Fig 1A; range from 480–2000). However, it then dropped sharply to approximately 300 at week 10 and continued to decline thereafter but at a much slower rate. In two animals, egg excretion reached zero by the end of the study. The profile of CAA was similar to that of egg output except that the levels were already high by four weeks, indicating that blood feeding was well underway (Fig 1A). The CAA level rose steadily as worms reached maturity, with mean values plateauing at week 8 before declining linearly from week 12 onwards. There was very little CAA detected at the last two sampling times. We estimated the worm burden in each monkey at peak by combining the EPG values with published data on egg output per female, the fraction of eggs excreted into the faeces [30] and daily faecal production by macaques [31] (S1 Table). The estimated burden ranged from 136 to 560 worms, equating to between 23% and 93% of applied cercariae. In contrast, at perfusion there was a wide discrepancy in the number of worms recovered (S1A Fig) with burdens ranging from 15 to 112, equivalent to between 2. 5% and 18. 7% of applied cercariae. Unexpectedly the monkeys increased their body weight by a mean of 17% over the first 4 weeks (t-test, P<0. 05) in spite of being exposed to 600 schistosome cercariae (Fig 1B), presumably due to absence of food competition since they were now singly housed. However, a marked decline in body weight, coincident with the onset of egg deposition, was observed between 4 and 8 weeks to a mean of 79% of the week 4 weight (t-test, P<0. 01). All monkeys suffered from severe diarrhea and inappetence during this period. A gradual recovery occurred thereafter so that by week 16 the group mean was not significantly different from the week 4 peak weight. The non-parametric Wilcoxon Signed Rank Test for the significance of percentage changes in weight over the time course gave the same results as the t-test (weeks 0–4,4–8 and 4–12, P<0. 05; weeks 4–16 and 4–20, not significant). In appearance, the worms recovered from rhesus macaques 22 weeks post-infection differed distinctly from those that developed in the rabbit permissive host (S1B and S1C Fig). Both males and females were smaller in size compared with their normal equivalent from permissive hosts. Female body length ranged from 2. 3 to 7. 48 mm, the majority of them being shorter or much shorter than controls; indeed there was minimal overlap in the two populations (t-test, P<0. 01) (Fig 2A). The size of males from rhesus macaques was also smaller than those from permissive hosts (t-test, P<0. 05) (Fig 2B), although there was considerable overlap in the two frequency distributions. Females were generally paler (S1B Fig) than those from rabbits, which are always black along the length of their bodies (S1C Fig and Fig 2C). When females were classified by the amount of hemozoin in their guts, approximately 37% came within the black category (Fig 2D and S1A Fig) although not strictly comparable to normal worms (Fig 2C), both in respect of body size and pigment intensity. The remaining two thirds of females had intermediate amounts (Fig 2E and S1A Fig) or no pigment (Fig 2F and S1A Fig) in their guts, suggesting they either had difficulty ingesting blood for some time or had stopped feeding altogether and were starving to death. The average body length for black, intermediate and white groups was 5. 29 mm, 4. 75 mm and 4. 15 mm, respectively (all significantly different; t-test, P<0. 05), indicating that the parameter correlated with the amount of blood consumption. Indeed, the most pallid females at perfusion looked ‘sick’ or moribund. The mean female: male ratio of surviving worms in rhesus macaques was 0. 751: 1, less than that in rabbits (0. 888: 1) (t-test, P<0. 05), suggesting females are more susceptible than males to elimination in rhesus macaques. Unlike S. mansoni and S. haematobium, mature S. japonicum females that develop in permissive hosts have a robust reproductive capacity, with large numbers of eggs in the uterus and a total egg output up to 2000 eggs/female/day [30]. The egg number in the uterus of the normal mouse and rabbit worms examined was not significantly different (t-test, P> 0. 05) so the data were pooled; it ranged from 56 to 192 (Fig 3A), (median = 135; mean = 124 ± SE 37). By contrast, surviving females from rhesus macaques generally had many fewer eggs in their uterus, 77% of them with <40 (Fig 3A), (median 23. 5; mean = 31. 4 ± SE 5. 8; t-test, P<0. 0001). Nevertheless, there was a wide range between individuals: four females had no eggs while two had more than 100, signifying that they retained their reproductive capacity. Ovary surface area was linearly related to number of eggs in the uterus (Fig 3B), with the survivors from rhesus macaques all having smaller ovaries and fewer eggs in their uterus, than those from the permissive hosts with large ovaries and many eggs (areas significantly different, t-test, P<0. 0001). The linear relationship was even more prominent between egg number and body length (Fig 3C). Indeed, the three categories of females from rhesus macaques, although varying in their level of gut pigment, all showed other marked changes in their reproductive system (S2 Fig). The vitelline cells in vitelline ducts were smaller and more widely spaced (S2A Fig) whilst the oviduct was inflated (S2C Fig), compared with controls (S2D Fig). Massive numbers of spermatozoa were adherent to regions of oviduct in both ‘white’ and ‘black’ females (S2A and S2B Fig), waiting to fertilise an egg. Apparently insemination continued but egg production was reduced or terminated. Having established that surviving worms had mostly ceased blood feeding, we wished to discover if there were changes in the cellular structure of feeding-related organs. The anterior esophagus in S. japonicum functions as a secretory organ, delivering the contents of light vesicles to the lumen where they can interact with incoming blood [18]. In worms from permissive mouse hosts the narrow cytoplasmic corrugations that line the compartment terminate in threads of cytoplasm that create a mesh-like structure in the center (S3 Fig, Part 1C). However, in rhesus worms SEM revealed that these corrugations were shorter and the fringe of cytoplasmic threads was much reduced, losing its mesh-like qualities (Fig 4A). The individual corrugations appeared more fragile, with fenestrations giving a honeycomb appearance (Fig 4B versus S3 Fig, Part 1D). In addition, the minute pits on the flat plates of cytoplasm located within the fringe were obscured to varying degrees by a deposit of material, to become rounded, lumpy spheres (Fig 4C versus S3 Fig, Part 1F), with an average size of 1. 54 x1. 20 μm2. The tips of corrugations in mouse worms contain clusters of light vesicles in males and a single vesicle in females (dimensions: surface area of a cluster, ~1 μm2; single vesicle, 0. 43x 0. 31 μm2) (S3 Fig, Part 2A-2D). In contrast, TEM revealed that the corrugation tips in rhesus worms had a greatly enlarged surface area, up to 5 μm2, mean size being 3. 5 μm2 (Fig 4D). They contained more vesicles and the individual vesicles were larger (0. 78 x 0. 63 μm2). Indeed the largest rhesus worm vesicle was 1. 13 x 1. 02 μm2, which translated into volume made it 28 times larger than the average mouse worm vesicle (Fig 4D versus S3 Fig Part 2A). The contents of the enlarged rhesus worm vesicles were proportionately similar to those from mouse worms but the quantities of both granular material and membrane stacks were amplified (Fig 4E versus S3 Fig Part 2F-2G). At high magnification the membrane stacks comprised alternating light and dark bands of material that would normally be destined to refresh the surface membranocalyx of the anterior esophagus (Fig 4F). The central core of granular material comprised dark specks scattered over a lighter ground substance, indicating two classes of constituent. Finally, confocal microscopy on the cryosections of the anterior esophagus revealed intense oval deposits of IgG on the lining (average size 2. 04 x 1. 14 μm2), plus a background of smaller, fainter spots (Fig 4G). We also detected fibrin in the oval deposits (S4 Fig), although blood clot formation does not normally occur in the esophagus and staining was incompletely superimposed on the pattern of rhesus IgG. S. japonicum worms recovered from permissive hosts almost always have a plug of tethered leucocytes in the lumen of the posterior esophagus around which incoming blood flows (S5A Fig). In contrast, the posterior esophageal lumen of starving rhesus worms was always completely occluded by a mass of tissue debris in which the remnants of ingested host leucocytes (nuclei) were visible (Fig 5A). The condition of the cytoplasmic plates that projected into the lumen was also altered. Normally these are separated by spaces in which the aggregated secretions are visible (S5B Fig). In rhesus worms the plates appeared closely adherent with spaces largely absent (Fig 5A and 5B). Indeed, at higher magnification the spacing between the outer surfaces of adjacent plates averaged 17 nm over long stretches, giving the appearance of tramlines running in parallel (Fig 5C). Electron-dense structures (Fig 5A) were also visible around the edges of the debris (arrowed). Close inspection revealed these to be aggregates of membranous material or even intact macro vesicles, presumably derived from the secretions of the anterior esophagus (Fig 5D). Immunocytochemistry showed intrinsic host antibody bound to the lining of the posterior esophagus, with the luminal edges being strongly positive, whilst the plate sides stained more weakly (Fig 5E and 5F). All these changes are evidence that worm starvation is the result of an occluded esophagus lumen with host antibody the causative agent. In order to discover which worm structures were targeted by rhesus IgG, we reacted permeabilized intact adult males and females with FITC-labeled second antibody. The tegument of both sexes was strongly positive, particularly around the anterior (Fig 6A and 6B). However, sections of the alimentary tract were also positive, the anterior esophageal lumen and lining being especially intense, and to a lesser extent the posterior esophagus and the transverse gut (Fig 6B). These observations strongly suggest that the esophagus is a major ‘battlefield’ between the worms and the host, whilst not ruling out the other targets. We attempted to develop an esophageal antigen preparation by physical enrichment, cutting off male worm heads. However, proteomic analysis revealed that the dominant constituents were still the abundant cytosolic and cytoskeletal proteins present in SWAP (S3 Table). We therefore monitored the longitudinal antibody profile to worm and eggs using crude soluble antigenic extracts from the two life-cycle stages. No response against SEA was observed over the first four weeks but by week 6 the antibody titer had risen dramatically (Fig 6C) as a consequence of egg deposition in tissues. The level continued to rise, reaching the peak value at week 10 but the apparent erratic decline thereafter was not statistically significant (regression r2 = 0. 048, P>0. 05). As with the response to eggs, there was no detectable antibody to SWAP over the first four weeks. However, unlike the response to eggs, the IgG against worm antigens increased steadily from week 4 onwards, before reaching a plateau as late as week 18 (Fig 6C) (regression r2, weeks 4 to 18 = 0. 5, P<0. 0001; weeks 18 to 22, r2 = 0. 01, P>0. 05). The higher OD values for the anti-egg compared with the anti-worm reactivity emphasized the much stronger response that the host made to eggs and their products than to worm constituents. Although no MEGs are annotated in the S. japonicum genome database we were able to identify cDNA sequences for seven of these genes by BLAST searching of the NCBInr database. Antibodies were successfully raised against synthetic peptides from SjMEGs 4. 1,4. 2,9, 11 and 14 and affinity-purified recombinant proteins for SjMEG-8. 2 and SjVAL-7 (S6 Fig). Immunocytochemistry on either permeabilized whole worms or cryosections revealed that all seven proteins were expressed in the posterior esophageal gland of the adult S. japonicum worms (Fig 7A–7G). The pattern of staining was identical in each case, with the cytoplasm of the cell bodies being strongly positive, whilst the individual nuclei appeared as dark holes most clearly visible in Fig 7F. The secretion of the SjMEGs 4. 2 and 9 was evident from their detection on host cells in the esophageal lumen (Fig 7B and 7G). The staining pattern for SjMEG-8. 2 on cryosections of the esophagus revealed its presence both on the posterior and the anterior lining (Fig 7C). As a pointer to immune reactivity we determined the extent of co-localization between host IgG in the esophagus and the presence of potential target proteins. A caveat was that such observations were only possible on worms recovered from high burden monkeys, which were likely to be the least reactive to worm secretions. In the third image of each triplet, a lemon yellow color indicated co-localization whereas regions of the lumen that remained red indicated that a given protein was not likely to be a target of IgG. In the case of MEGs 4. 1,9 and 11 proteins, IgG appeared to be almost completely superimposed with no free protein visible, suggesting they were likely targets (Fig 8A, 8C and 8D). Similarly, MEG-8. 2 protein in both posterior and anterior esophageal lumen, and on the lining, was completely superimposed by IgG (Fig 8E). For VAL-7, the co-localization between IgG and protein appeared in the posterior esophagus (Fig 8F). However, whilst MEG-4. 2 and IgG staining was superimposed in the posterior esophageal lumen (Fig 8B), free MEG-4. 2 was visible in the anterior esophagus indicating it was not associated with IgG. No colocalization of IgG and MEG-14 was observed. We also investigated the IgG response of individual rhesus macaques to the demonstrated esophageal gland secreted proteins. For this purpose, ELISA plates were coated with the same synthetic peptides/recombinant proteins used to elicit specific antibodies in rats for immunocytochemistry. The resulting data were displayed by monkey number (Fig 9A) and by synthetic peptides or recombinant constructs (Fig 9B) on the x axis. All monkeys made a detectable response to one or more targets but there was considerable variation among individuals. All six monkeys made a response to MEG-4. 2, MEG-11 and VAL-7, five to MEG-9, four to MEG-14 and three to MEG-8. 2. No monkey responded to the short MEG-4. 1 peptide. The maximum intensity of response to a given construct varied greatly among animals. Data plotted by monkey revealed the overall response to esophageal proteins in descending order, 4,5, 3,6, 2,1 but the impressive repertoire of monkeys 4 and 5 did not correlate with worm burdens at perfusion, these being high and low burden, respectively.
Initially, the rhesus macaque behaves as a permissive host for S. japonicum, with approximately 75% of applied cercariae being recovered as mature adults six weeks post infection, worm bodies of both genders being even longer than those from the permissive rabbit host [7]. Viable eggs are laid in large numbers and granulomatous liver pathology ensues [7]. However, unlike most other hosts of this zoonotic parasite, the monkeys develop a self-cure response, so that by 20 weeks or so only a small fraction of the original worm population remains [7]. In our study, the data on faecal egg output showed that a robust infection was established in all animals after cercarial exposure. The CAA level also provided good evidence for establishment of a considerable worm population. The body weight reductions that followed oviposition plus the diarrhea and inappetence at week 8 (typical clinical symptoms of acute schistosomiasis) were independent evidence for the magnitude of infection, whilst the subsequent increase in weight revealed that by week 12 the impact of the infection was waning. The coincident reduced egg output while CAA levels remained high, showed that worms were still present but already manifesting impaired reproductive capacity. Indeed, females recovered from rhesus macaques and permissive hosts clearly fell into two distinct populations, the former being shorter and having fewer eggs, whilst the latter were longer with more eggs. The much smaller size of females at week 22 compared to their counterparts at week 6 [5] and their declining reproductive capacity is striking; it is clear that their bodies shrink as their feeding is impaired. Nevertheless, the tegument and gastrodermis were intact with negligible changes in structure observed. The fact that the less pigment a female had in its gut the shorter it was, again demonstrates that blood feeding is vital for worm health, particularly in females. Even the rhesus females that we classified as ‘black’ actually contained much less hemozoin pigment than mature females from rabbit and mouse hosts, indicating that even the most active were processing less blood than normal worms. Furthermore, the lower female: male sex ratio suggests that females are more susceptible to rhesus macaque responses. The impact of rhesus host responses on males appears less marked. On average they were only 30% shorter than normal males, while many could still produce sperm and inseminate females, even debilitated females. These gender discrepancies can be explained by differences in the relative dependence of males and females on feeding via the tegument versus the gut [21]. The male worm relies on massive trans-tegumental uptake of glucose to supply its energy requirements (4. 5 x its own dry weight per day), with correspondingly less reliance on blood ingestion via the gut. Conversely, the female ingests four times her own dry weight in blood protein via the gut, with relatively much lower glucose uptake across the tegument. On this basis females would be more susceptible than males to immune effector mechanisms operating against esophageal functions that disrupted feeding. Whatever the mechanism, the upshot appears to be that males survive in rhesus macaques better than females; however, both sexes eventually succumb. The marked changes recorded in the reproductive system of females are most likely attributable to reduced nutrient uptake (rather than the converse). The only other major changes in cellular morphology, detected by electron microscopy, were in the esophageal region. Whilst not ruling out other causes for worm demise, the most economical explanation is that rhesus antibodies target the still poorly understood esophageal functions. Indeed, the anterior esophageal gland was only formally described in late 2014 [18]. Progress in understanding how neutralizing antibodies might block the functions of esophageal proteins is hampered by the lack of knowledge about the secretions and the roles of individual constituents in blood processing. Our observations on the structure of the anterior esophagus of rhesus worms reveal striking differences from the structures we have described in comparable worms from the permissive murine host [18]. There is a marked reduction in the surface area of the lining, with more fragile corrugations and fewer spaghetti filaments, which could reflect immunological pressure associated with the self-cure mechanism. However, the principal changes appear to be in the vesicle secretory process itself. Major discrepancies in the morphology of worms from rhesus macaques compared with normal worms are the swollen expanded tips of the anterior corrugations packed with greatly enlarged light vesicles, and the deposition of material on the flattened plates partially or completely obscuring the surface pits. Our data strongly suggest that the deposits comprise host antibody and also fibrin. The pits on the flattened plates may be homologous to porosomes [32], where vesicle docking occurs on apical plasma membranes in many cell types [32,33]. This process involves formation of a fusion pore/channel [34,35] which can remain open for an extended period, permitting egress of contents. Furthermore, in some cell types, vesicles adjacent to the plasma membrane fuse with it while underlying vesicles fuse with their neighbors to form a chain of secretion through the single fusion pore [36]. In worms from rhesus macaques, it is plausible that antibodies against one or more vesicle constituents enter via the fusion pore to form immune complexes that block discharge into the esophageal lumen. Staining of the oval blobs in the esophagus lining indicates that this may indeed occur. Giant vesicle formation could then result from vesicle to vesicle fusion after discharge of contents is blocked. Identification of the proteins secreted from the newly designated anterior esophageal gland is now in progress. Our TEM observations suggest that malfunction also occurs in the posterior esophageal compartment. One factor may be the close adherence of post esophageal plates, potentially cross-linked by antibodies to peptide or glycan epitopes. Indeed, the average 17 nm distance between the outer surfaces of adjacent plates approximates to the flexible distance between the two antigen binding sites on a single IgG molecule (~15 nm). Moreover, adherence of the plates would prevent the secretions of the posterior esophageal gland, such as the MEGs and VAL-7 from reaching the central lumen. A second factor is the plug of tethered leucocytes present in the posterior esophageal lumen. In worms from permissive hosts we have demonstrated by video microscopy that blood is able to flow round this plug unhindered [17,21]. Our observation of a more substantial plug in some rhesus worms suggests a vital process has been inhibited such that build-up of tissue debris occurs. Undoubtedly, such inhibition will interfere with blood processing in the posterior esophagus. It has been documented that debility and elimination of adult worms does not occur in monkeys with a light infection, where egg output remains unchanged over long periods [9,37]. This is one of the strongest arguments for an immunological basis to self-cure but correlative measurements of antibody level were not made in these early studies. Based on our premise that antibody is responsible for worm elimination, we investigated the nature of the antigenic targets. Confocal microscopy on permeabilized worms revealed the tegument, esophagus and gut were all prominent sites of IgG binding in live worms recovered from rhesus macaques. Note that antibody does not bind to internal (i. e. intracellular) proteins in these intact worms in vivo. In the context of other potential targets we have evaluated the antibody responses to a protein array comprising 172 putative tegument constituents [38]. A significant response was detected to eight proteins, only one of which (SGTP1) is a verified constituent of the apical plasma membrane in S. mansoni. In the current study we elected to focus on the esophagus because of its newly appreciated role as the site of secretions that mediate the initial stage of blood processing [17]. In addition, its structure was modified in rhesus worms that showed a very obvious reduction in blood feeding. Probing a conventional SWAP preparation with the rhesus serum provided a general profile of the host response to soluble somatic antigen preparations comprising highly immunogenic, and internal proteins [39]. When we generated a protein fraction from S. japonicum heads, estimated to give a 50-fold enrichment of esophageal gland proteins, the principal constituents were still of cytosolic and cytoskeletal origin so we were compelled to take a piecemeal approach to investigate the reactivity of the esophageal proteins we had identified. We previously showed that four genes (SmMEG-4. 1, SmMEG-4. 2, SmMEG-14, SmVAL-7) encoding secreted proteins were specifically expressed in the S. mansoni esophageal gland, and that SjMEG-4. 1 protein was present in the posterior esophageal gland of S. japonicum [17]. In the current study we used immunocytochemistry to demonstrate that three more homologues of the S. mansoni genes (SjMEG-4. 2, SjMEG-14 and SjVAL-7) plus three novel SjMEGs 8. 2,9 and 11, are localized in and secreted from the posterior gland of S. japonicum. (SjMEG-14 is predicted to be membrane-anchored.) The way in which the secreted MEG proteins we identified interact with components of ingested blood to initiate its processing, remains to be established. However, it is noteworthy that MEG-4. 1 from S. mansoni was shown to heavily O-glycosylated [17] and both MEGs 4. 1 and 8. 2 from S. japonicum are predicted to have similar properties. Furthermore, the close association we observed between these two proteins and the plug of leucocytes in the posterior lumen suggests that their combined physical properties may be responsible for cell tethering. Our ELISA data revealed that the IgG responses of the monkeys to the seven targets differed, perhaps unsurprising, as they are from an outbred population. Nevertheless, all monkeys reacted with the SjMEG-4. 2 C-terminus, MEG-11 and VAL-7. No reaction to MEG-4. 1 was detected in any monkey, probably due to the short peptide we used to coat the plates. It is worth noting that this approach is most likely to detect linear epitopes, whilst antibodies that bind in vivo to native proteins in live worms can potentially react with conformational, linear and glycan epitopes. The co-localisation approach does not discriminate between two or more superimposed secreted targets in the esophagus lumen, whereas detection of a free protein not coincident with the antibody staining pattern may rule it out as a mediator of the self-cure process. On this basis, MEG-4. 2, which did overlay in some places, appears the least likely candidate while the remainder (MEGs 4. 1,8. 2,9, 11 and VAL-7) must be considered potential targets of the self-cure process. It is intriguing that IgG binding has also been detected in the esophageal lumen of worms from mice [18], yet they do not eliminate adult worms. We suggest this is due to the intensity and/or the specificity of the response and we are currently seeking differences in the reactivity of rhesus, mouse and rabbit infection sera to esophageal secretions. Our results emphasize the central place of the worm esophagus in the mechanism of self-cure mounted against adult worms by the rhesus macaque. They provide a clear explanation for the cessation of feeding that leads to worm starvation and death and they implicate the protein products of MEGs as likely targets. A recent study has revealed that this group of genes has been subjected to greater selective pressure than any other during evolution of the Genus Schistosoma [40]. It is tempting to conclude that the host immune system has applied the pressure mainly on the esophagus, where the battle has been waged. Our immediate task is to identify more esophageal secreted proteins, especially from the anterior gland where the morphological changes are most visible. Targeting esophageal proteins provides a novel, hitherto unexplored, route to an effective schistosome vaccine. Furthermore, indications from early studies in the rhesus macaque are that once the immunity has been induced, it is long lasting [13]. It is also possible that when subsequent invasions of worms reach the blood-feeding stage in the portal tract they have the capacity to boost the established protective response. This is not a feature found in the immunity to S. mansoni induced in another primate host, the baboon, by radiation-attenuated cercariae where the protection rapidly wanes and is not boosted by challenge larvae [41]. There are pointers from the literature that self-cure occurs in other natural hosts of S. japonicum, including water buffalo [42], but on a longer time scale, although it is not clear whether this has an immunological basis. Analysis of host antibody responses to a broad spectrum of esophageal secreted proteins should therefore provide the key to select the best vaccine candidates. In this context our findings on the immunological-based self-cure mechanisms in rhesus macaques and their association with esophageal secretions should open a new chapter for vaccine development against this important helminth zoonosis. | Rhesus macaques can self-cure from a schistosome infection. Antibody is crucial to drive this process and adult worm elimination is preceded by cessation of blood feeding. Recently we have shown that the schistosome esophagus plays a central role in blood processing. We first confirm the self-cure process in rhesus macaques infected with Schistosoma japonicum and provide evidence that the self-cure mechanism involves blocking the worm esophagus function with antibody. In the anterior region, secretion of light vesicles is disrupted hence their contents are not released into the lumen to interact with blood components to fulfil their tasks. The plates in the posterior lining stick together whilst the lumen is occluded, hampering blood processing. Furthermore, rhesus IgG binds strongly to the worm esophageal lumen and co-localizes completely with five esophageal secreted proteins, SjMEGs 4. 1,8. 2,9, 11 and VAL-7. Our results indicate that rhesus macaques eliminate their adult worms by disrupting esophageal function making blood difficult to ingest; feeding stops eventually causing their demise because nutrient uptake across the body surface cannot fully compensate. | Abstract
Introduction
Methods
Results
Discussion | 2015 | Evidence That Rhesus Macaques Self-Cure from a Schistosoma japonicum Infection by Disrupting Worm Esophageal Function: A New Route to an Effective Vaccine? | 12,627 | 316 |
|
Exhausted T cells express multiple co-inhibitory molecules that impair their function and limit immunity to chronic viral infection. Defining novel markers of exhaustion is important both for identifying and potentially reversing T cell exhaustion. Herein, we show that the ectonucleotidse CD39 is a marker of exhausted CD8+ T cells. CD8+ T cells specific for HCV or HIV express high levels of CD39, but those specific for EBV and CMV do not. CD39 expressed by CD8+ T cells in chronic infection is enzymatically active, co-expressed with PD-1, marks cells with a transcriptional signature of T cell exhaustion and correlates with viral load in HIV and HCV. In the mouse model of chronic Lymphocytic Choriomeningitis Virus infection, virus-specific CD8+ T cells contain a population of CD39high CD8+ T cells that is absent in functional memory cells elicited by acute infection. This CD39high CD8+ T cell population is enriched for cells with the phenotypic and functional profile of terminal exhaustion. These findings provide a new marker of T cell exhaustion, and implicate the purinergic pathway in the regulation of T cell exhaustion.
In acute infections, antigen-specific T cells differentiate into activated effector cells and then into memory T cells which rapidly gain effector functions and re-expand on subsequent encounter with the same pathogen [1]. In contrast, during chronic infections, pathogen-specific T cells gradually lose effector functions, fail to expand, and can eventually become physically deleted [2]. These traits are collectively termed T cell exhaustion, and have been described both in animal models of chronic viral infection as well as in human infections with hepatitis C virus (HCV) and human immunodeficiency virus (HIV) [2–4]. Identifying reversible mechanisms of T cell exhaustion is therefore a major goal in medicine. Prolonged or high-level expression of multiple inhibitory receptors such as PD-1, Lag3, and CD244 (2B4) is a cardinal feature of exhausted T cells in both animal models and human disease [5–7]. Expression of PD-1 appears to be a particularly important feature of exhausted CD8+ T cells, as the majority of exhausted cells in mouse models of chronic infection express this receptor, and blockade of the PD-1: PD-L1 axis can restore the function of exhausted CD8+ T cells in humans and mouse models [2,6]. However, in humans, many inhibitory receptors also can be expressed by a large fraction of fully functional memory CD8+ T cells. PD-1, for instance, can be expressed by up to 60% of memory CD8+ T cells in healthy individuals, making it challenging to use PD-1 to identify exhausted CD8+ T cells in humans, particularly when the antigen-specificity of potentially exhausted CD8+ T cells is not known [8]. Studies in mice and humans suggest that exhausted CD8+ T cells are not a homogeneous population, but instead include at least two subpopulations of T cells that differentially express the transcription factors T-bet and Eomesodermin (Eomes) [9–11]. T-bethigh CD8+ T cells represent a progenitor subset with proliferative potential that give rise to Eomeshigh CD8+ T cells, which are terminally differentiated and can no longer proliferate in response to antigen or be rescued by PD-1 blockade [9,12]. Both populations express PD-1, but Eomeshigh exhausted cells express the highest levels of PD-1. However, no specific cell-surface markers of this terminally differentiated population of exhausted cells have thus far been identified. CD39 (ENTPD1) is an ectonucleotidase originally identified as an activation marker on human lymphocytes and as the vascular ecto-ADPase [13], but has subsequently been shown to be a hallmark feature of regulatory T cells [14–16]. CD39 hydrolyzes extracellular ATP and ADP into adenosine monophosphate, which is then processed into adenosine by CD73, an ecto-5' -nucleotidase [17]. Adenosine is a potent immunoregulator that binds to A2A receptors expressed by lymphocytes causing accumulation of intracellular cAMP, preventing T cell activation and NK cytotoxicity [18–20]. Loss of CD39 in Tregs markedly impairs their ability to suppress T cell activation, suggesting that the juxtacrine activity of CD39 serves to negatively regulate T cell function [15]. However, blood CD8+ T cells have generally been reported to be CD39– [14,21–23], and the expression of this marker on exhausted T cells has not been examined. In this study, we demonstrate that, in contrast to CD8+ T cells from healthy donors, antigen-specific CD8+ T cells responding to chronic viral infection in humans and a mouse model express high levels of biochemically active CD39. CD39+ CD8+ T cells co-express PD-1 and are enriched for a gene signature of T cell exhaustion. In the mouse model of chronic LCMV infection, high levels of CD39 expression demarcate terminally differentiated virus-specific CD8+ T cells within the pool of exhausted CD8+ T cells. Thus, CD39 provides a specific, pathological marker of exhausted CD8+ T cells in chronic viral infection in humans and mouse models of chronic viral infection.
We surveyed the expression of CD39 by CD8+ T cells from healthy adult subjects without chronic viral infection. Consistent with previous reports we found that only a small fraction (mean 6%) of CD8+ T cells in healthy individuals expressed CD39 (Fig 1A and 1B) [14,21–23]. This small population of CD39+ CD8+ T cells in healthy donors was primarily found in the central and effector memory compartments while virtually no naive CD8+ T cells expressed CD39 (S1 Fig). We next focused on CD39 expression by antigen-specific CD8+ T cells specific for latent viruses in healthy subjects and found that only a very small fraction of CMV- or EBV-specific CD8+ T cells expressed CD39 (Fig 1A and 1B) (mean 3% and 7% respectively). We next measured CD39 expression by T cells specific for the chronic viral pathogens HCV and HIV. We measured CD39 expression in 57 subjects with acute HCV infections (23 with acute resolving infection and 34 with chronically evolving infection), and in 40 subjects with HIV infection (28 chronic progressors and 12 controllers; clinical characteristics of the subjects are summarized in S1 Table). We found a mean of 51% of HCV-specific CD8+ T cells and 31% of HIV-specific CD8+ T cells expressed CD39, a number significantly higher than CD8+ T cells specific for EBV or CMV, or in total CD8+ T cell populations from healthy individuals (Fig 1A and 1B). A slightly greater fraction of virus-specific CD8+ T cells from HCV-infected subjects expressed CD39 than did those from HIV-infected subjects. In subjects with chronic infection, the frequency of CD39-expressing cells in the virus-specific (tetramer+) CD8+ T cell population was significantly higher than in the total CD8+ T cell population (Fig 1C and 1D). However the fraction of total CD8+ T cells expressing CD39 in the CD8+ T cell compartment of individuals with HCV or HIV infection was slightly increased compared to healthy controls (Fig 1E), consistent with the presence of other, unmeasured virus-specific CD8+ T cells that were also CD39+ in the tetramer−fraction of CD8+ T cells. Thus CD39 is expressed infrequently by CD8+ T cells in healthy donors, but marks a large fraction of pathogen-specific cells CD8+ T cells in patients with chronic infection. CD39 expressed by regulatory T cells catalyzes the hydrolysis of ADP to 5’-AMP [14–16] but its enzymatic activity can be regulated by a range of post-transcriptional mechanisms [PMID. We therefore tested CD39 expressed by CD8+ T cells from patients infected with chronic HCV was functional using ATP hydrolysis as a surrogate marker of CD39 activity [24–26]. We sorted CD39– and CD39+ CD8+ T cells from six HCV-infected individuals (four with chronic infection and two with resolved infection) and incubated equal numbers of cells in the presence of extracellular ATP (eATP). Remaining levels of eATP were measured in the supernatant by HPLC. As a control, we assessed ATP hydrolysis by CD4+ CD25+ CD39+ regulatory T cells (Tregs) sorted from the same individuals (Fig 2A). Within the CD39+ CD8+ T cell population the level of CD39 expression was lower than in Tregs (Fig 2B). Consistent with reduced CD39 expression relative to Tregs, ATP hydrolysis by CD39+ CD8+ T cells was less than that by Tregs (Fig 2C). However ATP hydrolysis by CD39+ CD8+ T cells was significantly greater than that of CD39– cells (Fig 2C). Thus CD39 expressed by CD8+ T cells in HCV infection is enzymatically active and capable of hydrolyzing ATP. CD8+ T cells specific for chronic viruses such as HCV and HIV express increased levels of PD-1 [3,27]. We therefore examined the relationship between CD39 and PD-1 expression by virus-specific CD8+ T cells in 54 patients with HCV (23 chronically infected and 31 resolvers) and 40 patients infected with HIV (28 chronic progressors, 7 viremic controllers and 5 elite controllers). In both diseases we found a significant association between the level of expression (mean fluorescence intensity, MFI) of CD39 and PD-1 on antigen-specific CD8+ T cells in subjects with HCV and with HIV (r = 0. 70, P <0. 0001 and r = 0. 54, P<0. 05, respectively) (Fig 3A and 3B). We next examined the relationship between CD39 and PD-1 expression and viral load in HCV and HIV infection. We found that in both the HCV and HIV infection there was a modest but significant correlation between viral load and the level of CD39 expression on virus-specific CD8+ T cells measured by MFI (Fig 3C). The fraction of CD39+, virus-specific CD8+ T cells was significantly higher in HIV progressors compared with those from HIV controllers (S2 Fig). A similar, but non-significant, trend was seen comparing CD39 expression in HCV-specific CD8+ T cells in patients with chronic versus resolved disease. However, in HCV, a significantly higher fraction of virus-specific CD8+ T cells co-expressed both CD39 and PD-1 in patients with chronic versus resolved disease (S2 Fig). Consistent with these findings, there was a significant correlation between viral load and the fraction of virus-specific CD8+ T cells that were CD39+ PD-1+ double positive in both HCV and HIV infection (S2 Fig). PD-1 expression was also modestly correlated with the viral load in HCV and in HIV-infected patients (Fig 3D) [3,27]. Thus CD39 expression by virus-specific CD8+ T cells is greatest in setting of high antigen burden. In order to characterize more broadly the phenotype of CD39+ CD8+ T cells from individuals with chronic infection, we compared the global gene expression profiles of sorted CD39+ and CD39– CD8+ T cells from 8 HCV-infected subjects (3 with acute resolving infection and 5 with chronically evolving infection; S4 Table). Limited numbers of cells precluded the comparison of CD39+ and CD39– CD8+ T cells within HCV-specific cells, leading us to focus on the total CD8+ population of antigen-experienced CD8+ T cells (S4 Table). Because naive CD8+ T cells express little CD39 (S1 Fig), we excluded this population from the sorted cells (S3 Fig) to enable direct comparison of antigen-experienced CD39+ and CD39– CD8+ T cells. We first used unbiased clustering approaches to identify whether CD39+ and CD39– CD8+ T cells showed distinct patterns of gene expression. Analysis of gene expression profiles using consensus hierarchical clustering (Fig 4A) showed two distinct clusters of samples that corresponded almost exactly to CD39+ and CD39– populations, suggesting that that in both acute and chronic infection, CD39 expression demarcates two types of CD8+ T cells with markedly different patterns of gene expression. Supervised analysis of differential gene expression identified 619 genes differentially expressed (FDR<0. 15) between CD39+ and CD39– CD8+ T cells (S4 Table). Inspection of the list of differentially expressed genes revealed many with known roles in CD8+ T cell biology including increased expression of the inhibitory receptors PD-1 and CTLA-4 in CD39+ CD8+ T cells. To identify biological processes that were differentially active in CD39+ vs. CD39– cells, we performed gene set enrichment analysis using the Gene Ontology collection of gene sets [28]. We found no significant enrichment of GO terms in the CD39– CD8+ subset. In contrast, 21 gene sets significantly enriched (FDR<0. 1) in CD39+ population, almost all of which were related to mitosis and cell-cycle associated genes or cytoskeleton organization (Fig 4B). This suggests that CD39+ CD8+ T cells in chronic viral infection show coordinate up-regulation of genes related to proliferation. The expression of CD39 by CD8+ T cells in chronic but not acute/latent infection, suggests that it may be a marker of T cell exhaustion. We therefore tested whether the profile of CD39+ CD8+ T cells was enriched for genes expressed by exhausted CD8+ cells. Previous studies of gene expression in CD8+ T cells in the mouse model of chronic viral infection with the Clone 13 strain of LCMV have identified signatures of T cell exhaustion that are also enriched in exhausted CD8+ T cells in humans [29–31]. We therefore curated a signature of 200 genes up-regulated by exhausted CD8+ T cells responding to chronic infection relative to functional memory CD8+ T cells generated by acute infection (LCMV Armstrong strain). We found that the exhausted CD8+ T cell signature from LCMV model was significantly enriched in CD39+ vs. CD39– CD8+ T cells in subjects with HCV infection (Fig 4C). We focused on the “leading edge” genes contributing most to the enrichment [32], which correspond to genes up-regulated both in the mouse exhausted signature and in the human CD39+ profile. As expected, the leading edge genes included PD-1 (PDCD1), a feature of both human CD39+ CD8+ T cells and of exhausted CD8+ T cells in the mouse model (Fig 4D). In addition we found that up-regulation of many genes associated with proliferation including BUB1, TOP2A and MKI67 was common to mouse exhausted CD8+ T cells and human CD39+ CD8+ T cells. Thus CD39+ CD8+ T cells in HCV infection and exhausted CD8+ T cells in a mouse model of chronic infection share transcriptional features that include genes related to proliferation. Because the mouse signature of CD8+ T cell exhaustion was significantly enriched in the transcriptional profile of CD39+ CD8+ T cells in HCV-infected patients, we next asked if CD39 was up-regulated by CD8+ T cells in the mouse model of chronic viral infection. To address this question we compared two well-described mouse models of viral infection using two strains of LCMV: LCMV Armstrong that causes an acute infection that is resolved in up to 8 days; and LCMV Clone 13 that persists in mice for up to 3 months and leads to T cell exhaustion [5,6]. We measured CD39 expression and compared it to PD-1 expression in CD8+ T cells responding to each infection. While naive CD8+ T cells expressed neither CD39 nor PD-1 (Fig 5A), both were rapidly and coordinately up-regulated by antigen-experienced cells following either infection (day 7 post infection [d7 p. i. ], Fig 5B). However, in acute infection, the fraction of CD39 bright PD-1+ population decreased with time. In contrast, high expression of CD39 and PD-1 was maintained in Clone 13 infection. The accumulation of CD39 bright PD-1+ cells among the total CD8+ population was most apparent in the H-2Db GP276-286 tetramer-specific CD8+ T cells (Fig 5B). Thus after chronic viral infection, antigen-specific CD8+ T cells can be identified by high expression of both CD39 and PD-1. This difference in expression of both markers between chronic and acute infection is noticeable as early as d7 p. i. but becomes more pronounced with time after infection. Having determined that high, persistent expression of CD39 is a feature of LCMV-specific CD8+ T cells during chronic LCMV infection, we next sought to further characterize the phenotype of CD39+ CD8+ T cells during Clone 13 infection. We analyzed CD39 expression in antigen-experienced, CD44+ CD8+ T cells and found that mice infected with Clone 13 developed a population of cells with particularly high expression of CD39 (CD39high). This population was entirely absent in mice infected with the acute LCMV Armstrong strain at d35 p. i. , which only exhibited the presence of intermediate levels of CD39 staining (CD39int) (Fig 6A). Further characterization of the two sub-populations in Clone 13 infected mice revealed that the CD39high cells showed more down-regulation of CD127 (Fig 6B) and higher expression of PD-1 (Fig 6C) than did the CD39int population. Because the highest levels of PD-1 are characteristic of terminally exhausted CD8+ T cells in chronic infection [12,33], we tested whether CD39high T cells in chronic infection showed other phenotypic characteristics of terminal exhaustion. Analysis of expression of two additional co-inhibitory receptors, CD244 (2B4) and Lag3, showed that a significantly higher fraction of CD39high cells co-expressed multiple receptors, consistent with terminal exhaustion. In contrast, CD39int CD8+ T cells were generally negative for all three receptors analyzed (Fig 6D and 6E). We next examined the expression of the transcription factors T-bet and Eomes. We found that the CD39high subset of CD8+ T cells was comprised primarily of Eomeshigh T-betlow terminally exhausted phenotype, while the CD39int CD8+ T cells showed a comparable distribution of both (Fig 6F). Similarly, we found that in CD8+ T cells from subjects with either HCV or HIV infection, the CD39+ CD8+ T cell compartment contained a significantly higher ratio of Eomeshigh T-betlow: Eomeslow T-bethigh relative to CD39– CD8+ T cells (S4 Fig). Thus in both humans and mice with chronic viral infection, CD39+ CD8+ T cells show a phenotype consistent with previous descriptions of terminal exhaustion [9]. We next examined the functional properties of CD39high and CD39int CD8+ T cells from mice with chronic LCMV infection. Co-production of cytokines IFN-γ and TNFα is a feature of virus-specific T cells responding to acute infection and in the early stages of chronic infection but is progressively lost as exhaustion evolves [2]. To compare the functionality of CD39high and CD39int virus-specific CD8+ T cells, we isolated CD8+ T cells from mice with chronic infection at d35 post-infection and stained for IFN-γ and TNFα following in vitro stimulation with GP33-41 peptide. We found a significantly smaller fraction of antigen-specific coproduced IFN-γ and TNFα in CD39high CD8+ T cells compared to CD39int CD8+ T cells (Fig 7A and 7B). To confirm this finding, we analyzed the function of transferred P14 CD8+ T cells in chronic infection. The P14 TCR transgene recognizes the GP33-41 peptide of LCMV presented on H-2Db. We found that both the frequency of IFN-γ-producing and IFN-γ/TNFα co-producing P14 T cells was significantly lower in CD39high CD8+ T cells compared to CD39int CD8+ T cells (Fig 7C and 7D). The defect in cytokine secretion was not only observed in terms of the frequency of cytokine-secreting cells, but also in the amount of cytokine detected per cell. Even among cells that did secrete IFN-γ, we found the MFI of expression to be significantly lower in CD39high CD8+ T cells compared to CD39int CD8+ T cells (Fig 7E and 7F). Thus high levels of CD39 expression demarcate a population of exhausted cells with the poorest function in chronic infection.
The state of CD8+ T cell exhaustion is characterized by widespread changes in gene expression relative to functional memory CD8+ T cells [5]. However, in humans, identification of specific T cell exhaustion markers that are not shared by more functional CD8+ T cell populations has been challenging [8]. We show that high-level expression of the ectonucleotidase CD39 is characteristic of CD8+ T cells specific for chronic viral infections in humans and mice, but is otherwise rare in the CD8+ T cell compartment of healthy donors. Persistent, high-level expression is also seen in the LCMV mouse model of chronic viral infection, suggesting that CD39 expression is a phenotypic marker of CD8+ T cell exhaustion. Moreover, within the exhausted population in the mouse model, CD39high CD8+ T cells express the highest levels of PD-1, co-express multiple inhibitory receptors and have profoundly impaired function. We found that in both mice and humans, CD39 is expressed preferentially by CD8+ T cells that are T-betlow/Eomeshigh. These data suggest that CD39 expression by CD8+ T cells is a pathological finding and demarcates the population of CD8+ T cells previously identify as being terminally exhausted [9]. The fact that peripheral blood CD8+ T cells in humans can express CD39 is surprising. Previous data have shown that CD39 expression is restricted to CD4+ regulatory T cells, Th17 cells, and small populations of regulatory-like CD8+ T cells [14,21–23]. Indeed, we find that in the bulk population of CD8+ T cells in healthy donors only a small minority of CD8+ T cells expresses CD39. However, CD39 is abundantly expressed by virus-specific CD8+ T cells in two human chronic infections (HIV and HCV). This helps explain why CD39+ CD8+ T cells have not been appreciated in earlier studies that have focused on healthy individuals, and suggests that, in steady-state conditions, the expression of CD39 by CD8+ T cells is a pathological occurrence that is related to the development of T cell exhaustion. Whether the small fraction of CD8+ T cells expressing CD39 in healthy donors represents acutely activated CD8+ T cells, or those exhausted by asymptomatic chronic pathogens or inflammatory signals is an important question for future studies. Several features of CD39-expressing CD8+ T cells suggest that CD39 is a diagnostically valuable marker of T cell exhaustion. First, in both human and mouse CD8+ T cells responding to chronic infection, CD39 is co-expressed with PD-1, an inhibitory receptor expressed by the majority of exhausted T cells [5,6]. Second, CD39 expression correlates with viral load in subjects with HIV and HCV infection suggesting that the conditions of high levels of inflammation and antigen load that lead to exhaustion also increase CD39 expression in the virus-specific pool of CD8+ T cells, as has been observed for PD-1 [3,34]. Third, gene signatures characteristic of exhausted mouse CD8+ T cells are enriched in CD39+ cells relative to CD39– CD8+ T cells in subjects with HCV infection, underscoring the association between CD39 expression and T cell exhaustion. Finally, chronic LCMV infection in the mouse model increases CD39 expression by exhausted virus-specific CD8+ T cells, and elicits a population of CD39high cells that are absent in functional memory cells. Previous studies show that CD39, like PD-1, is transiently up-regulated by acute T cell activation [14,35]. Additional studies will therefore be required to determine the extent to which T cell activation (rather than exhaustion per se) contributes to the up-regulation of CD39 and PD-1 in chronic infection. However, the strong association between CD39 expression and the hallmark phenotypic features of T cell exhaustion in humans and a mouse model suggests that it can serve as a valuable marker of the exhausted T cells state. The expression of molecules, such as PD-1, that inhibit T cell function has been used to identify exhausted CD8+ T cells in several studies of human chronic infection and cancer [2]. However, there are important distinctions between the pattern of CD39 expression and that of inhibitory receptors. Many inhibitory receptors, such as PD-1 [3,8, 36] and CD244 [37,38] are also expressed by a substantial fraction of CD8+ T cells in healthy donors that are not exhausted. In contrast, CD39 expression is found in only a very small minority of CD8+ T cells from healthy donors. This expression pattern suggests that CD39 expression, particularly in combination with PD-1, may be useful as a more specific phenotype of exhausted CD8+ T cells, at least in HCV and HIV infection. In addition, CD39 may provide a useful marker to isolate exhausted CD8+ T cells in settings such as tumor-specific responses where very few reagents are available to identify antigen-specific T cells. Importantly, while CD39 is rare in the CD8+ compartment in healthy donors, it is expressed by CD4+ Tregs–as is PD-1 –making it difficult to distinguish between exhausted CD4+ T cells and Tregs by CD39 expression alone. Analysis of global expression profiles of CD39+ versus CD39– CD8+ T cells in HCV-infected subjects showed that the CD39+ fraction was strongly enriched for genes related to proliferation. This may at first seem counterintuitive, given the functional defects that have been described in exhausted CD8+ T cells [2,5]. However, data from the mouse model of chronic infection suggest that, unlike memory CD8+ T cells, exhausted CD8+ T cells are dependent on continuous exposure to viral antigen to ensure their survival and undergo extensive cell division at a rate higher than that seen in physiological homeostatic proliferation of the memory CD8+ T cell pool [39]. Exhausted CD8+ T cells therefore have a paradoxical increase in their proliferation in vivo despite reduced proliferative potential in vitro [40], explaining the increased expression of proliferation-associated genes in CD39+ CD8+ T cells in HCV infection and in mouse exhausted CD8+ T cells [9,41]. Recent studies of exhausted CD8+ T cells have revealed that two distinct states of virus-specific CD8+ T cells exist in chronically infected mice and humans [9,10]. Differential expression of the T-box transcription factors T-bet and Eomes characterize two populations, which form a progenitor-progeny relationship. T-bethigh cells display low intrinsic turnover but are capable of proliferation in response to persisting antigen, giving rise to Eomeshigh terminal progeny. In contrast, Eomeshigh CD8+ T cells responding to chronic infection had reduced capacity to undergo additional proliferation in vivo. The T-betlow /Eomeshigh exhausted subset of CD8+ T cells correspond to the PD-1 bright population that has also been shown to be unresponsive to PD-1: PD-L1 blockade. These data suggest that the differential expression of these transcription factors identifies subpopulations of exhausted CD8+ T cells with fundamentally different fates and functional profiles. Our data show that in the LCMV mouse model of chronic infection and in HIV infection, the CD39high subset of CD8+ T cells demarcates T-betlow /Eomeshigh cells. Consistent with this, CD39+ CD8+ T cells in the mouse model express the highest levels of PD-1, co-express multiple inhibitory receptors and show marked functional defects. These findings suggest that CD39 may be a marker not only of the exhausted state, but specifically of the most terminally exhausted cells, at least in the mouse model. Additional studies of the fate of transferred CD39+ vs. CD39– exhausted CD8+ T cells in the mouse model, and broader surveys of CD39 expression in human chronic infections will be required to determine whether this marker can be used as a surrogate for terminal exhaustion. However, the strong association between CD39 expression and the key features of terminal exhaustion suggests that it may prove a useful marker to help distinguish between" reversible" and" irreversible" T cell exhaustion. Moroever, the fact that isolating CD39+ cells does not require intracellular staining (as is required for T-bet and Eomes) makes this marker useful for studying the function of this terminally exhausted cells ex vivo. The fact that CD39 is expressed by a slightly larger fraction of HCV-specific CD8+ T cells than HIV-specific CD8+ T cells may be related to differences in the timing of blood sampling during the course of infection, or may be due to differences in the extent of antigen-load and inflammation in the two infections. Alternatively, it may be consistent with a model in which HCV-specific CD8+ T cells are in a more “terminal” state of exhaustion than CD8+ T cells specific for HIV. This possibility is supported by profound loss of HCV-specific CD8+ T cells over the course of chronic infection [42] that is not seen in the HIV-specific CD8+ T cell pool, consistent with the clonal deletion seen in mouse models of extreme CD8+ T cell exhaustion [43,44] It is tempting to speculate that expression of CD39 contributes to the dysfunction of exhausted T cells [45]. For instance, the expression of CD39 might enable CD8+ T cells to provide negative regulation in an autocrine or juxtacrine fashion via adenosine [18–20] in the same manner as Tregs [15,35]. The fact that CD39 requires both a substrate (ATP) and a downstream enzyme (CD73) to generate adenosine could provide a mechanism to ensure that this negative signaling occurred only in certain contexts such as in inflamed, damaged tissues where the extracellular concentrations of ATP are high and CD73-expressing cells are present [46]. Moreover, CD39-expressing CD8+ T cells may contribute to the general inhibitory milieu by contributing to the inhibition of activated T cells that express the adenosine receptor but are not yet exhausted. It will therefore be important to determine whether inhibition of CD39 activity could provide an additional therapeutic strategy to rescue the function of exhausted T cells.
Healthy human donors were recruited at the Kraft family Blood Donor Center, Dana-Farber Cancer Institute. All human subjects with HCV infection were recruited at the Gastrointestinal Unit and the Department of Surgery of the Massachusetts General Hospital (Boston, MA) (S1 Table). Individuals with chronic HCV infection (n = 82) were defined by positive anti-HCV antibody and detectable viral load. Patients with spontaneous clearance of HCV, termed resolvers (n = 30), were defined by positive anti-HCV antibody but an undetectable viral load for at least 6 months. The estimated time of infection was calculated either using the exposure date or the time of onset of symptoms and peak ALT (which are assumed to be 7 weeks post infection). All HCV patients were treatment naive and studied at 5. 9 and 219. 7 weeks post infection. HCV RNA levels were determined using the VERSANT HCV RNA 3. 0 (bDNA 3. 0) assay (Bayer Diagnostics). All HIV infected subjects (n = 40) were recruited at the Ragon Institute at the Massachusetts General Hospital (Boston, USA) or the Peter Medawar Building for Pathogen Research (Oxford, UK) (S2 Table). HIV controllers included elite controllers (n = 5) defined as having HIV RNA below the level of detection (<75 viral copies per ml) and viremic controllers (n = 7) with HIV RNA levels < 2,000 viral copies per ml. HIV chronic progressors (n = 28) were defined as having > 2,000 viral copies per ml. All subjects were off therapy. Viral load during chronic infection was measured using the Roche Amplicor version 1. 5 assay. Major histocompatibility complex (MHC) class I HIV Gag-specific tetramers were produced as previously described [47] or obtained from Proimmune. CMV- and EBV-specific MHC class I dextramers conjugated with FITC and APC were purchased from Immudex. Mouse MHC class I tetramers of H-2Db complexed with LCMV GP276-286 were produced as previously described [48,49]. Biotinylated complexes were tetramerized using allophycocyanin-conjugated streptavidin (Molecular Probes). The complete list of multimers can be found in supplemental materials (S3 Table). The following anti-human (hu) and anti-mouse (m) fluorochrome-conjugated antibodies were used for flow cytometry: huCD8α (RPA-T8), huCD4 (OKT4), huCD3 (OKT3), huCD39 (A1), huPD-1 (EG12. 2H7), huCD25 (BC96), huCCR7 (G043H7), huCD45RA (HI100), huT-bet (4B10), mCD8α (53–6. 7), mCD4 (GK1. 5), mCD3 (145-2C11), mCD244. 2 (m2B4 (B6) 458. 1), mPD-1 (RMP1-30), mLag3 (C9B7W), mCD44 (IM7), mCD127 (A7R34), mTNFα (MP6XT22) (all from Biolegend), mT-bet (04–46; BD Pharmingen), mCD39 (24DMS1), mIFN-γ (XMG1. 2), huEomes (WD1928) and mEomes (Dan11mag) (eBioscience). Intracellular staining was performed following surface staining and fixed and permeabilized using the FoxP3/Transcription Factor Staining Buffer Set (eBioscience). Cells were sorted by BD FACS ARIA II and all other analyses were performed on BD LSR II and BD LSR Fortessa flow cytometers equipped with FACSDiva v6. 1. Gates were set using Full Minus One (FMO) controls. Data were analyzed using FlowJo software v9. 8 (Treestar). For intracellular cytokine analysis of mouse T cells, 2x106 splenocytes were cultured in the presence of GP33-41 peptide (0. 2 μg/ml) (sequence KAVYNFATM), brefeldin A (BD), and monensin (BD) for 4. 5 hours at 37°C. Following staining for surface antigens, cells were permeabilized and stained for intracellular cytokines with the Cytofix/Cytoperm kit according to manufacturer' s instructions (BD Biosciences). Wild-type C57BL/6J mice were purchased from The Jackson Laboratory. Female mice (6–8 weeks old) were infected with 2 x 105 plaque forming units (p. f. u.) of LCMV-Armstrong intraperitoneally or 4 x 106 p. f. u. of LCMV-Clone 13 intravenously and analyzed at indicated time points by homogenizing the spleen into a single-cell suspension, Ammonium-Chloride-Potassium lysis of red blood cells, followed by antibody staining. For experiments involving P14 cell transfers, Ly5. 1+ P14s were isolated from peripheral blood, and 500 P14 cells were transferred i. v. into 5–6 week old wild-type female mice one day prior to infection. Viruses were propagated as described previously [48–50]. The concentration of ATP hydrolyzed by CD8+ T cells from subjects with HCV infection (n = 6) was assessed by high performance liquid chromatography (HPLC) as previously described [51]. Briefly, 10,000 CD39+ CD8+ T cells were sorted and placed on ice to minimize ATP production by cells. 20 μM of ATP was added and incubated for 1 h at 37°C in 5% CO2 to allow for cellular activity to increase and CD39-mediated ATP hydrolysis to occur. Samples were then placed in an ice bath for 10 min to halt enzymatic activity, collected, and centrifuged for 10 min at 380 x g and 0°C. Cells were discarded and supernatant centrifuged again to remove remaining cells (2350 x g, 5 min, 0°C). The resulting RPMI samples (160 μl) were treated with 10 μl of an 8 M perchloric acid solution (Sigma-Aldrich) and centrifuged at 15,900 x g for 10 min at 0°C to precipitate proteins. In order to neutralize the pH of the resulting solutions and to remove lipids, supernatants (80 μl) were treated with 4 M K2HPO4 (8 μl) and tri-N-octylamine (50 μl). These samples were mixed with 50 μl of 1,1, 2-trichloro-trifluoroethane and centrifuged (15,900 x g, 10 min, 0°C) and this last lipid extraction step was repeated once. The resulting supernatants were subjected to the following procedure to generate fluorescent etheno-adenine products: 150 μl supernatant (or nucleotide standard solution) was incubated at 72°C for 30 min with 250 mM Na2HPO4 (20 μl) and 1 M chloroacetaldehyde (30 μl; Sigma-Aldrich) in a final reaction volume of 200 μl, resulting in the formation of 1, N6-etheno derivatives as previously described [51]. Samples were placed on ice, alkalinized with 0. 5 M NH4HCO3 (50 μl), filtered with a 1 ml syringe and 0. 45 μM filter and analyzed using a Waters HPLC system and Supelcosil 3 μM LC-18T reverse phase column (Sigma), consisting of a gradient system described previously, a Waters autosampler, and a Waters 474 fluorescence detector [52]. Empower2 software was used for the analysis of data and all samples were compared with water and ATP standard controls as well as a sample with no cells to determine background degradation of ATP. CD8+ T cells from subjects with HCV infection were sorted and pelleted and re-suspended in TRIzol (Invitrogen). RNA extraction was performed using the RNAdvance Tissue Isolation kit (Agencourt). Concentrations of total RNA were determined with a Nanodrop spectrophotometer or Ribogreen RNA quantification kits (Molecular Probes/Invitrogen). RNA purity was determined by Bioanalyzer 2100 traces (Agilent Technologies). Total RNA was amplified with the WT-Ovation Pico RNA Amplification system (NuGEN) according to the manufacturer' s instructions. After fragmentation and biotinylation, cDNA was hybridized to HG-U133A 2. 0 microarrays (Affymetrix). Data have been deposited in Gene Expression Omnibus with accession code GSE72752. Prior to analysis, microarray data were pre-processed and normalized using robust multichip averaging, as previously described [53]. Differentially gene expression and consensus clustering [54] were performed using Gene-E software (www. broadinstitute. org/cancer/software/GENE-E/), and gene set enrichment analysis was performed as described previously using gene sets from MSigDB [55] or published resources [29,32]. Consensus hierarchical clustering was performed using the top 10% of genes that varied across the dataset, without reference to sample identity. Consensus cluster assesses the “stability” of the clusters discovered using unbiased methods such as hierarchical clustering i. e. the robustness of the putative clusters to sampling variability. The basic assumption is that if the data represent a sample of items drawn from distinct sub-populations, a different sample drawn from the same sub-populations, would result in cluster composition and number should not be radically different. Therefore, the more the attained clusters are robust to sampling variability, the greater the likelihood that the observed clusters represent real structure. The result of consensus clustering is a matrix that shows, for each pair of samples, the proportion of clustering runs on sub-sampled data in which those two items cluster together (shown on a scale of 0 to 1). Enrichment Map analysis of GSEA results was performed as described [56]. The gene signature of exhaustion was generated by identifying the top 200 genes upregulated in CD8+ T cells responding to chronic vs. acute LCMV infection in microarray data from a previously published study [29]. All human subjects were recruited with recruited with written informed consent in accordance with Dana-Farber Cancer Institute IRB approval DFCI 00–159, Partners IRB approvals 2010P002121,2010P002463,1999P004983, and Oxford Research Ethics Committee approval 06/Q1604/12. The mouse work was performed under a protocol 01214 approved by the HMA Institutional Animal Care and Use Committee (IACUC), in strict accordance with the recommendations in the Guide for the care and use of Laboratory Animals of the National Institutes of Health. The Harvard Medical School animal management program is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International (AAALAC). | Chronic viral infection induces an acquired state of T cell dysfunction known as exhaustion. Discovering surface markers of exhausted T cells is important for both to identify exhausted T cells as well as to develop potential therapies. We report that the ectonucleotidase CD39 is expressed by T cells specific for chronic viral infections in humans and a mouse model, but is rare in T cells following clearance of acute infections. In the mouse model of chronic viral infection, CD39 demarcates a subpopulation of dysfunctional, exhausted CD8+ T cells with the phenotype of irreversible exhaustion. CD39 expression therefore identifies terminal CD8+ T cell exhaustion in mice and humans, and implicates the purinergic pathway in the regulation of exhaustion. | Abstract
Introduction
Results
Discussion
Materials and Methods | 2015 | CD39 Expression Identifies Terminally Exhausted CD8+ T Cells | 10,006 | 172 |
|
A central issue of myogenesis is the acquisition of identity by individual muscles. In Drosophila, at the time muscle progenitors are singled out, they already express unique combinations of muscle identity genes. This muscle code results from the integration of positional and temporal signalling inputs. Here we identify, by means of loss-of-function and ectopic expression approaches, the Iroquois Complex homeobox genes araucan and caupolican as novel muscle identity genes that confer lateral transverse muscle identity. The acquisition of this fate requires that Araucan/Caupolican repress other muscle identity genes such as slouch and vestigial. In addition, we show that Caupolican-dependent slouch expression depends on the activation state of the Ras/Mitogen Activated Protein Kinase cascade. This provides a comprehensive insight into the way Iroquois genes integrate in muscle progenitors, signalling inputs that modulate gene expression and protein activity.
In Drosophila as in vertebrates the proper function of the muscular system relies on the generation of a stereotyped pattern of discrete muscles and their intimate connection with the nervous system, which together control the adequate release of contraction power to fulfil the functional requirements of the organism. The formation of a muscle pattern is therefore of great importance and consequently many efforts have been devoted to solve the central problem of the acquisition of muscle identity. The embryonic Drosophila muscle pattern comprises thirty elements in each abdominal hemisegment (Figure 1G). Each muscle is a syncytial fibre whose unique characteristics, i. e. , position, size, attachment to tendon cells, innervation and pattern of gene expression allow its unambiguous identification [1], [2]. Muscle specification is a stepwise process that ensures the local singling out of a population of myoblasts, the founder myoblasts, each of them containing the necessary information to give rise to a unique muscle. The origin of founder myoblasts can be traced to late embryonic stage 10 when groups of mesodermal cells (the promuscular clusters) start expressing the proneural gene lethal of scute and acquire myogenic competence [3]. Opposing activities of Notch and Receptor Tyrosine Kinase signalling pathways ensure that only one cell in the cluster will segregate as a muscle progenitor [4]. This will divide asymmetrically to generate two sibling founder myoblasts or a founder myoblast and an adult muscle precursor [3], [5], [6]. The unselected cells of the promuscular clusters, by activation of the Notch signalling pathway, will initiate the expression of the transcriptional regulator Myoblasts incompetent (also called Gleeful and Lame duck) and become fusion competent myoblasts that by fusing to founders will give rise to multinucleated fibres [7]–[9]. Regarding muscle identity, each progenitor and founder exhibits a specific code of gene expression that confers to muscles their unique characteristics. The components of these codes are accordingly named muscle identity genes (reviewed in [2], [10], [11]). The identity code is transmitted to all the nuclei in the syncytium through the process of myoblast fusion [12]. According to their patterns of expression muscle identity genes can be grouped into three categories. Type I includes genes expressed by progenitors and whose expression is maintained in sibling founders and muscles. Examples are apterous, ladybird (lb) and Pox meso (Poxm) [13]–[15]. Type II identity genes are expressed in progenitors but differentially regulated in sibling founders, being lost from one of them and the corresponding muscles. Examples are Krüppel (Kr), even-skipped (eve), collier and slouch (slou) [3], [4], [16]–[18]. And finally type III refers to genes expressed by progenitors and founders of muscles sharing common characteristics. vestigial (vg), expressed by all internal muscles, is the only known member of this class [12], [19]. Regarding the onset of their expression a few muscle identity genes, such as Kr, eve and collier, are already expressed in the promuscular cluster, before the segregation of muscle progenitors [4], [16], [18], [20] whereas other genes, like Connectin (Con), initiate their expression in already segregated progenitors [4], [16], [18], [20]. In this study we identify araucan (ara) and caupolican (caup), two members of the Iroquois gene complex (Iro-C), as novel type III muscle identity genes. The Iro-C genes encode homeoproteins conserved throughout the animal kingdom. They are organized in genomic clusters of three paralogous genes, one in the case of Drosophila and usually two in most vertebrates [21]. They participate in a wide variety of developmental processes, mainly related to the specification and patterning of diverse territories of the body, including the lateral mesonotum and dorsal cephalic region of Drosophila, the neural ectoderm of Xenopus and cranial placode derivatives of zebrafish [22]–[30]. Here we show by means of genetic approaches that ara and caup function redundantly in the specification of the lateral transverse (LT) muscles, since in the absence of both genes LT1–4 muscles loose their LT fates and acquire those of other muscles. At present there is compelling evidence that muscle progenitors can integrate positional and temporal signalling inputs. This promotes the expression of unique combinations of muscle identity genes, which confers on them their ultimate fate [14]–[18], [31], [32]. There has been extensive analysis on the regulation of some of these genes, such as eve and collier [4], [33], [34], which has allowed to propose candidate cis-regulatory modules for founder muscle specific expression [35]. However, very little is known about how progenitors integrate the activity of the transcription factors encoded by these genes, about the identity of their direct targets (save in the cases of Kr and Lb [36]–[38]), and of their hierarchical relationships and their putative post-transcriptional regulation. In this report we have focused on these issues in relation to the function of the ara/caup identity genes. We demonstrate that the implementation of the lateral transverse muscle fate requires the repression mediated by Ara/Caup of the muscle identity genes slou and vg, to avoid reiteration of other muscle fates regulated by these transcription factors. In addition, we identify slou as a potential direct target of Ara/Caup. Furthermore, our tissue culture and in vivo experiments show that the repression of slou by Ara/Caup in LT precursors requires the activity of the Ras/Mitogen Activated Protein Kinase (Ras/MAPK) pathway to be kept at a very low level, since otherwise Caup is converted from a repressor to an activator of slou. This is to our knowledge the first evidence of the interplay between the Receptor Tyrosine Kinase signalling pathways and the activity of a muscle identity transcription factor. Therefore, during Drosophila embryogenesis, and for the acquisition of the lateral transverse muscle fate, the homeoproteins Ara and Caup appear to act at a nodal point in muscle progenitors, as they integrate positional and temporal signalling inputs that modulate their activity on subordinate identity genes.
The patterns of expression of ara and caup in the embryonic ectoderm have been previously reported [39], [40]. In this work we focus on the embryonic ara and caup mesodermal expression. In situ hybridization showed that here both genes were similarly expressed (Figure 1 and results not shown). At early stage 11 caup (and ara) transcripts and proteins are detected in groups of cells of the presumptive visceral trunk mesoderm (Figure 1A-1A′″, the available anti-Caup antibody recognises both Ara and Caup proteins). By mid stage 11 they are expressed at the same dorso-ventral level in the visceral mesoderm and in the dorsolateral ectoderm (Figure 1B-1B′″). Expression in the visceral mesoderm declined at late stage 11 when it became detectable in groups of cells of the somatic mesoderm (promuscular clusters [3], Figure 1C and 1C′), from where a subset of muscle progenitors (P) still expressing ara/caup, will segregate slightly later (stage 12, Figure 1D). Expression was maintained in sibling founder myoblasts (Fs in Figure 1D′) derived from ara-caup-expressing progenitors and in the muscles they give rise to (Figure 1E-1E″), namely LT1–4, dorsal transverse 1 (DT1) and segment border muscle (SBM) (Figure 1F and 1G). The expression in the somatic mesoderm of the third member of the Iroquois complex, mirror [41] did not overlap with that of ara-caup (not shown). The early expression of ara/caup in all lateral muscles with vertical orientation, suggested a possible role as muscle identity genes. Therefore, we compared their expression with that of several muscle identity genes. For the LT1–4 muscles, ara/caup were co-expressed with Kr in the promuscular clusters from which progenitors PLT1/LT2 and PLT3/LT4 are singled out (Figure 2A). ara/caup expression was maintained at high levels in both progenitors that also express Kr (Figure 2B). Whereas Kr expression decayed in founders LT1 and LT3 before the onset of myoblast fusion and in LT2 and LT4 muscles from stage 15 onwards [18], expression of ara/caup was maintained in the four founders (Figure 2D and 2E). These also expressed Con, co-expression that was maintained in the mature LT1–4 muscles (Figure 1G). In the case of muscle DT1, the onset of ara/caup expression coincided with that of Con and slou in the progenitor of DT1 and dorsal oblique 3 (DO3) muscles (Figure 2C) and it appeared to be maintained in DT1 founder (Figure 2E) and mature muscle at low levels (Figure 1G). Finally, ara/caup co-expressed with lb in the SBM founder (Figure 2D), but were not be detected in the lb-expressing progenitor and promuscular cluster. In summary, different muscle lineages expressed ara/caup at different steps of the myogenic programme (Figure 2F). In the LT1–4 case ara/caup and Kr were detected at the earliest lineage stage, that is in promuscular clusters, preceding Con expression in progenitors (not shown); in the DT1/DO3 lineage ara/caup and slou were first detected in the already singled out DT1/DO3 progenitor and in the SBM ara/caup expression was first detected in the SBM founder after lb expression. During imaginal development Ara and Caup can functionally substitute each other in all territories where their function has been investigated [22], [23], [28]. Thus, to analyse their role in embryonic myogenesis and evaluate the possible contribution of mirror to any phenotype we might find, we used three deficiencies: Df (3L) iroDFM3, which removes both ara and caup, (and probably affects mirror regulation, [23], [28]), Df (3L) iroEGP6, which removes ara and caup without affecting mirror and its regulatory region, and Df (3L) iroEGP5, which only removes mirror [42]. Whereas Df (3L) iroEGP5 embryos did not show any detectable phenotype in the lateral region (not shown), a distortion of the lateral larval muscle pattern (visualised with antibody MAC141 to Tropomyosin) was found in both Df (3L) iroDFM3 and Df (3L) iroEGP6 embryos (Figure 3A–3C). In more than 95% of cases muscles with LT morphology were absent (Figure 3E). Instead, some fibres with abnormal orientation appeared in the lateral and ventral regions, but never inserted at the LT attachment sites (asterisks in Figure 3B and 3C). The loss of LT muscles was further verified by loss of expression of the specific LT muscle marker CG13424, recently renamed lateral muscles scarcer (lms) [43] at stage 15 and the absence of Con expression in the lateral somatic mesoderm (Figure S1). Both DT1 and SBM fibres developed with normal morphologies (Figure 3A–3C and Figure S1). To examine the individual contribution of ara and caup to the phenotype we resorted to embryos mutant for only one of these genes (ara in ararF209, [28], or caup in iroEGPΔ1, [42]). The larval muscle pattern was normal in both mutants (not shown). Thus, similarly to imaginal development, ara and caup appear to play redundant roles during embryonic myogenesis. The absence of muscles with LT morphology in ara/caup mutants could be due to a failure of otherwise well specified muscles to find the right insertion to tendon cells, due to ectodermal requirement of Iro-C genes, or to a misspecification of the muscles. Two independent results indicated that Iro-C genes are required autonomously in the mesoderm to specify the LT fate. First, the normal development of LT muscles in Df (2L) 5 embryos devoid of Iro-C gene expression at the ectoderm (Figure S2 and [44]). And second, the rescue of the muscle phenotype of Df (3L) iroDFM3 embryos by Ara supplied exclusively in the mesoderm (using myocyte enhancer factor 2 (mef2) -GAL4 as driver, Figure 3D). We next examined whether the loss of LT muscles was due to either a failure in the segregation of muscle progenitors (absences and/or duplications) or to an early transformation of the fate of LT progenitors. To discern between these possibilities we combined the reporter line rP298, which expresses ß-galactosidase in all progenitors and founders [32], [45] with Df (3L) iroEGP6. We focussed on the previously well-established muscle lineages labelled by Slou/S59 [3], [17] and the LT1–4 lineages labelled by Kr [18]. With these markers in the lateral-ventral region of rP298 embryos we can identify the following founders (Figure 4A-4A″ and insets below). In the dorsalmost lateral mesoderm we find the sibling founders DT1 and DO3 (expressing slou) and the lateral longitudinal 1 (LL1) founder and its sibling (expressing Kr). Immediately below segregate the four LT founders (expressing Kr). And more ventrally appear the sibling ventral acute 1 (VA1) and VA2 founders (which express Kr and slou) and the VA3 founder and its sibling, the ventral adult precursor (that express slou). In Df (3L) iroEGP6 embryos we observed the same number of identifiable founders (Figure 4B and 4B′). There were however significant differences in terms of patterns of gene expression. Namely, the presumptive LT3–4 founders now expressed slou in addition to Kr (Figure 4B and 4B′ and insets below). This code of muscle identity gene expression is similar to that of VA1 and VA2 founders (Figure 5A), suggesting an early transformation of LT3–4 to VA1–2 muscles. The absence of all muscles with LT morphology in ara/caup mutant embryos prompted us to examine whether, in addition to the putative transformation of LT3–4 towards VA1–2, there was a similar change of fate for LT1–2. LT progenitors express Kr, caup, Con and lms, PLL1/LL1sib expresses Kr and vg, and PVA1/2 Kr, slou, Con and Poxm (Figure 5A and 5B). Using a combination of these markers we found in the lateral region of Df (3L) iroDFM3 embryos an ectopic muscle that expressed Kr+Vg, the code of LL1 (LL1*, Figure 5C) and an ectopic muscle VA2 (VA2* in Figure 5E–5G). This change of muscle identity could take place in founders or at the progenitor state. If this were the case, we anticipated that both muscles resulting from sibling founder myoblasts should be duplicated in Df (3L) iroDFM3 embryos. Indeed, using anti-Poxm, which labels VA1–3 ([14] and Figure 5B), and antibodies to Kr and Slou, which are maintained only in VA2 (Figure 5B and 5D), we identified two VA2 muscles (that co-express Poxm and Kr) and two Poxm-expressing VA1 muscles in late stage 14 Df (3L) iroDFM3 embryos (Figure 5F). The presence of the duplicated VA1 and VA2 muscles was more evident at stage 15 when Poxm was only weakly expressed in VA2 muscles (Figure 5G). We concluded that Ara and Caup were required to specify LT progenitors and that implementation of this fate implies the repression of specific muscle identity genes, such as slou in PLT3/4 and vg in PLT1/2. Moreover, it seemed that the only muscles affected by the lack of ara/caup were those in which these genes were already expressed in the corresponding promuscular clusters, since the fate of DT1 and SBM, visualised by the expression of slou, Con and lb, was apparently unaffected in Df (3L) iroDFM3 embryos (Figure 2F, Figure 5D and 5E, and Figure S1E–S1H). Our data suggested that Ara/Caup might act as repressors of slou in the Drosophila mesoderm. Therefore we decided to investigate whether slou might be a direct target of Ara/Caup. An “in silico” search of a previously reported slou cis-regulatory region [46] identified two putative Iro binding sites (BS) at positions +129 (BS1) and −1642 (BS2), relative to the transcription start site, which match the consensus ACAN2–8TGT ([47] and Figure 6A). We cloned this regulatory region in a Luciferase reporter vector and measured Luciferase activity in Drosophila Schneider-2 (S2) cells transiently transfected with this construct and increasing amounts of HA-tagged Caup. Contrary to expectations, we found that addition of Caup-HA increased the basal Luciferase activity driven by the slou regulatory region in a dose dependent manner (blue bars in Figure 6B), indicating that Caup acts as a transcriptional activator of slou under these conditions. The reported regulation of the chicken Irx2 factor by MAPK (that switches it from repressor to activator) could explain this result [48]. Since Western Blot analysis of S2 lysates using an antibody against diphospho-extracellular-signal related kinase (dpErk) showed the MAPK pathway to be active in S2 cells (Figure 6C) and we have obtained experimental evidence showing the presence of phosphorylated Caup in S2 cells with constitutively active MAPK pathway (N. B, A. S. T and S. C, manuscript in preparation), we hypothesized that the activation effect of Caup in S2 cells could be due to the Ras/MAPK cascade turning Caup from transcriptional repressor into activator. Indeed, the inhibition of the Ras/MAPK pathway by the PD98059 MAP-erk kinase-1 (MEK1) inhibitor induced a Caup-dose dependent decrease in Luciferase activity driven by the slou regulatory sequences (Figure 6B, red bars). This result could not be attributed to a direct effect of the inhibitor over the slou promoter, since its addition did not modify the basal Luciferase activity of the construct (Figure 6B). To test whether Caup-dependent transcriptional regulation relied on a direct interaction of Caup with the slou regulatory region we performed electrophoretic mobility shift assays (EMSA) with in vitro translated Caup and wild-type and mutated Caup-BS. These assays indicate efficient binding of Caup to BS1, which is abolished by BS1 mutation and deletion (Figure 6D). In contrast, Caup appears not to bind BS2 under these experimental conditions (not shown). Next we examined the functional relevance of BS1 and BS2 in the Luciferase reporter assay. Deletion of BS2 had no major effect on Caup-dependent luciferase expression compared to the wild-type promoter (Figure 6E and 6F compare with Figure 6B). This result suggested that Caup does not bind to BS2 (as indicated by the EMSA data). Unexpectedly, deletion of BS1 resulted in a more efficient activation of luciferase expression than that driven by the wild type regulatory region (Figure 6E). This suggested that binding of Caup to BS1 somehow impaired transcription. Note that the activation of luciferase driven by the BS1 mutated regulatory region was still dependent on the MAPK pathway (Figure 6E and 6F). This suggests that such activation appears to depend on the binding of a MAPK-dependent phosphorylated protein, which we hypothesize might be Caup, to a so far unidentified binding site. Thus, the analysis in S2 cells confirmed the relevance of BS1, but not of BS2 on Caup-dependent regulation. Additionally, we have analysed the evolutionary conservation of these putative Caup-BS among several Drosophila species (Figure S3). Notably, BS1 is located in a highly conserved region and its sequence is identical across the melanogaster group, whereas neither BS2 nor the adjacent sequences are conserved. These data further reinforce the relevance of BS1 for Caup-dependent slou regulation. Our results are thus consistent with a direct effect of Caup on slou regulation. However, it cannot be ruled out the possibility of the existence in vivo of a transcription factor, acting downstream of ara/caup, that could repress slou through BS1 or through a still unidentified regulatory sequence of slou. To further examine in vivo the regulatory activity of Caup on slou (Figure 7B, 7C), we ectopically expressed caup or ara in VA1–3 using Con-GAL4 and checked whether they would repress slou in the VA2 muscle. This was indeed the case (Figure 7B, 7D, 7F-7F″ and not shown). Loss of slou expression caused by ectopic caup reproduced the morphological defects in VA2 previously described in slou mutants (Figure 7F-7F″ and [17]). To analyse whether the morphological effect of Caup on muscle VA2 development was only due to Caup-dependent repression of slou, we forced the expression of both genes using the Con-GAL4 driver. In this experimental condition Caup was unable to repress UAS-slou expression and the VA2 muscle and its morphology seemed unaffected (Figure 7F-7G″). Once verified the repressor activity of Caup on slou during myogenesis, to analyse the regulatory potential of BS1 in vivo we generated transgenic flies harbouring the wild-type or the BS1 deleted version of the slou regulatory region. The wild-type regulatory region only partially reproduced the slou endogenous expression, as it drove lacZ expression in the CNS but not in the muscles (not shown and Figure S4). In contrast, the construct lacking BS1 behaved congruently with our S2 cells results, since it drove ectopic expression of lacZ in the lateral muscles (Figure S4). Curiously, up-regulation of lacZ was found in the 4 lateral muscles and not only in the ones that show slou expression in the absence of Ara/Caup (Figure 4B). Thus we interpret that this construct, while missing some of the regulatory sequences required for slou mesodermal expression, it contains those required for Caup mediated repression in the mesoderm. In addition, the absence of strict correlation between the phenotypes of deletion of BS1 and lack of Ara/Caup, might indicate the ability of other transcription factor (s) to regulate slou expression in LT1–2 through BS1. To investigate whether the effect of the MAPK cascade on the transcriptional activity of Ara/Caup found in the S2 cell assay is also at work during myogenesis we examined whether there is a correlation between MAPK signalling and Caup transcriptional regulatory activity. We looked at the state of activation of this pathway in the LT promuscular cluster, where Ara/Caup repress slou, and found that it did not appreciably express dpErk (Figure 7H). Therefore, a repressor activity of Ara/Caup correlates in vivo with the absence of MAPK signalling. Next, we tested whether forced activation of the MAPK pathway in the mesoderm could interfere with the repressor activity of endogenous Caup in LT promuscular clusters. This was indeed the case, since activation of the MAPK pathway using twist-GAL4; 24B-GAL4 to drive the activated form of Ras85D (rasV12 [49]) allowed co-expression of caup and slou in this cluster (Figure 7I). Similarly, late co-expression of rasV12 and caup (Con-Gal4 driver) blocked the repression activity of Caup on slou (Figure 7D and 7E). Finally, to test whether MAPK signalling not only prevented Caup-dependent repression of slou but also converted Caup from repressor to activator, we looked at the expression of slou after early pan-mesodermal Caup expression (mef2-GAL4). As shown in Figure 7J, Ara was indeed able to ectopically activate slou in Drosophila epidermal growth factor receptor (DER) -dependent eve-expressing muscles.
The study of myogenesis in Drosophila has increased the understanding of how the mechanisms that underlie the acquisition of specific properties by individual muscles are integrated within the myogenic terminal differentiation pathway. Thus, the current hypothesis proposes that distinct combinations of regulatory inputs leads to the activation of specific sets of muscle identity genes in progenitors that regulate the expression of a battery of downstream target genes responsible for executing the different developmental programmes (reviewed in [2], [10], [38]). However, the analysis of the specific role of individual muscle identity genes and of their hierarchical relationships is far from complete since the characterisation of direct targets for these transcriptional regulators is very scarce [36], [37]. Here, we report the identification of ara and caup, two members of the Iroquois complex, as novel type III muscle identity genes. We find that the homeodomain-containing Ara and Caup proteins are necessary for the specification of the LT fate. ara/caup appear to be bona fide muscle identity genes. Indeed, similarly to the identity genes Kr and slou [17], [18], absence of ara/caup does not interfere with the segregation of muscle progenitors or their terminal differentiation, but modifies the specific characteristics of LT1–4 muscles, which are transformed towards VA1, VA2, LL1 and LL1 sib fates. These transformations may be due in part to the up-regulation of slou and vg in the corresponding muscles. Thus, a recent report [50] shows that forced expression of vg in LT muscles induces changes in muscle attachments similar to the ones observed in LT1 in ara/caup mutant embryos. However, it should be stressed that although in ara/caup mutants LT muscles are lost in more than 95% of cases, they are not completely transformed into perfect duplicates of the newly acquired fates. For instance, while the specific LT marker lms is lost in 91% of cases, ectopic slou expression is detected in only 75% of cases. These partial transformations might be due to differences in the signalling inputs acting in the mesodermal region from where these muscles segregate (see below). Our unpublished data also showed that forced pan-mesodermal expression of ara/caup alter the fates of many muscles both in dorsal and in ventral regions without converting them into LT muscles (i. e. , they do not ectopically express lms). Similarly, Kr and slou ectopic expression is not sufficient to implement a certain muscle fate [17], [18]. The failure to recreate a given muscle identity by adding just one of the relevant muscle identity proteins reveals the importance that cell context, that is, the specific combination of signalling inputs and gene regulators present in each cell, have in determining a specific muscle identity. Our analysis of the myogenic requirement of ara/caup has revealed several features about how these genes act to implement LT fates. Thus, although they are expressed in six developing embryonic muscles, only four of them, LT1–4, are miss-specified in the absence of Ara/Caup. The remaining two, DT1 and SBM, seem to develop correctly, according to morphological as well as molecular criteria. It is worth noting that the requirement for ara/caup genes in these six muscles correlates with the onset of their expression. Thus, in the affected LT1–4 muscles Ara/Caup can be first detected at the earliest step of muscle lineages, that is in the promuscular clusters. In contrast, in the unaffected muscles ara/caup start to be expressed later, in the DT1/DO3 progenitor and the SBM founder. This suggests that in muscle lineages ara/caup have to be expressed very early to repress slou and vg to implement the LT fate. Several data support this interpretation. For instance, the observation that ara/caup are co-expressed with slou in DT1, whereas they repress slou in LT3–4, may be related to the fact that slou expression precedes that of ara/caup in the DT1 lineage. Should this be so, one would expect that ectopic expression of ara using the early driver mef2-GAL4, would repress slou in DT1, as it actually does (Figure S5), whereas this repression is not evident using the late driver Con-GAL4. Furthermore, the hypothesis of the relevance of the timing of muscle identity gene expression for muscle fate specification might also apply to the case of slou, where a similar correlation between the strength of the loss-of-function slou phenotypes in specific muscles and the onset of slou expression has also been found [17]. It should be stressed that the generation of the LT code depends not only on the early presence of Ara/Caup on the promuscular clusters but also on the absence (or strong reduction) of DER/Ras activity at that precise developmental stage and location (Figure 8). There is a dynamic regulation of MAPK signalling in the lateral mesoderm. Caup-expressing muscles develop from DER-independent clusters whereas the duplicated muscles observed in ara/caup mutants derive from progenitors that segregate very near the LT progenitors [3], but originate in DER-dependent promuscular clusters that are specified slightly later in development [4], [51]. Furthermore we have observed both by in vivo and in cell culture that low MAPK activity is required for Caup-dependent slou repression. Therefore, we interpret the role of Ara/Caup in the implementation of LT fate as follows (Figure 8). At mid stage 11 in the myogenic mesoderm, groups of mesodermal cells acquire myogenic competence as a result of interpreting a combinatorial signalling code that reflects their position along the main body axes, as well as the state of activation of different signalling pathways [4]. Accordingly, these clusters initiate the expression of lethal of scute and a unique code of muscle identity genes, as has been shown in great detail for eve expression in the dorsal mesoderm [34], [35]. In the case of the dorso-lateral mesoderm this code includes ara/caup and Kr and implements the LT fate. Since the level of activation of the Ras/MAPK cascade is low in these clusters, Ara/Caup will behave as transcriptional repressors, preventing the activation of slou or vg in LT1–2 and LT3–4 clusters, which would be otherwise activated in this location. Thus, Ara/Caup implement the LT fate by repressing the execution of the alternative fates (Kr+, Slou+, Con+, Poxm+ and Kr+, Vg+) that would give rise to duplicates of PVA1/VA2 and PLL1/LL1sib, respectively, and by allowing a different identity gene code (Kr+, Caup+, Con+, lms+) that generates the LT fate. Slightly later the Ras/MAPK pathway becomes active at the dorsolateral region (Figure 8). This changes the combinatorial signalling code and coincides with a change in the muscle identity genes expressed by the promuscular clusters that segregate from this position, which now accumulate Kr but not Ara/Caup. Progenitors born from them will express either slou or vg and give rise to VA1–2 and LL1/LL1sib fates, all DER-dependent [51]. Our S2 cells experiments suggest a molecular mechanism by which the Ras/MAPK pathway modulates the transcriptional activity of Ara/Caup on slou. Thus, low MAPK activity and direct binding of Caup to BS1 site of the slou gene would favour strong repression of slou. BS1 could be embedded in a silencer regulatory element or its binding to Caup may block transcription of the downstream located luciferase gene. On the contrary, Caup-dependent activation of slou would be dependent on MAPK signalling. We hypothesize that MAPK–dependent Caup phosphorylation could modulate its interaction with different transcriptional co-factors or/and its binding site affinity. Furthermore, our in vivo evidence indicates a repressor function of presumably non-phosphorylated Caup on slou since forced activation of the Ras pathway allows co-expression of slou and caup. On the other hand, the ectopic expression of slou induced by caup-over-expression is suggestive of a possible activator function of phosphorylated Caup. The role of IRO proteins in cell fate specification is conserved in both vertebrates and invertebrates (reviewed in [52]). Here we have shown that the interplay between MAPK signalling and IRO activity found in vertebrate neuroepithelium [48] is also at work in Drosophila myogenesis. We have also identified a potential direct target of Ara/Caup, slou and propose vg as a candidate gene to be regulated by Ara/Caup. In both cases the genes subordinated to ara/caup encode transcription factors that might in turn regulate the expression of other genes, genes that must be repressed in LT muscles in order to acquire the LT fate. These results, therefore, provide insights into the way Ara/Caup control lateral muscle identity and on the role of signalling pathway inputs to modulate the activity of these transcription factors, with consequences in their downstream targets. It also highlights the importance that the specific combination of muscle identity genes, their hierarchical relationships and their temporal activation have in determining the identity of a given muscle cell, very alike to what is at work during the acquisition of neural fates [53].
The following stocks were used: Df (3L) iroDFM3, ararF209 [28], Df (2L) 5 [54], Df (3L) iroEGP6, Df (3L) iroEGP5, Df (3L) iroEGPΔ1 [42], rP298 [32], mef2-GAL4 [55], Con-GAL4 [56], twist-GAL4; 24B-GAL-4 (a gift from M. Baylies), UAS-ara, UAS-caup [28], UAS-caup-HA (N. Barrios, unpublished) and UAS-rasV12 [49]. Ectopic expression was generated by means of the GAL4/UAS system [57]. Whole-mount in situ hybridisation with digoxygenin-labelled RNA probes and immunocytochemistry were performed as described previously [58]. Stained embryos were embedded in Araldite and sectioned (3 µm) following standard procedures. The following primary antibodies were used at the indicated dilutions: rat anti-Caup (1∶50) [23], guinea pig anti-Kr (1∶500) [59], mouse anti-Lb (1∶1) [15], rabbit anti-Poxm (1∶10) [14], rat and rabbit S59 (that recognises Slou, 1∶50) [3], rabbit anti-Alien (1∶500) [60], mouse anti-Con (1∶10) [61], rabbit anti-Vg (1∶500) [62], rat-anti- Tropomyosin (MAC141; 1∶100; Babraham Tech), rabbit anti-Myosin (Myo; 1∶300) [63], rat anti-HA (1∶1000; Roche); rabbit anti-ß-Gal (1∶5000; Cappel) and mouse anti- dpErk (1∶50; Sigma). Images were obtained with confocal microscopes MicroRadiance (BioRad) and LSM510META (Zeiss) and analysed using the software Zeiss LSM Image or LaserSharp and Adobe Photoshop 7. 0. In most cases the figures correspond to z-projections of series of confocal sections. The 5′-upstream region of slou (from −1828 to +153 nt) was amplified via PCR and cloned in pGLHS43 vector, a modified version of the pGL2-basic vector (Luciferase reporter plasmid, Promega), obtained after substitution of the SV40 promoter by the Drosophila heat-shock 43 minimal promoter (a gift from A. Baonza). The putative Caup BS1 and BS2 were deleted using the “Quick Change” site-directed mutagenesis kit (Stratagene, SantaClara, CA). The sequences of the primers used to delete BS1 were 5′-GAGTTCTTAATCCAGCCGTGTTGTGTGCCTGTGGCAAGTCAATAG-3′ and its reverse complement and for BS2,5′-CCATATACATATGTGTGCATGTATGCATAAGTGTGAGTGTGAGTGGG -3′ and its reverse complement. pAC5. 1-Caup-HA plasmid was obtained after cloning caup ORF with an HA tag in the Drosophila expression vector pAC5. 1 (Invitrogen). Drosophila S2 cells were cultured in Insect-Xpress medium (Lonza) supplemented with 7% fetal bovine serum and grown at 25°C. For Luciferase assays S2 cells were seeded at a density of 2×106 and co-transfected with 1 µg of the different firefly Luciferase reporter constructs DNA, 30 ng of control plasmid (expressing Renilla Luciferase driven by the promoter of Drosophila RpIII128, [64]) and either 0,0. 25,0. 5 or 1 µg of pAC5. 1-Caupolican-HA plasmid per well using Nucleofector Technology (Lonza). Luciferase activity in the cell extracts was measured using Dual-Glo Luciferase assay system (Promega) following the manufacturer' s protocol. Briefly, 20 µl extract was added to 100 µl F-luc assay reagent, mixed gently for 5 s and placed in a luminometer. After counting F-luc activity for 10 s, 100 µl stop-and-glo reagent was added to the tube, mixed gently for 5 s and placed in the luminometer for R-luc count. The R-luc activities were used as internal control to correct for the difference in transfection efficiency of different reporter plasmids. Therefore, F-Luc/R-Luc activities were used for data analysis. To investigate whether the MEK/ERK pathway was involved in transcriptional regulation driven by the slou promoter, S2 cells were treated or not with 50 µM PD-98059 (Sigma) for 2 hrs before Luciferase activity measurement. All data reported are means from three or four independent experiments, each performed in triplicates. Primary antibodies used in immunoblots were mouse anti-dpErk (1 µg/ml; Sigma), rat anti-HA (200 ng/ml; Roche) and mouse anti-βtubuline (1∶5000; Developmental Studies Hybridoma Bank). The 5′-upstream region of slou used in S2 cells in the Luciferase reporter assays (both the wild type sequence and that missing the putative Caup BS1) were subcloned at the EcoRI site of the C4PLZ enhancer tester plasmid that contains a weak P-element promoter [65]. These lacZ reporter plasmids were introduced into y w1118 embryos by standard P-element transformation. Caup binding ability to the slou promoter region was analyzed by EMSA. Pairs of single-stranded, Cyc3 and unlabeled 40-mer oligonucleotides containing the wild-type putative Caup binding sites BS1, BS2 and their mutant or deleted versions were allowed to anneal to generate double-stranded probes. Sequences of primers are shown in Figure 6D for BS1 and in Dataset S1. Caup protein was synthesized in vitro by using the coupled transcription/translation rabbit reticulocyte lysate system (TNT Promega). The indicated amount of µl of TNT reaction mixture was incubated with 20 ng of labelled probe. Protein–DNA complexes were allowed to form at room temperature for 30 min in a total volume of 20 µl of binding buffer (50 mM HEPES, pH 7. 5,10 mM MgCl2,10 mM KCl and 1 mM DTT). After incubation, free DNA and protein–DNA complexes were resolved by 6% non-denaturing polyacrylamide gel electrophoresis. Gel fluorescence was analyzed in a Typhoon Scanner (GE healthcare). | In Drosophila, as in vertebrates, the muscular system consists of different types of muscles that must act in coordination with the nervous system to control the adequate release of contraction power required for the proper functioning of the organism. Therefore, the acquisition of specific identities by individual muscles is a key step in the generation of the muscular system. In Drosophila, muscle progenitors (specific myoblasts that seed the formation of mature muscles) integrate positional and temporal signalling inputs, resulting in the expression of unique combinations of muscle identity genes, which confer on them specific fates. Up to now, very little was known of how this integration takes place at a molecular level and how a particular code is translated into a specific muscle fate. Here we show that the acquisition of the lateral transverse muscle fate requires the repression mediated by Araucan and Caupolican, two homeoproteins of the Iroquois Complex, of other muscle identity genes, like slouch and vestigial. The repressor or activator function of the Iroquois proteins depends on the activity of the Ras signalling pathway. Therefore, our work places Iroquois genes at a nodal point that integrates signalling inputs and regulates protein activity and cell fate determination. | Abstract
Introduction
Results
Discussion
Materials and Methods | developmental biology
model organisms
genetics
biology
molecular cell biology
genetics and genomics | 2011 | Drosophila Araucan and Caupolican Integrate Intrinsic and Signalling Inputs for the Acquisition by Muscle Progenitors of the Lateral Transverse Fate | 10,469 | 289 |
RNA modification plays an important role in modulating host-pathogen interaction. Flavivirus NS5 protein encodes N-7 and 2′-O methyltransferase activities that are required for the formation of 5′ type I cap (m7GpppAm) of viral RNA genome. Here we reported, for the first time, that flavivirus NS5 has a novel internal RNA methylation activity. Recombinant NS5 proteins of West Nile virus and Dengue virus (serotype 4; DENV-4) specifically methylates polyA, but not polyG, polyC, or polyU, indicating that the methylation occurs at adenosine residue. RNAs with internal adenosines substituted with 2′-O-methyladenosines are not active substrates for internal methylation, whereas RNAs with adenosines substituted with N6-methyladenosines can be efficiently methylated, suggesting that the internal methylation occurs at the 2′-OH position of adenosine. Mass spectroscopic analysis further demonstrated that the internal methylation product is 2′-O-methyladenosine. Importantly, genomic RNA purified from DENV virion contains 2′-O-methyladenosine. The 2′-O methylation of internal adenosine does not require specific RNA sequence since recombinant methyltransferase of DENV-4 can efficiently methylate RNAs spanning different regions of viral genome, host ribosomal RNAs, and polyA. Structure-based mutagenesis results indicate that K61-D146-K181-E217 tetrad of DENV-4 methyltransferase forms the active site of internal methylation activity; in addition, distinct residues within the methyl donor (S-adenosyl-L-methionine) pocket, GTP pocket, and RNA-binding site are critical for the internal methylation activity. Functional analysis using flavivirus replicon and genome-length RNAs showed that internal methylation attenuated viral RNA translation and replication. Polymerase assay revealed that internal 2′-O-methyladenosine reduces the efficiency of RNA elongation. Collectively, our results demonstrate that flavivirus NS5 performs 2′-O methylation of internal adenosine of viral RNA in vivo and host ribosomal RNAs in vitro.
Many members within the Flavivirus genus from Flaviviridae family are important human pathogens, including the four serotypes of Dengue virus (DENV-1 to -4), yellow fever virus (YFV), West Nile virus (WNV), Japanese encephalitis virus (JEV), and tick-borne encephalitis virus (TBEV). These viruses are naturally transmitted by mosquitoes or ticks, causing global burden and threat to public health [1]. The flaviviral genome is a plus-sense RNA of about 11 kb in length. The 5′ end of the flavivirus genome contains a type I cap, followed by the conserved dinucleotide sequence AG (m7GpppAmG). The genomic RNA consists of a 5′ untranslated region (UTR), a single open-reading-frame, and a 3′ UTR. The open-reading-frame encodes a long polyprotein that is processed by viral and host proteases into three structural proteins (capsid [C], premembrane [prM], and envelope [E]) and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) [2]. Structural proteins form viral particles, and participate in virus entry and assembly. Nonstructural proteins function in viral RNA replication [2], evasion of innate immune response [3]–[6], as well as virus assembly [7], [8]. Two flavivirus nonstructural proteins have enzymatic activities. NS3 functions as a viral serine protease (together with NS2B as a cofactor) [9], [10], a NTPase [11], an RNA triphophatase [12], and an RNA helicase [13]. NS5 acts as a methyltransferase (MTase) [14], [15] and an RNA-dependant RNA polymerase (RdRp) [16], [17]. We previously showed that the N-terminal domain of flaviviral NS5 protein posses both N-7 and 2′-O methylation activities required for the formation of 5′ RNA cap [15]. The MTase catalyzes the two distinct methylation reactions in a sequential manner, GpppA-RNA→m7GpppA-RNA→m7GpppAm-RNA. Both reactions use S-adenosyl-L-methionine (SAM) as the methyl donor and generate S-adenosyl-L-homocysteine (SAH) as a by-product. The order of two sequential methylations is dictated by the fact that the 2′-O methylation reaction prefers the substrate m7GpppA-RNA to GpppA-RNA, whereas the N7 methylation reaction has no preference between substrates GpppA-RNA and GpppAm-RNA [18]. Biochemical and structural studies indicate that flaviviral MTase catalyses the N7 and 2′-O methylations through an RNA cap-repositioning mechanism [18], [19]. Functional analysis showed that the N7 methylation of flaviviral RNA cap is critical for efficient translation [15], whereas the 2′-O methylation functions in subverting innate host antiviral response through escape of IFIT-mediated suppression [20]. Most eukaryotic mRNAs contain co- or post-transcriptional modifications, including the 5′ cap structure, internal bases methylation, splicing of introns, and polyadenylation. N6-methyladenosine (m6A) represents a major internal modified nucleoside. The m6A is found in cellular mRNAs from mammals, plants, insects, and yeast [21]–[26] as well as in some viral RNAs [27]–[29]. The m6A modification functions in mRNA processing [25], [30], intracellular transporting, and translation [31]. Besides m6A, 2′-O methylation of ribose represents another common internal nucleoside modification. The 2′-O methylation is found in splicesomal small nuclear RNAs (snRNAs) and ribosomal RNAs [32]. Although the exact function of internal 2′-O methylations remains elusive, these modifications are clustered in regions of functional importance, such as regions engaged in RNA-RNA interactions [32]. The distinct chemical properties of 2′-O methyl group could modulate RNA structure, thermal stability, biochemical interactions, and other aspects of the modified RNA [33]. Here we report that flavivirus NS5 performs methylation at the 2′-OH position of internal adenosine (Am) of RNA. The 2′-O methylation occurs specifically at internal adenosine, not at guanosine, cytidine, or uridine. Mutagenesis analysis indicates that K61-D146-K181-E218 tetrad of the DENV-4 MTase forms the active site to catalyze internal methylation. Functional studies, using flavivirus luciferase replicon and genome-length RNAs, indicate that internal Am modification reduces viral RNA translation and RNA synthesis. Furthermore, we found that recombinant flavivirus NS5 can methylate host ribosomal RNAs in vitro.
We developed a scintillation proximity assay (SPA) detect methylation of RNA without a 5′ cap structure (Figure 1A). A pppA-RNA (with 5′ triphosphate) representing the first 211 nt of DENV genome sequence was in vitro transcribed in the presence of biotinylated CTP. The biotinylated pppA-RNA was incubated with DENV-4 MTase in the presence of [methyl-3H]-SAM. The methylation reaction was then incubated with SPA beads coated with streptavidin. If the RNA is methylated, binding of the biotinylated RNA to the streptavidin SPA beads brings the [methyl-3H]-labeling into close proximity to the scintillant (embedded in the beads), leading to a signal that can be measured by a scintillation counter. As shown in Figure 1B, the pppA-RNA gained [methyl-3H]-signal upon treatment of DENV-4 MTase. In contrast, no 3H-activity was detected after the pppA-RNA was treated with DENV-4 RdRp domain. Addition of the RdRp domain to the MTase domain did not improve the methylation activity, whereas the full-length (FL) NS5 showed higher activity than the MTase domain alone. Interestingly, similar amounts of 3H-activity were detected after the WNV pppA-RNA and DENV-4 pppA-RNA were treated with DENV-4 FL NS5 and WNV FL NS5, respectively (data not shown). These results demonstrate that (i) DENV-4 MTase can methylate viral RNA without a 5′ cap structure; (ii) the RdRp domain could enhance the MTase activity, but only when the two domains are physically connected; and (iii) WNV and DENV-4 NS5 can cross methylate heterologous viral RNA. To exclude the possibility that the observed methylation occurs at the first nucleotide A (where 2′-O methylation occurs after the pppA-RNA was capped with a 5′ G [i. e. , GpppA-RNA]), we prepared a pppGGA-RNA that contained two extra G residues (underlined) to the 5′ end of authentic viral sequence. Methylation reactions showed that, compared with the pppA-RNA, the addition of two G residues did not change the methylation signals (data not shown). These results suggest that (i) the observed methylation activity is not dependent on the position of the first A residue; (ii) the 3H-signals could be derived from internal methylation of the RNAs without 5′ cap. We expanded the above observation to WNV, another member of flavivirus. Recombinant proteins of WNV FL NS5, MTase domain, RdRp domain were prepared. SPA analysis using pppA-RNA representing the first 190 nt of the WNV genome sequence showed that both FL NS5 and MTase domain, but not RdRp domain, could methylate the pppA-RNA (Figure 1B). As a negative control, vaccinia virus VP39, a known 2′-O MTase of RNA cap, did not methylate pppA-RNA containing DENV-4 or WNV sequence (Figure 1B). In contrast, VP39 efficiently methylated m7GpppA-RNA (to m7GpppAm-RNA) and GpppA-RNA (to GpppAm-RNA); the methylation signals derived from m7GpppA-RNA were higher than those derived from GpppA-RNA (Figure 1C), confirming that VP39 prefers methylating RNA cap with the N7 position of guanine pre-methylated [34]. As expected, both DENV-4 and WNV MTases could methylate GpppA-RNA and m7GpppA-RNA; the signals derived from the former substrate were greater than those derived from the latter substrate (Figure 1C). This is because flavivirus MTase could methylate two positions on substrate GpppA-RNA (to m7GpppAm-RNA), whereas it can only methylate one position on substrate m7GpppA-RNA (to m7GpppAm-RNA). Comparison of methylation signals showed that (i) the VP39-mediated 2′-O methylation is more robust than the flavivirus MTase-mediated cap methylations; and (ii) flavivirus MTase methylates RNA cap more efficiently than internal nucleoside. Taken together, these results indicate that flavivirus NS5 can methylate RNA without a cap structure, possibly through methylating internal nucleoside (s). Using substrate pppA-RNA (5′ 211 nt of DENV genomic RNA) and DENV-4 MTase, we determined the optimal condition for internal methylation activity. As shown in Figure 2, the activity reached maximum when performed at 22°C to 30°C in pH 9. 0 buffer containing 50 mM NaCl and 5 mM MgCl2. Addition of MnCl2 inhibited the internal methylation activity. To identify which nucleoside is internally methylated by DENV-4 MTase, we performed methylation reactions using homopolymer RNAs (polyA, polyG, polyC, or polyU). Since the homopolymer RNAs were not biotinylated, the methylation reactions were purified through an RNeasy column (Qiagen) to remove the un-incorporated [methyl-3H]-SAM. The purified RNAs were then measured for the level of 3H-methyl incorporation using a scintilation counter. The results showed that DENV-4 MTase efficiently methylated polyA (Figure 3A). No 3H-methyl incorporation was detected with polyG, and the incorporations with polyC and polyU were approximately 30-fold less efficient than that of polyA (Figure 3A). The results indicate that (i) DENV-4 MTase preferentially methylates adenosine; (ii) the internal methylation activity does not require a viral RNA sequence. To explore which position of adenosine is methylated, we synthesized three (A) 12 RNA derivatives, each of which was 3′ terminally biotinylated. Oligo (A) 12 contained unmodified adenosines; oligo (Am) 12 contained adenosine with ribose 2′-OH position methylated; and oligo (m6, m6A) 12 contained adenosine with adenine N6 position double methylated. SPA-based methylation assays showed that oligo (A) 12 was an active substrate for DENV-4 MTase (Figure 3B). In contrast, no methylation activity was observed for oligo (Am) 12, while oligo (m6, m6A) 12 had a 53% reduction of the methylation activity than that of oligo (A) 12 (Figure 3B). These results argue that the methylation occurs at the ribose 2′-OH position of adenosine. The reduction of methylation activity of oligo (m6, m6A) 12 could be due to steric hindrance between the double N6 methyl groups of (m6, m6A) 12 and MTase during the methylation reaction. Next, we introduced 2′-O-methyladenine or N6 methyl adenine (m6A) into DENV-4 pppA-RNA (representing the 5′ 211 nt of DENV-4 genome). The pppA-RNA was in vitro transcribed using 2′-O-methyladenine triphosphate (AmTP) or N6 methyl adenine triphosphate (m6ATP) in the absence of unmodified ATP. SPA-based methylation assays showed that the unmodified pppA-RNA and the (m6A) -modified pppA-RNA yielded similar levels of methylation signals (Figure 3C). In contrast, only background methylation signal was observed when using the (Am) -modified pppA-RNA. The results again indicate that DENV-4 MTase methylates the ribose 2′-OH position of adenine. Rigorous chemical identification of Am was achieved by mass spectrometry. High mass-accuracy LC-QTOF analysis of the hydrolysate of DENV-4 MTase-treated polyA revealed only the canonical ribonucleosides (data not shown) and a signal with m/z 282. 1187, as shown in the extracted ion chromatogram in Figure 4A. This m/z value yields a molecular formula of C11O4N5H15, which corresponds to a methylated adenosine species. Subsequent analysis by collision-induced dissociation (CID) revealed fragmentation of m/z 282. 1187 to an ion with m/z 136. 0620 (Figure 4C), which corresponds to adenine base, with loss of a 2′-O-methyl ribose moiety. To confirm that the unknown species was 2′-O-methyladenosine (Am), the LC-QTOF analysis with CID was repeated with synthetic Am, which yielded the same HPLC retention time (Figure 4A), m/z value, and CID fragmentation pattern as the unknown compound (Figure 4B). Analysis of the RNA for other methylated adenosine (e. g. , m1A, m6A, m62A, t6A, i6A) by direct analysis or comparison to chemical standards yielded no detectable signals. These data demonstrate that the methylation catalyzed by the DENV-4 in polyA is specific for the 2′-OH position of adenosine. We examined whether flavivirus MTase has sequence preference for internal adenosine methylation. A set of 3′ truncated DENV-1 RNAs were in vitro transcribed; each RNA contained a 5′ pppAG sequence without a cap structure (Figure 5A). Equal amounts of FL and truncated viral RNAs (0. 5 µg) were treated with DENV-4 MTase in the presence of [3H-methyl]-SAM. As shown in Figure 5B, no significant difference in methylation signals was observed between the FL and truncated RNAs, indicating that the MTase does not have sequence preference within viral genome for internal adenosine methylation. This conclusion was further supported by the results that (i) recombinant DENV MTase and WNV MTase could internally methylate WNV RNA and DENV RNA, respectively, at a similar efficiency (data not shown); (ii) cellular ribosomal 18 S and 28 S RNAs were equally methylated by the DENV-4 MTase (Figure 5B). Quantitative LC-MS/MS analysis revealed the MTase-induced increases in Am levels in 18 S and 28 S rRNA of 67. 9% and 16. 4%, respectively. We performed a structure-based mutagenesis analysis of the DENV-4 MTase to identify amino acids that are critical for internal methylation. Crystal structures of flavivirus MTases share three conserved structural elements (Figure 6A): a SAH-binding pocket, a GTP-binding pocket, and a RNA-binding site [14], [35]. For every structural pocket, we prepared a panel of mutant DENV-4 MTases, each containing an Ala substitution of one amino acid (Figure 6B). In addition, Ala substitution was also performed on the K-D-K-E motif, the active site for the 2′-O cap methylation [15]. All mutant MTases were analyzed using a DENV pppA-RNA (representing the first 211 nt of genomic RNA) in a SPA-based methylation assay. Figure 6C summarizes the internal methylation activities of 18 mutant MTases of DENV-4. (i) For the K61-D146-K181-E217 motif (Figure 6B, residues in yellow), Ala substitution of each residue within the tetrad abolished the methylation activity (Figure 6C), suggesting that the K-D-K-E motif forms the active site for internal methylation. (ii) For the SAM-binding pocket (Figure 6B, residues in blue), mutations of K105, D131, and I147 reduced the methylation activity to <20% of the WT activity, whereas mutation of H110 reduced the methylation activity by about 47% (Figure 6C). These results indicate that the SAM pocket is critical for internal methylation by positioning the methyl donor SAM. (iii) For RNA-binding site, each of the five mutations within the RNA-binding site (Figure 6B, residues in red) reduced the activity by >60% (Figure 6C), indicating the importance of these residues in internal methylation activity. (iv) For the GTP-binding pocket (Figure 6B, residues in green), only one (K14) of the five mutations reduced the activity by >60%. Interestingly, S150A mutant increased the activity by 70% (Figure 6C). It is currently not known how the residues within the RNA-binding site and GTP-binding pocket contribute to the methylation activity. Nevertheless, the mutagenesis results indicate that distinct amino acids of the DENV-4 MTase are required for internal methylation activity. DNEV-1 and WNV replicons expressing Renilla luciferase (RlucRep) were used to analyze the role of internal methylation in viral translation and RNA replication. Transfection of BHK-21 cells with flavivirus RlucRep RNA was previously shown to yield two distinctive peaks, one at 2 to 6 h and another at ≥24 h post-transfection (p. t.). The two luciferase peaks represent viral translation of input RNA and RNA translation of newly synthesized RNA, respectively [36]. As shown in Figure 7A (top panel), DENV-1 and WNV replicon RNAs (containing the 5′ m7GppAm cap) were treated with SAM and cognate MTase, resulting in internally methylated RNAs. As a control, the replicon RNAs were treated with cognate E217A MTase (a mutant that is inactive in internal methylation). Equal amounts of the treated replicon RNAs were electroporated into BHK-21 cells. The transfected cells were assayed for luciferase activities at various time points after electroporation. For both DENV-1 and WNV replicon, the WT MTase-treated replicon generated 10–22% less luciferase activity than the mutant MTase-treated replicon at 2 to 6 h p. t. (Figure 7A), suggesting that internal methylation slightly reduces the translation of viral RNA. At 24 and 48 h p. t. , the luciferase signals derived from the WT MTase-treated replicons were about 26–42% of the luciferase signals derived from the mutant MTase-treated replicons, suggesting that internal methylation suppresses viral RNA synthesis. To exclude the possibility that the observed difference in luciferase activity was caused by a difference in transfection efficiency between the replicons with and without internal methylation, we used RT-PCR to quantify the intracellular levels of viral RNA at various time points post-transfection. During the first 13 h p. t. , similar levels of viral RNAs were detected between the cells transfected with the WT and mutant MTase-treated RNAs (Figure 7B). This result indicates that (i) the RNA transfection efficiencies were comparable; (ii) internal methylation does not change the stability of the transfected RNA. From 18 to 24 h p. t. , the RT-PCR products derived from the WT MTase-treated replicon was much less than those derived from the mutant MTase-treated replicon; this difference became less dramatic from 31 to 48 h p. t. (Figure 7B). The replication difference observed at 18 to 24 h p. t. was most likely due to the difference in internal methylation of the input replicon RNAs. Since both WT and mutant MTase-treated replicons contained WT NS5 gene, the intracelluarly translated WT NS5 protein would methylate progeny viral RNA (and possibly also the transfected replicon RNAs), resulting in less difference in RNA replication observed at 31 to 48 h p. t. (Figure 7B). Next, we analyzed the effect of internal methylation of genome-length RNA on virus production. Genome-length RNA of DENV-1 was in vitro transcribed from an infectious cDNA clone, 5′ capped with m7GpppAm (using vaccinia virus capping enzymes and VP39 MTase), and treated with DENV WT or mutant E217A MTase. Transfection of BHK-21 cells with equal amounts of MTase-treated genome-length RNAs showed similar values of specific infectivity, 4. 3×104 and 4. 8×104 PFU per µg of transfected RNA, respectively (Figure 7C). However, cells transfected with the WT MTase-treated RNA produced slightly less virus than the cells transfected with the mutant MTase-treated RNA (Figure 7C), suggesting that internal methylation could attenuate virus production. We performed genome-length sequencing of the viruses recovered on day 7 post-infection. No adaptive mutation was detected. The results suggest that the decrease in difference of viral titers at later time points between the WT and mutant MTase-treated RNAs is due to a simple dilution of the un-methylated transfected RNA with methylated progeny RNAs produced in subsequent rounds of replication and re-infection. A SPA-based RNA elongation assay was established to compare the viral RdRp activities between RNA templates with and without 2′-O-methyladenosine. As shown in Figure 7D (left panel), a 5′ terminally biotinylated RNA oligo was annealed to a template (A) 20 or (Am) 20. Incorporation of 3H-labelled UTP into the biotinylated RNA in the presence of recombinant DENV-4 NS5 was measured. The amount of 3H-UMP incorporation into the (Am) 20 template was about 16-fold less than that into the (A) 20 template (Figure 7D, right panel). A similar result was observed when RNA elongation activity was monitored by incorporation of 32P-UMP and the RNA products were analyzed on a denaturing polyacrylamide gel (data not shown). Collectively, the results indicate that 2′-O-methyladenosine in RNA template reduces the efficiency of RNA elongation. To demonstrate the biological relevance of internal adenosine methylation, we purified WT and MTase E217A mutant DENV-1 virions. Viral genomic RNAs were extracted from the virions and enzymatically digested to ribonucleoside form, followed by LC-MS analysis of the ribonucleosides (see Materials and Methods for details). The analyses revealed that genomic RNA extracted from the WT virion contained Am at a frequency of 3. 4±0. 05 Am per genome (Figure 4D), while genomic RNA purified from the MTase mutant virion did not contain significant levels of Am (Figure 4E). The results clearly demonstrate that internal 2′-O-methyladenosine exists in DENV genomic RNA, though at a low frequency.
The current study has provided four lines of evidence to demonstrate that flavivirus MTase performs internal 2′-O methylation of adenine (Am). (i) Recombinant NS5 of DENV-4 and WNV can methylate viral RNA without a 5′ cap structure. Recombinant NS5 of other DENV serotypes (DENV-2 and -3) can also perform internal methylation (data not shown). (ii) DENV-4 MTase methylates polyA, but not polyG, polyC, or polyU. This is in stark contrast to flavivirus N-7 and 2′-O cap methylations which require RNA substrates with distinct viral sequence and structural elements [19], [37]. This is also different from the requirement of cellular mRNA m6A methylation, which occurs only within the GAC or AAC sequences (where A is methylated) [28], [38]. (iii) RNAs containing Am are not active substrates for internal methylation, whereas RNAs containing m6A are active substrates for internal methylation. (iv) Mass spectrometric analysis showed that the methylated product was 2′-O-methyladenosine. Importantly, we showed that genomic RNA extracted from DENV virion contains internal 2′-O-methyladenine albeit at low frequency. It should be noted that the internal adenosine methylation activity of flavivirus MTase is much lower than the N-7 and 2′-O cap methylations (compare Figures 1B and 1C). The observed internal methylation activity seems unique to flavivirus MTase since vaccinia virus VP39, a well known 2′-O MTase of RNA cap, did not show any internal methylation activity. Flavivirus internal Am modification exhibits a number of properties similar to that of 2′-O cap methylation. (i) Both methylations occur at the ribose 2′-OH position of adenosine (i. e. , m7GpppApN→m7GpppAmpN and NpApN→NpAmpN). For 2′-O cap methylation, we previously showed that substitution of the wild-type m7GpppA with m7GpppG completely abolished the 2′-O cap methylation of WNV RNA [19]. For 2′-O internal methylation, DENV MTase does not seem to have preference for RNA sequence context within viral genome; it can even methylate host ribosomal RNAs at an equal efficiency (Figure 5). (ii) Both methylations transfer a methyl group from SAM molecule that is bound to the same pocket of the enzyme. This is supported by two evidences: only one SAM-binding site is observed in flavivirus MTase crystal structure; mutations of the SAM-binding pocket abolished both cap methylations [18], [37] as well as internal adenosine methylation (Figure 4C). (iii) Both activities use the K61-D146-K181-E217 tetrad as an active site. Ala-substitution of each of the tetrad lead to complete loss of 2′-O cap methylation [18], [37] and internal adenosine methylation (Figure 4C). (iv) Both reactions require a similar optimal buffer conditions (e. g. , optimal pH at 9. 0) [35], [37]. These similarities suggest that the two reactions share a common mechanism of catalysis. We recently solved the co-crystal structure of DENV-3 MTase in complex with SAH and an m7GpppA-RNA oligo (Lescar et al. , submitted for publication). The co-crystal structure supports a mechanism that, during 2′-O cap methylation or internal adenosine methylation, K181 is deprotonated, leading to the deprotonation of the 2′-OH of ribose. The deprotonated 2′-OH of ribose then interacts with the CH3 group from SAM molecule, resulting in the formation of the SN-2-like transition state to accomplish the methyl transfer. A similar mechanism has previously proposed for vaccinia VP39 and other MTases [39], [40]. We explored the function of internal Am by analyzing its effects on viral RNA translation and replication. Using luciferase replicons of DENV-1 and WNV treated with respective recombinant MTases, we found that internal Am reduced the efficiency of RNA translation by approximately 10–22% (Figure 7A). This is in contrast to the observation that m6A modification enhances mRNA translation [31]. It is currently not clear how the two distinct methylations of adenosine modulate translation with opposite outcomes. For RNA replication, we found that internal Am significantly reduced viral RNA synthesis of DENV-1 replicon (indicated by luciferase reporting signals; Figures 7B and 7C) as well as the replication of genome-length RNA (Figure 7C). The lower efficiency of RNA synthesis could result from a decrease in input RNA translation. Alternatively, the internal Am could directly attenuate RNA replication during the first round of viral replication. The latter explanation was supported by the biochemical results showing that 2′-O-methyladenosine in RNA template reduces the efficiency of NS5-mediated RNA elongation (Figure 7D). In addition, since the 2′-O methylation of viral RNA cap functions in subverting innate host antiviral response [20], [41], it is possible that internal methylation of viral RNA could also modulate virus-host interactions. It should be noted that because the same K-D-K-E active site of MTase is responsible for 2′-O methylations of both 5′ RNA cap and internal adenosine, the observed evasion of immune response could be due to lack of methylation (s) of RNA cap and/or internal adenosine. Indeed, we found that 2′-O MTase mutant virus triggered stronger innate immune response than the WT virus did in cell culture (manuscript in preparation). It is currently impossible to differentiate the effect of the two types of methylations (5′ RNA cap and internal adenosine) on evasion of host immune response. In eukaryotes, 2′-O methylation is abundant in splicesomal snRNA and ribosomal RNA; however, 2′-O methylation has not been reported for mRNA. In spliceosomal snRNA, 2′-O methylation occurs at the branch point adenosine; such modification was shown to block pre-mRNA splicing in Xenopus oocytes [42], [43]. In ribosomal RNA, 2′-O methylation could increase the stability of RNA conformation [32]. We showed that flavivirus MTase can methylate cellular ribosomal RNAs in vitro (Figure 5B). This observation raises the possibility that the viral MTase may modulate host RNAs in infected cells. During flavivirus infection, only a small portion of expressed NS5 protein is located within the replication complex; majority of the viral NS5 protein is dispersed outside the replication site [44]. DENV NS5 translocates into nucleus in infected cells, and the distribution of DENV NS5 between cytoplasm and nucleus is regulated by the phosphorylation status of the NS5 protein [44]. Future studies are needed to examine whether host RNAs are modified by flavivirus MTase in infected cells.
All chemicals and reagents were of the highest purity available and were used without further purification. Nuclease P1 and phosphodiesterase I were purchased from USB (Cleveland, OH, USA). Coformycin were obtained from the National Cancer Institute Open Chemical Repository (Bethesda, MD USA). Deferoxamine mesylate, tetrahydrouridine, butylated hydroxytoluene, alkaline phosphatase, and 2′-O-methyladenosine were purchased from Sigma Chemical Co. (St. Louis, MO, USA). Thermo Hypersil aQ HPLC column was purchased from Thermo Fisher Scientific (Waltham, MA, USA). Experiments were performed on an Agilent LC/QTOF 6520 system (Santa Clara, CA, USA). Two types of RNA substrates were prepared for methylation analysis. Type one RNA contains non–viral sequence. Synthetic RNA oligos include (A) 12, (Am) 12 (“m” indicates 2′-O-methyl adenosine), and (m62A) 12 (dimethylation of the exocyclic N6-position of adenosine). The 3′ end of each oligo is biotinylated. These oligos were synthesized by Dhamacon, Inc (Lafayette, CO). In addition, polyA, polyC, polyG, and polyU without any biotinylation were also used in the methylation assay. These homopolymers were purchased from Sigma Aldrich. For the 18S and 28S rRNA species, total RNA was isolated from CCRF-SB B-lymphocytic leukemia cells (ATCC, Manassas, VA, USA) by homogenizing cells (repetitive pipetting) in 1 ml of Trizol reagent followed by extraction with 0. 2 ml volume of chloroform and isopropanol precipitation of the aqueous phase. The 18 S and 28 S rRNA species were purified by size-exclusion HPLC using an Agilent 1200 HPLC system (Agilent Technologies, Santa Clara, CA, USA) equipped with an Agilent Bio-SEC5 column (2000 Å, 300 mm×7. 8 mm) eluted at 60°C with 100 mM ammonium acetate at 0. 5 ml/min. Collected fractions were desalted using Ambion Millipore 10K MWCO columns (Millipore, Billerica, MA, USA) and the quality and concentration of the resulting rRNAs was assessed by analysis on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA; RNA 6000 Nano Kit). Type two RNA contains viral sequence. RNAs representing the first 190 nucleotides (nt) of WNV genome or the first 211 nt of DENV-1 genome were in vitro transcribed from PCR-generated DNA templates as reported previously [15], [37]. The in vitro transcription was performed using MEGAshortscription Kit (Applied Biosystems). Biotinylated RNAs were produced using biotinylated-CTP and regular CTP at a ratio of 1∶2. RNAs containing 2′-O-methyladenosines or N6-methyladenosines (m6A) were in vitro transcribed using 2′-O-methyladenosine triphosphate (AmTP) or N6-methyladenosine triphosphate (m6ATP) in the absence of unmodified ATP. RNAs with 5′ m7GpppA or GpppA cap were prepared by incubation of in vitro transcribed pppA-RNA with vaccinia virus capping enzyme (Epicetre) in the presence of GTP with or without SAM, respectively. RNA with 5′ m7GpppAm cap was prepared by vaccinia capping enzyme and VP39 2′-O MTase following the manufacturer' s protocol (Epicetre). All RNA transcripts were purified twice through Sephadex G-25 spin columns (GE Healthcare), extracted with phenol-chloroform, precipitated with ethanol, and resuspended in RNase-free water. Three assays were performed to detect internal RNA methylation. The first assay used SPA format in a 96-well plate (Figure 1A). Biotinylated RNA species (6 pmol) were incubated with 18 pmol of full-length NS5 (or MTase domain) and 1 µCi of [3H-methyl]-SAM (PerkinElmer) in buffer containing 50 mM Tris-HCl (pH 9. 0) and 50 mM NaCl at room temperature for 1 h. The reactions were terminated by addition of an equal volume of 2× stop solution (containing 100 mg of SPA beads coated with streptavidin in 50 mM Tris-HCl [pH 7. 0], 40 mM EDTA, and 150 mM NaCl). The 96-well plate was agitated at room temperature for 15 min, and measured for 3H-methyl incorporation (into RNA) by a MicroBeta counter (Perkin-Elmer). The full-length NS5 and MTase domain from both WNV and DENV-4 were used in the methylation assays. The MTase domains of WNV and DENV-4 contained the first 300 and 272 amino acids of their respective NS5 proteins. The preparations of NS5 and MTase proteins were reported previously [15], [37]. The second assay measured [3H-methyl] incorporation into non-biotinylated RNA substrates. The reaction (20 µl) contained 50 mM Tris-HCl (pH 9. 0), 2 µg MTase of DENV-4,1 µCi of [3H-methyl]-SAM, and 1 µg of oligo RNA, viral RNA, 18 S rRNA, or 28 S rRNA. After incubation at room temperature for 1 h, the unincorporated [3H]-SAM was removed by RNeasy kit (Qiagen, Valencia, CA USA) according to the manufacture' s instruction. The RNA samples were then mixed with 50 µl of optiphase supermix (Perkin Elmer), and measured for [3H-methyl] incorporation by a MicroBeta counter. A third assay involved LC-MS analysis of Am following treatment of polyA, 18S rRNA and 28S rRNA (1 µg) with DENV-4 MTase (2 µg) in a reaction (total volume 20 µl) containing 50 mM Tris-HCl (pH 9. 0), 2 mM DTT and 50 µM SAM, with incubation at room temperature for 1. 5 h. In addition, genomic RNA purified from DENV virion was directly analyzed using LC-MS (see below). For analysis of internal adenosine methylation of genomic RNA, WT and MTase E217A mutant DENV-1 virions, grown in mosquito C6/36 cells, were purified. Briefly, C6/36 were infected with DENV-1 at an MOI (multiplicity of infection) of 0. 1 and incubated at 29°C for five days. Cell culture supernatants were then harvested and virus were precipitated using 8% PEG8000 (w/v) at 4°C overnight. Precipitated virus was then resuspended in NTE buffer (12 mM Tris-HCl, 120 mM NaCl, 1 mM EDTA, pH 8. 0) and purified by spinning the virus through a 24% (w/v) sucrose cushion at 75,350× g for 1. 5 h at 4°C. Virus pellet was resuspended into 4% (w/v) potassium tartrate in NTE buffer and centrifuged at 149,008× g for 2 h at 4°C. Virus was further purified in a 10–30% potassium tartrate gradient by spinning at 126,444× g for 2 h at 4°C. Virus band of WT or E217A mutant was collected and concentrated using Millipore Amicon Ultra 100 K MWCO (Molecular Weight Cutoff). Virus samples were analyzed on a 15% SDS-PAGE stained with Coomassie brilliant blue to visualize viral capsid, membrane (pre-membrane), and envelope proteins. The amount of purified WT and E217A mutant virus were similar. Genomic RNAs were extracted from the purified virions using trizol (Invitrogen), quantified using NanoDrop, and subjected to enzymatic hydrolysis as described below. Identification and quantification of Am in samples of MTase-treated polyA, and 18S and 28S rRNA, and in DENV genomic RNA was achieved by analysis of RNA-derived ribonucleosides by HPLC-coupled mass spectrometry. For all analyses, samples of RNA (1–3 µg) were treated with 1 U/µl nuclease P1,2. 5 mM deferoxamine mesylate (antioxidant), 10 ng/ml coformycin (adenosine deaminase inhibitor), 50 µg/ml tetrahydrouridine (cytidine deaminase inhibitor), and 0. 5 mM butylated hydroxytoluene (antioxidant) at 37°C. After 3 h, alkaline phosphatase and phosphodiesterase I were added to a final concentration of 0. 1 U/ml. The sample was incubated at 37°C overnight, followed by removal of enzymes by filtration through a 10,000 kDa-molecular weight cut-off Amicon spin filter. The resulting filtrate was lyophilized prior to mass spectrometric analysis. For identification of Am in polyA treated with MTase, the lyophilized hydrolysis products were dissolved in deionized water and analyzed by HPLC-coupled, electrospray ionization (ESI) quadrupole time-of-flight mass spectrometry (LC-QTOF). To resolve ribonucleosides, the digested sample (5 µl) was loaded onto a Thermo Hypersil aQ column (100×2. 1 mm, 1. 9 µm particle size) at 25°C and eluted at a flow rate of 50 µl/min with an acetonitrile gradient using the following mobile phases: Solvent A: 0. 1% formic acid in 10 mM ammonium acetate (pH 7. 5); and Solvent B: 0. 1% formic acid in acetonitrile. The percentage of Solvent B was as follows: 0–5 min, 0. 5%; 5–14 min, 5%; 14–19 min, 5%; 19–31 min, 23%; 31–34 min, 23%; 34–49 min, 83%; 49–54 min, 83%; 54–60 min, 0. 5%; 60–70 min, 0. 5%. The eluent was analyzed on an Agilent QTOF 6520 mass spectrometer with an ESI source operated in positive ion mode and the mass spectrometer was operated in ion scanning mode (m/z 100–1000) with the following parameters: gas temperature, 350°C; drying gas, 10 l/min; fragmentor voltage, 100 V; skimmer voltage, 65 V; and capillary voltage, 3500 V. Data processing was performed using MassHunter Workstation software version B. 03. 01 (Agilent Technologies, Santa Clara CA USA). The retention times and exact molecular weights of Am and other methylated ribonucleoside species were compared to chemical standards (Cm, Um, Gm, m3C/m5C, m62A, m1A/m6A, and m7G). To quantify Am, samples of RNA (MTase-treated 18S and 28S rRNA, and DENV genomic RNAs) were hydrolyzed to ribonucleosides, as described above, followed by resolution of the ribonucleosides on a Thermo Hypersil aQ column (100×2. 1 mm, 1. 9 µm particle size) at 25°C at a flow rate of 0. 3 ml/min with an acetonitrile gradient in 0. 1% (v/v) formic acid in water as follows: 0–1. 5 min, 0%; 1. 5–2. 7 min, 6%; 2. 7–4. 4 min, 17. 8%; 4. 4–5. 0 min, 18%; 5. 0–5. 5 min, 18%; 5. 5–8. 1 min, 30%; 8. 1–9. 3 min, 40%; 9. 3–10. 2 min, 65%; 10. 2–11. 1 min, 95%; 11. 1–12. 0 min, 95%; 12. 0–12. 9 min, 0%; 12. 9–15. 0 min, 0%. The HPLC column was coupled to a triple quadrupole mass spectrometer (LC-MS/MS) operated in positive ion, multiple reaction monitoring (MRM) mode for the Am molecular transition of m/z 282→136. Voltages and source gas parameters were as follows: gas temperature, 300°C; sheath gas temperature, 325°C; gas flow, 8 l/min; nebulizer, 40 psi; and capillary voltage, 4000 V. Quantification of Am and adenosine in the MTase-treated and DENV viral RNA samples was achieved by integrating the extracted ion chromatographic peaks for molecular transitions m/z 282→136 and m/z 268→136, respectively, followed by interpolation from linear external calibration curves prepared with Am (0. 5–10 nM) or adenosine (0. 1–10 µM) dissolved in the hydrolysis buffer as a matrix control. The number of Am per viral genome was calculated by multiplying the measured value of Am per adenosine by the number adenosines in the 10,735 ribonucleotide DENV-1 genome (2861; NCBI Genome Database; http: //www. ncbi. nlm. nih. gov/genome? term=dengue%20virus%201). Renilla luciferase (Rluc) -reporter replicons of DENV-1 (Western Pacific 74 strain; GenBank accession U88535) [45] and WNV (New York strain 3356) [36] were used to examine the effect of internal methylation on viral translation and RNA synthesis. Replicon RNAs were in vitro transcribed using mMESSAGE mMACHINE kits (Applied Biosystems). The 5′ end of replicon or genome-length RNA was treated with vaccinia virus capping enzyme and VP39 to generate m7GpppAm-RNA following the manufacturer' s protocols (Epicetre). The reactions were extracted with phenol∶chloroform and precipitated using ethanol. The resulting replicon RNAs (4 µg) were treated with 2 µg of WT and mutant WNV or DENV-1 MTases in the presence of 50 µM SAM in the methylation buffer described above. After incubation at room temperature for 1 h, the reaction mixtures were directly electroporated to 8×106 BHK-21 cells [46]. The electroporated cells were resuspended in 20 ml of DMEM medium with 10% FCS. 0. 5∼1. 0 ml of cells were seeded onto a 12-well plate (2∼4×105 cells per well), and assayed for luciferase activities at indicated time points. The luciferase assay was performed as reported previously [47]. Besides measuring luciferase activity, we also quantified intracellular viral RNA at various time points after electroporation. For each time point, total cellular RNA was extracted using RNeasy kit (Qiagen). The extracted RNAs (3 µg) were subjected to standard RT-PCR quantification using one primer pair targeting viral NS5 gene (forward primer 5′-TGAAGCTAAGGTGCTTGAGC-3′ and reverse primer 5′-AGCCACATCTGGGCATAAGA-3′) and another primer pair targeting housekeeping gene actin (forward primer 5′-AGAGGGAAATTGTGCGTGAC-3′ and reverse primer 5′-CAATGGTGA TGACCTGGCCA-3′) The RT-PCR reactions were performed using SuperScript III one-step RT-PCR kit (Invitrogen), and the products were analyzed on a 1% agarose gel. Genome-length RNA of DENV-1was in vitro transcribed from a full-length cDNA plasmid linearized by SacII [48]. Using the same protocol for replicon experiment (described above), the genome-length RNA containing a 5′ m7GpppAm cap (2 µg) was treated with WT and mutant DENV-1 MTases. The internally methylated genome-length RNAs were electroporated into BHK-21 cells [46]. The transfected cells were resuspended in 20 ml of DMEM medium, and subjected to virus production and specific infectivity assays. For virus production assay, 18 ml of the resuspended cells plus 10 ml of medium were cultured in a T-175 flask. Viral titers of culture fluids collected on day 1–7 post-transfection (p. t.) were determined using a single-layer plaque assay [37]. For specific infectivity assay, a series of 1∶10 dilutions of the transfected cells were prepared using DMEM medium. One ml of cell suspension at each dilution was seeded onto confluent BHK-21 cell in six-well plates (The plates were seeded with 5×105 BHK-21 at 16–24 h before the assay day). After incubating the plates for 6 h (to allow the transfected cell to attach to the monolayer of BHK-21 cells), culture medium was aspirated and replaced with an overlayer medium (RPMI 1640,2% FBS, 1% penicillin-streptomycin, and 0. 8% methylcellulose). The plates were incubated at 37°C with 5% CO2 for 5 days. The cells were then fixed with 10% formaldehyde for 20 min at room temperature, rinsed with tap water, and stained with 1% crystal violet for 5 min. The plates were again rinsed with tap water (to remove staining) and visually counted for plaques. The specific infectivity was calculated as the number of infectious virus upon transfection of 1 µg of genome-length RNA. The sequences of RNA template and primer are shown in Figure 7D. The two RNAs (12. 5 µM) were annealed in 50 mM Tris-HCl (pH 7. 0) and 100 mM NaCl by heating at 95°C for 3 min followed by cooling to room temperature (23°C). The RNA elongation reaction (25 µl) contained 50 mM Tris-HCl (pH 7. 0), 50 mM NaCl, 5 mM MgCl2,2 mM MnCl2,0. 25 µM annealed RNA template/primer, 1 µM cold UTP, 1 µM 3H-UTP, 4 mM DTT, and 50 nM full-length DENV-4 NS5. After incubating the reaction at room temperature for 1 h, the reactions were terminated by addition of an equal volume of 2× stop solution (containing 100 mg of SPA beads coated with streptavidin in 50 mM This-HCl [pH 7. 0], 40 mM EDTA, and 150 mM NaCl). The 96-well plate was agitated at room temperature for 15 min, and measured for 3H-UTP incorporation as described above. | We report that flavivirus NS5 has a novel internal RNA methylation activity. Recombinant proteins of NS5 and its N-terminal methyltransferase domain of West Nile virus and Dengue virus (DENV) specifically methylates polyA, but not polyG, polyC, or polyU. RNAs with internal adenosines substituted with 2′-O-methyladenosines are not active substrates for internal methylation, suggesting that the internal methylation occurs at the 2′-OH position of adenosine. Mass spectroscopic analysis confirmed that the internal methylation product is 2′-O-methyladenosine. Furthermore, the 2′-O-methyladenosine could also be detected in DENV genomic RNA. The 2′-O methylation of internal adenosine does not require specific RNA sequence context because the DENV methyltransferase can methylate RNAs spanning different regions of viral genome and host ribosomal RNAs at equal efficiencies. Mutagenesis analysis showed that K61-D146-K181-E217 motif of the DENV methyltransferase forms the active site of internal methylation activity; in addition, distinct residues on the surface of the enzyme are critical for the internal methylation activity. Functional analysis showed that internal methylation attenuated viral RNA translation and replication. Overall, our results demonstrate that flavivirus NS5 performs 2′-O methylation of internal adenosine of viral RNA in vivo and host ribosomal RNA in vitro. Such 2′-O-methyladenosine modification may modulate virus-host interaction. | Abstract
Introduction
Results
Discussion
Materials and Methods | biochemistry
biology
microbiology
molecular cell biology | 2012 | 2′-O Methylation of Internal Adenosine by Flavivirus NS5 Methyltransferase | 13,669 | 406 |
Peptidoglycan hydrolases are a double-edged sword. They are required for normal cell division, but when dysregulated can become autolysins lethal to bacteria. How bacteria ensure that peptidoglycan hydrolases function only in the correct spatial and temporal context remains largely unknown. Here, we demonstrate that dysregulation converts the essential mycobacterial peptidoglycan hydrolase RipA to an autolysin that compromises cellular structural integrity. We find that mycobacteria control RipA activity through two interconnected levels of regulation in vivo—protein interactions coordinate PG hydrolysis, while proteolysis is necessary for RipA enzymatic activity. Dysregulation of RipA protein complexes by treatment with a peptidoglycan synthase inhibitor leads to excessive RipA activity and impairment of correct morphology. Furthermore, expression of a RipA dominant negative mutant or of differentially processed RipA homologues reveals that RipA is produced as a zymogen, requiring proteolytic processing for activity. The amount of RipA processing differs between fast-growing and slow-growing mycobacteria and correlates with the requirement for peptidoglycan hydrolase activity in these species. Together, the complex picture of RipA regulation is a part of a growing paradigm for careful control of cell wall hydrolysis by bacteria during growth, and may represent a novel target for chemotherapy development.
Mycobacterium tuberculosis is the causative agent of tuberculosis and accounts for up to 10 million symptomatic infections a year [1]. The spread of multi-, extensively- and now totally- drug resistant strains [2] has created a pressing need to understand essential mycobacterial processes in an effort to define novel targets for chemotherapy. One highly essential bacterial process is peptidoglycan (PG) synthesis and remodeling, which is critical for providing structural integrity in nearly all bacteria. PG forms a continuous macromolecular mesh that is part of the bacterial cell wall and is required for correct cellular morphology and opposition to osmotic forces. Despite extensive biochemical and genetic characterization of the enzymes responsible for the synthesis and degradation of PG (reviewed in [3], [4]), the mechanism by which these enzymes coordinate their activities remains poorly defined. It is clear, however, that dysregulation of this homeostatic balance frequently has lethal effects on the bacterium—inactivation of peptidoglycan synthases, either through the use of penicillin derivatives or overexpression of dominant negative forms of PG synthetic enzymes, induces lysis of cells [5], [6]. In many cases, this lethality can be suppressed by inactivation of several peptidoglycan hydrolases [5], [7], suggesting that PG hydrolase autolysin activity is restrained by functional interactions with PG synthases. This idea is consistent with a ‘make-then-break’ approach to cell wall synthesis where new PG subunits are first incorporated before the existing sacculus is cleaved to allow expansion [8]. One example of this is the formation of the septal PG—cells ensure that the septal PG is formed before PG hydrolases cleave apart the daughter cells. Recent work suggests that the activity of PG synthetic and hydrolytic enzymes is regulated by the formation of protein complexes. In E. coli, the PG amidases AmiA, AmiB and AmiC can interact with non-enzymatic partners that upregulate septal peptidoglycan hydrolysis [9]. Conversely, the major bifunctional PG synthases, PBP1A and PBP1B in E. coli interact with and rely on essential lipoprotein partners for function [10]. In addition to interactions with non-enzymatic partners, several affinity chromatography and genetic studies have identified interactions between PG modulating enzymes themselves [11]. While the exact interactions may be species-specific, in general, PG synthases can associate with both other PG synthases and with PG hydrolases. Likewise, PG hydrolases can form predicted hydrolytic complexes with other autolysins [11]–[13]. These results suggest a general paradigm where PG modulating enzymes of both similar and opposing functions assemble as multi-protein complexes that spatially and temporally coordinate PG synthesis and hydrolysis during bacterial growth and division. An immediate challenge is to translate the many identified interactions into functional in vivo effects on the growth and division of bacteria. Previously, we have studied regulation of the essential M. tuberculosis PG hydrolase, RipA (Rv1477). RipA belongs to the NLPC/p60 family, and has been characterized as a D, L D-glutamate-diaminopimelic acid (DAP) endopeptidase that cleaves within the pentapeptide bridges of the PG sacculus, thereby removing cell wall crosslinks [14]. The RipA homologue in Listeria (P60) and in Mycobacterium marinum (IipA) can be deleted, but this causes septal resolution defects [15], [16]. In contrast, RipA is essential in M. tuberculosis [17], and depletion of RipA produces a chaining phenotype in M. smegmatis, which causes severe growth inhibition [18]. This is unlike the case in E. coli, where extensive chaining and growth inhibition requires inactivation of several PG hydrolases [19]. In this work, we interrogate the mechanism by which RipA activity is regulated in vivo during vegetative growth. We report that RipA requires careful control to support growth and division without compromising the cell' s structural integrity—RipA becomes a lethal autolysin when its activity is dysregulated. Under physiological conditions, RipA relies on protein interactions to correctly control its degradative capacity. These interactions are also necessary for proteolytic cleavage of RipA to produce active enzyme. RipA cleavage and activation is more robust in M. smegmatis than in the pathogenic M. tuberculosis or M. bovis BCG, which may be a reflection of the different PG hydrolysis requirements between fast and slow growing mycobacteria. However, bypassing RipA cleavage by overexpressing fully active truncated enzyme compromises the structural integrity of both M. smegmatis and M. tuberculosis, suggesting that RipA cleavage may be rate-limited in order to synchronize PG hydrolysis with the growth rate of the bacterium. These results suggest a model in which RipA is regulated by several interconnected post-transcriptional mechanisms—proteolytic processing produces active enzyme, while protein-protein interactions upstream and downstream of cleavage ensure RipA functions correctly at the septum.
When RipA is depleted, daughter cells are unable to separate and instead, grow as chains (Figure 1A). While cells require peptidoglycan hydrolysis to accomplish cell separation, excessive cell wall degradation can compromise structural integrity and lead to lysis. We hypothesized that RipA sits in this precarious situation, where the cell cannot tolerate either too little or too much RipA activity. We investigated whether excessive RipA activity is toxic to mycobacteria by inducing M. smegmatis RipA (RipASm) from a tetracycline-inducible episomal plasmid in M. smegmatis. Unlike the chaining phenotype we have previously observed with RipA depletion, RipA overexpression caused the rod-shaped cells to become spherical and lyse, (Figure 1A, 1D (time-lapse movie in Video S1). This is dependent on catalytic activity, as overexpression of a catalytic mutant, RipASm C408A, does not display this phenotype (Figure 1A, 1D, (time-lapse movie in Video S2). The spherical phenotype of RipASm overexpression led to a severe growth defect by optical density (Figure 1B) and over one hundred fold killing, as determined by CFU enumeration (Figure 1C). Thus, excessive RipA activity in the cell is highly lethal. To determine whether a more physiological level of RipA could be converted to a lethal autolysin, we dysregulated RipA activity through the use of the beta-lactam antibiotic meropenem. Beta-lactam antibiotics block PG precursor incorporation, which causes excessive PG hydrolase activity and cell lysis [7]. While M. tuberculosis is relatively resistant to most beta-lactams, recent work has shown that meropenem is more resistant to the endogenous mycobaterial beta-lactamase, and is highly effective at killing M. tuberculosis, especially in combination with the beta-lactamase inhibitor clavulanate [20]. Meropenem targets PBP2 and PBP3 in E. coli, as well as L, D transpeptidases in M. tuberculosis [21], [22]. Since RipA is known to interact with the PG synthase PBP1, which is required for normal vegetative growth [23] and morphology in mycobacteria [24] (depletion of the protein leads to rounded cells), we asked whether meropenem treatment can dysregulate RipA and convert the enzyme into a lethal autolysin. We first treated M. smegmatis with 10 µg/mL meropenem and assessed morphological changes over time by microscopy. Treated cells filament and swell at the poles and septa, which are the sites of mycobacterial PG incorporation (Figure 2A, arrows). This morphological toxicity correlated with a decrease in optical density over time, which suggested lysis (Figure 2B). This was borne out by CFU analysis, which showed that 80% of treated cells were killed within 6 hours of meropenem treatment (Figure 2C, bar 2). The bulging at sites of PG incorporation after meropenem treatment suggested an excess of PG hydrolase activity. Since RipA localizes both to poles and septa, we hypothesized that RipA may play a role in killing meropenem treated cells. To test this idea, we depleted M. smegmatis of RipA before meropenem treatment (Figure S1A) and then assessed survival with meropenem treatment by CFU enumeration. We found that unlike RipA replete cells (Figure 2C, bar 2), meropenem did not kill RipA depleted cells (Figure 2C, bar 4). Furthermore, while RipA replete cells bulged under meropenem treatment as expected (Figure 2D, arrows), the RipA depleted cells appeared refractory to swelling (Figure 2D). As a control, cells were treated with SDS (which causes non-specific cell wall and membrane stress), as well as streptomycin, which targets protein synthesis. These control cells showed no survival (Figure S1B) or morphological (Figure S2) differences between RipA replete and depleted cells, demonstrating RipA specifically interacts with meropenem-affected pathways. Given that RipA enzymatic activity is modulated through protein-protein interactions with different PG synthetic and hydrolytic partners [18], [24], the meropenem data suggest that RipA is held in check in complexes by an interacting protein, such as PBP1. Furthermore, although there are many hydrolases that could contribute to cell death when PG synthesis is blocked, we found that at least for the synthases blocked by the clinically relevant beta-lactam antibiotic meropenem, RipA is quantitatively the single most important hydrolase. Our results demonstrate that RipA dysregulation is highly detrimental to the cell. Thus, mycobacteria must control the activity of RipA during growth—there must be enough PG hydrolase activity around to support growth and division, but not an excessive amount so as to compromise structural integrity. One way this control may be regulated is at the transcriptional level. We assessed whether the cell downregulates RipA expression using quantitative PCR. Since RipA is required for septation, which does not occur in non-replicative conditions, we compared RipA expression between exponential and stationary phases. We found that RipA remained expressed from exponential phase through the transition into stationary phase (Figure S3), suggesting there may be post-transcriptional mechanisms responsible for restraining RipA activity when it is not needed. Thus, we investigated whether achieving tight control of RipA activity may rely on post-transcriptional processes. We hypothesized that removal of wildtype RipA from its endogenous niche in vivo would inhibit correct septal resolution and therefore phenocopy RipA depletion. If RipA requires downstream interactions for activity, e. g. with members of septal complexes or with post-translational enzymatic regulatory proteins, then we should be able to create a dominant negative RipA mutant, where the critical catalytic cysteine [25] is mutated to a nonfunctional alanine. Overexpression of the RipASm C408A catalytic mutant should result in competition between nonfunctional RipA and endogenous RipA for required post-translational activation processes. If RipA requires these interactions to function, then we would observe chaining. Indeed, when we induced RipASm C408A, cells grew as short chains (Figure 1A, white arrows), suggesting that RipA interactions are necessary for correct septal PG hydrolysis. While overexpression of the RipASm C408A mutant produced a severe growth defect (Figure 1B, 1C), it was not accompanied by the widespread lysis that was observed upon wildtype RipASm overexpression (Figure 1C). The apparent drop in optical density upon longer induction of the RipASm C408A strain (Figure 1B) was due to clumping of the culture, which though affecting optical density, did not lead to a drop in CFU, indicating growth inhibition rather than lysis (Figure 1C). However, we did observe occasional lysis of the RipASm C408A strain in addition to the dominant negative chaining phenotype. Some cells within a chain produced a slight bulging phenotype, which is indicative of an increase in PG hydrolytic activity (Figure 1A, red arrows). These bulging cells, like in RipA dysregulated cells, can go on to lyse (Figure 1D, Video S2), though this does not lead to detectable cell death by CFU enumeration (Figure 1C). It is possible that displacement of endogenous wildtype RipA from complexes at the septum leads to activity at ectopic sites in a subset of cells. Alternatively, RipA overexpression could stimulate other endogenous PG hydrolases. When we used a RipA polyclonal antibody that recognizes a C-terminal epitope, we observed truncated RipA species from mycobacteria by Western blotting. When we overexpressed RipASm, we found several bands smaller than the predicted full length protein (Figure 3A, lane 3). Likewise, we saw these truncated bands when we overexpressed RipASm C408A in M. smegmatis (Figure 3A, brackets). These products were not due to non-specific cytoplasmic degradation of overexpressed RipA, as we fractionated RipASm C408A overexpressing cells and found that RipA processed species were enriched in the cell wall fraction (Figure S4A). The efficiency of fractionating mycobacteria was confirmed by Western blotting against RpoB (cytosolic) and mycobacterial antigen 85 (cell wall) markers (Figures S4B, S4C). Thus, these results suggest that RipA undergoes physiological post-translational processing in the periplasmic or cell wall compartment. To further demonstrate that RipA processing is physiological and not an artifact of overexpression, we used Western blotting to estimate the size of RipA in wildtype M. smegmatis whole cell lysates. In mid-exponential phase cells, RipA formed a smear of ∼30 kDa (Figure 3A, lane 2, red arrow) with no detectable full length protein present. This signal was specific for RipA, as cells depleted for RipA (Figure 3A, lane 1) had decreased signal compared to wildtype cells (Figure 3A, lane 2) when equal amounts of total protein were analyzed (Figure 3B). Furthermore, processed endogenous RipA partitioned to the cell wall compartment (Figure S4A). The smear of processed RipA suggests there are multiple processing sites. This may represent multiple cleavage products or further modifications to the protein. A recent crystal structure of RipA suggests a protease labile loop exists between the N inhibitory and C terminal PG hydrolase domains; this loop is hypothesized to be the site of cleavage in vitro, which is required for RipA enzymatic activation [25]. We mutated candidate residues in the loop in an attempt to identify cleavage sites but were unsuccessful in blocking RipA processing (Figure S5). Truncated RipA species were found associated with the cell wall compartment and in culture filtrates. In the culture filtrate, a RipA fragment appeared as a single band at approximately 25 kDa (Figure 3C, asterisk). We demonstrated this signal was specific by C-terminally tagging endogenous RipA on the chromosome of M. smegmatis with a FLAG epitope. This 25 kDa species exhibited altered mobility due to the epitope and could be detected by both anti-RipA (Figure 3C, left panel) and anti-FLAG antibodies (Figure 3C, right panels). These results indicate that RipA exists physiologically in a smaller form than the predicted full length protein. The observation of RipA cleavage suggested this process could be required for the protein' s function in vivo. To test whether RipA processing is correlated with division, we titrated overexpression of RipASm C408A and quantified the amount of induction needed to mediate chaining. We found that low level overexpression with 30 ng/mL inducer was sufficient to cause chaining (Figure 4A) without saturating the processing machinery, since these cells did not accumulate full length RipA (Figure 4B). Instead, mildly overexpressed RipA was processed down to two sets of smaller species at around 23 kDa and 12 kDa (Figure 4B). We also saw a dose-dependent saturation of the endogenous processing capacity, with high induction leading to accumulation of full length RipA (Figure 4B), as well as loss of processed endogenous RipA (Figure S6A). When we quantified recombinant protein levels by comparing densitometry with endogenous RipA protein (Figure 4C), we found that even mild RipASm C408A overexpression at 30 ng/mL of inducer (processed recombinant protein is approximately 10% of endogenous wild type RipA levels) was sufficient to cause chaining and cellular toxicity (Figure 4A). Together with qPCR data showing that RipASm C408A induction does not affect endogenous RipA transcription (Figure S6B), these data suggest direct competition between the RipAsm C408A mutant and endogenous RipA for processing machinery, and that the processed, not full length, form of RipA is required for division. However, though correlated with function, the processed species we observed could be the product of an inactivating event. To test this, we took advantage of the observation that the M. tuberculosis homologue of RipA (RipATB) functions differently in M. smegmatis than its native counterpart, despite having the same general domain architecture (Figure S7). In contrast to RipASm, which is toxic when overexpressed, overexpression of RipATB in M. smegmatis, surprisingly caused no toxicity or cell morphological differences (Figure 5A, B), despite similar protein levels (Figure 3A). We examined whether RipASm toxicity was correlated with its processing by performing Western blot analysis on RipATB overexpressing M. smegmatis. In contrast to overexpression of RipASm, when wildtype RipATB is overexpressed we observed only a single full length band at 55 kDa (Figure 3A, right arrow). The absence of processing correlates with the lack of detectable RipATB toxicity in M. smegmatis. Thus, we hypothesize that proteolytic cleavage is required for activating RipA in vivo. However, it could be formally possible that RipATB cannot recognize M. smegmatis peptidoglycan or is not intrinsically active enough in M. smegmatis to cause morphological defects. To test if RipATB can be enzymatically functional in M. smegmatis, we deleted the predicted N-terminal inhibitory segment by fusing the truncated active domain of RipATB to the RipA secretion signal peptide (RipATB-AD). As a control, we also produced a construct in which the M. smegmatis RipA active domain (RipASm-AD) can be secreted. None of the strains produced growth defects when uninduced (Figure S8). As expected, RipASm-AD like full length RipASm, was fully functional when induced and disrupted cell wall integrity, leading to bulging of the cells and a concomitant growth defect (Figure 6A, B). When RipATB-AD was secreted, we found that it was also functional and behaved in the same way as RipASm-AD (Figure 6A, B). Thus, the catalytic domain of RipATB can be active in M. smegmatis, but full length RipATB is not toxic because it does not undergo efficient processing in M. smegmatis. Given the potentially toxic nature of hyperactive RipA we hypothesized that RipA processing and activation may be less robust in slow-growing mycobacteria in order to match their much slower rate of growth and consequent lower requirement for peptidoglycan hydrolysis. To investigate this model, we first determined whether RipA is processed in M. tuberculosis by overexpressing RipATB. By Western blot analysis, we found multiple immunoreactive smaller species of RipATB, suggesting processing in M. tuberculosis (Figure 7D, brackets). However, the induction of RipATB in M. tuberculosis did not produce morphological changes or growth defects, even after five days of induction (Figure 7A, B). This overexpression produced about 3 fold more protein (most of which is in the processed form) than endogenous full length RipA (Figure 7E), which is similar to the amount of overexpression needed to observe cell chaining in M. smegmatis with the RipASm C408A allele (Figure 4C). The lack of morphological changes in M. tuberculosis is also in contrast with the marked lethality of RipASm overexpression in M. smegmatis (Figure 1). One explanation for this dichotomy may be that the slow growth of M. tuberculosis might mask morphological or growth defects caused by RipA overexpression. To test this possibility, we bypassed RipA processing by secreting truncated RipATB-AD in M. tuberculosis. Induction of RipATB-AD produced a severe growth defect in M. tuberculosis (Figure 7C), and concomitant cell rounding (Figure 7A) similar to that seen when RipASm was overexpressed in M. smegmatis. These results show that unchecked RipA activity is toxic even to slow-growing mycobacteria. Since full length RipATB induction does not produce this toxicity, M. tuberculosis may have intrinsically less robust RipA processing than in M. smegmatis. Indeed, in M. tuberculosis, we observed a band of 55 kDa on Western blots probed with anti-RipA antisera. This band is the same size as full length RipATB and appears in both uninduced and induced samples (Figure 7D, arrow). As this form was not detected in wildtype M. smegmatis lysates (Figure 3A), it may represent endogenous, unprocessed full length RipA. The same full length band was also observed in M. bovis BCG cell lysates (Figure S9). These data, along with the active domain overexpression analysis, support the idea that slow-growing mycobacteria process RipA less efficiently in order to keep this potentially lethal activity in check.
Bacteria rely on peptidoglycan (PG) for shape and structure. The prevailing view of PG remodeling requires the concerted action of synthetic enzymes ligating new subunits into the existing PG lattice followed by hydrolysis of the PG sacculus by autolysins to allow cellular expansion or division. This process is accomplished through the action of large holoenzyme complexes in the periplasm consisting of both PG synthetic and hydrolytic enzymes. Disruption of PG synthesis in these complexes can dysregulate cognate PG hydrolases, which can then become autolysins that lyse the cell [26]. Thus, the coordination and regulation of PG hydrolases is a critical process for the survival of the bacterium. Here we find that RipA in M. tuberculosis and M. smegmatis can behave as an autolysin, resulting in the formation of spherical cells and lysis when overexpressed or dysregulated. Overexpression of a RipA dominant negative mutant not only causes loss of septal resolution and chaining but also leads to uncontrolled activity of endogenous PG hydrolases and lysis in a subset of cells. Thus, RipA requires downstream interactions to govern its correct function during septal resolution, as well as prevent lethal ectopic hydrolase activity. The relatively low amount of dominant negative RipA (about 10% of endogenous RipA) required for chaining suggests that the cell has finely tuned the amount of active RipA in the cell to near the level required for division; even loss of 10% of these active RipA species (which is manifest in a partial loss of endogenous RipA processing (Figure S6A) leads to a block in septal resolution. While it is clear that RipASm C408A overexpression can interfere with endogenous RipA activation, given its known interactions with two other PG remodeling enzymes that localize to the septum—RpfB and PBP1 [18], [24], [27]—it is possible the dominant negative mutant also incorporates into and inhibits functional PG remodeling complexes. A combination of these two activities may contribute to the RipASm C408A mutant' s potency at inducing chaining at relatively low levels of induction. Supporting the presence of regulatory RipA septal complexes, we showed chemical inhibition of peptidoglycan incorporating PBPs (of which the RipA binding partner PBP1 is a member) results in cell rounding and lysis. Loss of PG synthetic activity within a PG remodeling complex may allow cognate PG hydrolases (such as RipA) to become hyperactive and lyse the cell. We found that RipA depleted cells were specifically protected against meropenem-induced killing, but remained sensitive to other unrelated stresses. The depletion of the RipA likely affects the expression of RipB, which resides downstream in the same operon and has the same in vitro enzymatic specificity as RipA [14]. However, RipB is not essential for growth [28], and we have previously shown that RipA appears to be more phenotypically active than RipB in vivo. RipA, but not RipB, can complement the growth inhibition and cell chaining defects observed in the ripAB depletion strain [18]. While we cannot discount the possibility that RipB contributes to meropenem-mediated killing, it seems more likely that RipA is the main enzyme responsible for this lethal phenotype. Together, our data suggests that meropenem-induced killing is RipA dependent. However, we do observe a slight but significant difference in growth between RipA depleted cells in the presence and absence of meropenem (Figure 2C, lanes 2 and 4) that suggests there may also be some RipA independent growth inhibition (but not lysis) due to meropenem treatment. This may reflect the fact that meropenem can target several transpeptidases [21], [22]. Despite this, since RipA appears to mediate meropenem' s bactericidal capacity, and thus appears to be a more attractive target for drug development, we would expect that a chemical activator of RipA might act synergistically with meropenem treatment. From these data alone, it may be possible that a RipA inhibitor would be contraindicated in combination with meropenem, as it would antagonize the effect of PBP blockade, but we have previously observed that RipA depletion can sensitize cells to carbenicllin, a β-lactam antibiotic that also targets various transpeptidases [18]. In previous assays, in contrast to meropenem, carbenicillin sensitization required long term RipA depletion—it had no bactericidal effect on cells depleted for RipA in the same time scale as our meropenem studies (Figure S1B). These data suggest that extended treatment with a RipA inhibitor may weaken cells enough to cause sensitivity to PBP inhibitors to which the cell was previously resistant. It would be interesting to determine whether RipA blockade can, in fact, synergize with existing PG targeting antibiotics in vivo. Because of the threat of lethal autolysin activity, cells can control PG hydrolases through several, interconnected regulatory mechanisms. RipA is no exception, as we have found that in addition to protein interactions that modulate its function, RipA requires proteolytic activation. RipA exists primarily as smaller processed forms in M. smegmatis. Recent work with RipA in vitro has mapped a protease labile loop between a putative N terminal blocking domain and the C terminal p60 PG hydrolase domain [25]. The size of our truncated RipA species could contain the predicted size of the p60 domain itself after cleavage within this loop, but we were unable to determine the exact cleavage site (s) for RipA proteolytic activation in vivo using site directed mutagenesis—mutation of two pairs of highly scissile aspartate-proline peptide bonds [29] at DP301 and DP315 to alanines had no effect on the ability of RipA to be cleaved in M. smegmatis (Figure S5). This is consistent with the activation of Auto amidase in Listeria monocytogenes, which is also produced as a zymogen and becomes active only after proteolytic processing and removal of an N terminal inhibitory domain [30]. For Auto it was not possible to isolate single amino acid substitutions that abolish processing; instead, only deletion of the loop prevented proteolytic cleavage [30], which suggests that the activation loop is intrinsically labile and might be cleaved by many different proteases. Like Auto amidase, RipA' s labile loop can be cleaved by many proteases in vitro [25], and thus may be a target of several proteases in vivo. This may explain why we see a smear of RipA truncated species in wildtype mycobacteria, as opposed to a single truncated band. Given the work of Ruggiero et al [25], it was likely that RipA is produced as a zymogen in vivo, like Auto amindase. However, another recent report suggested a different effect of the N terminal domain in blocking RipA enzymatic activity [14], [25]. While both studies agreed that the N terminal domain appears to partially block the C terminal endopeptidase active site, the authors reached opposite conclusions as to whether the N terminus is inhibitory. Ruggiero et al [25] found that truncated RipA containing only the C terminal p60 domain was able to cleave purified PG, while full length RipA had minimal activity [25]. In contrast, Böth et al [14] showed that full length RipA was capable of degrading small synthetic PG fragments, and truncation of the N terminus produced no increase in enzymatic activity. However, in the latter work, the authors did not perform enzymatic digests using full length RipA on purified PG, as performed by Ruggiero et al. It is possible that the reported differences between these studies reflects the ability of small PG fragments to enter the RipA active site, despite partial occlusion by the N terminal domain, while access of larger substrates such as crosslinked and polymerized PG is blocked. Our results favor the zymogen model, as we have found that processing of the N terminal domain is required for full RipA enzymatic activation in vivo. Likewise, the lack of processing of RipATB in M. smegmatis likely accounts for the absence of its toxicity upon overexpression. While it is possible that full length RipA could serve some degradation function on smaller substrates in vivo, our results suggest that its main peptidoglycan remodeling activity requires removal of the N terminus, which contains a functional inhibitory domain. Furthermore, using the less efficiently processed RipATB homologue, we showed that protein interactions are not only necessary for regulating functional septal complexes but also promote RipA proteolytic activation. Full length RipASm is toxic when overexpressed in M. smegmatis, but full length RipATB does not produce the same phenotype. However, when we bypassed processing and expressed the truncated RipATB active domain in M. smegmatis, we observed a full gain of toxicity. These results suggest that the interactions between RipASm and the cellular factors necessary for processing do not occur with the RipATB homologue. Since septation is a highly conserved process, these data may not necessarily indicate different RipA binding partners in slow and fast growing mycobacteria but rather that the M. smegmatis and M. tuberculosis binding partners have evolved together and may have higher affinities for one another. Together, our work demonstrates that RipA regulation occurs at multiple levels post-transcriptionally. We did not see any transcriptional downregulation in cells transitioning into non-replicating conditions (Figure S4). In fact, there was a significant increase in RipA transcription during the transition to stationary phase, but the functional consequence of this observation remains unknown. Furthermore, while overexpression of the dominant negative RipASm C408A allele modulates processing of endogenous RipA (Fig S6A), this is due to competition for processing and not transcriptional feedback, even at high induction conditions (Figure S6B). This lack of transcriptional modulation is consistent with the observation that ripA expression has only limited variation across dozens of published experimental conditions (summarized on TBDB [31]), including general and antibiotic stresses. The only conditions under which ripA expression has been found to change are under non-replication conditions and, recently, when cells are blocked in cell division [32]. In the latter work, Plocinska et al found that ripA can be regulated by the MtrAB two-component system. Specifically, inhibiting septum formation prevents MtrB, which localizes at the septum, from activating the MtrA response regulator, leading to ripA downregulation. The authors proposed an interesting model in which ripA transcription could be upregulated by MtrB when it assembles at the division site; however, it remains unclear whether MtrAB regulation of ripA transcription occurs during normal growth or, instead, represents a stress response when cell division is inhibited. While the question of ripA transcriptional regulation during vegetative growth remains to be tested, our work suggests that post-translational mechanisms like processing may represent a key way of controlling RipA hydrolytic activity during growth. Therefore, we propose that protein-protein interactions help establish RipA function at the septum, where it is then aided in becoming proteolytically cleaved in the periplasm (Figure 8). After enzymatic activation, functional RipA can rely on both upstream and downstream protein interactions to help place it in the correct context during cell division—inhibition of cognate PG synthases can lead to dysregulated cell wall hydrolysis. The benefit of having multiple levels of RipA regulation is that the cell can exert a tighter control over RipA' s activation and potential autolysin activity. We found that RipA in M. tuberculosis is subject to less proteolytic activation than in M. smegmatis. The slow-growing mycobacteria like M. tuberculosis and M. bovis BCG might well have slower rates of PG hydrolysis and, consequently, reduced RipA activity. Indeed, we find significant amounts of full length RipATB in slow-growing mycobacterial lysates, a form that is not present in M. smegmatis lysates. Furthermore, while expression of the active domain of RipATB leads to severe growth inhibition with concomitant bulging, overexpression of full length RipATB has no such effect in M. tuberculosis, suggesting that slow-growing mycobacteria proteolytically activate less RipA than their fast-growing counterparts. The mechanism behind this additional control over RipA activation is not known, but may be at the level of expression or functional kinetics of the protease (s) responsible for RipA cleavage. In fact, there is an additional stretch of amino acids in RipASm compared to RipATB, which sits at the beginning of the N terminal inhibitory domain (Figure S7). Ultimately, an integrated mechanism for controlling PG hydrolases may represent a broad paradigm among cell wall degrading proteins. Multiple levels of regulation might be required to synchronize their activity to the cellular requirement while avoiding overactivity and toxicity. In support of this, expression of a dominant negative RipA allele at 10% of the endogenous RipA levels leads to abnormal chaining. Thus, M. smegmatis appears to carefully titrate the amount of processed RipA to nearly the minimum levels it requires for division. Finally, beyond division mechanics, post-translational PG hydrolase regulation has the added benefit of inducing changes quickly in response to changing environmental conditions, especially in times of low transcription such as non-replicative conditions [33], [34]. The byproducts of PG hydrolysis can act as sensors for the bacterial environment, whether in vitro or within a host. For example, in B. subtilis, muropeptides have been found to be sufficient to induce spore resuscitation [35], [36] while in M. tuberculosis, RpfB, a lysozyme that is known to be a RipA interacting partner [27], is required for regrowth from both in vitro and in vivo non-replication states [37], [38]. The exact mechanism behind mycobacterial resuscitation remains unclear, but muropeptide-based signaling could play a major role. In fact, we found processed RipA species in the culture filtrates and recent work by Mir et al [39] demonstrated that the addition of muropeptides to dormant M. tuberculosis facilitated resuscitation, possibly through the binding and signaling of the essential mycobacterial integral membrane kinase, PknB. Thus, soluble PG remodeling proteins might play a role in fostering communication across a bacterial population. In summary, this work has further defined two connected, but distinct, mechanisms to regulate the activity of RipA, a potential autolysin that is essential for septal resolution in mycobacteria. The complexity of this regulation, which involves protein interactions as well as proteolytic activation, underscores the importance of carefully coordinating cell wall hydrolysis during growth and division. By dissecting the molecular regulation of PG hydrolases, we gain fundamental insight into how the bacterial cell wall is dynamically maintained and also open up avenues for novel chemotherapeutics, especially against major human pathogens such as M. tuberculosis.
E. coli XL-1 Blue (Stratagene, Santa Clara, CA) were grown at 37°C in LB broth or agar and used for cloning. Selection was performed using kanamycin (50 µg/mL), hygromycin (100 µg/mL), ampcillin (100 µg/mL) or zeocin (25 µg/mL) when appropriate. Mycobacterium smegmatis mc2155 was grown at 37°C, unless otherwise indicated, in Middlebrook 7H9 broth supplemented with ADC (bovine albumin fraction V (Sigma) (5 g/L) -dextrose (2 g/L) -catalase (3 mg/L) and 0. 05% Tween80. Selection of M. smegmatis was achieved by supplementation of kanamycin (25 µg/mL), hygromycin (50 µg/mL) or zeocin (25 µg/mL). M. tuberculosis H37Rv and M. bovis BCG were grown in liquid Middlebrook 7H9 broth and plated on Middlebrook 7H10 agar supplemented with OADC (oleic acid-albumin-dextrose-catalase) (BD Biosciences, Franklin Lakes, NJ). M. smegmatis in which the RipA endogenous promoter has been replaced by a tetracycline inducible promoter was previously constructed and characterized in [18]. Mtb RipA (Rv1477) and Msmeg RipA (MSMEG_3145) mutants were constructed through PCR stitching using the following primers: Mtb C383A RipA Forward (CCGTCGGCTTCGACGCCTCAGGCCTGGTGTTG) Mtb C383A Reverse (CAACACCAGGCCTGAGGCGTCGAAGCCGACGG) Msmeg RipA C408A Forward (ACCGTCGGCTTCGACgcCTCGGGTCTGATG) Msmeg C408A Reverse (CATCAGACCCGAGgcGTCGAAGCCGACGGT). Msmeg RipA DP301AA DP315AA double point mutants were constructed by PCR stitching using the following primers: Msmeg RipA DP315AA Forward (GCGATCCCGAGCGCGTTCGTCAGCGGTGcCgCCATCGCGATCATCAAC) Msmeg RipA DP300AA Reverse (GAACGCGCTCGGGATCGCAGGCAGGGTCgCGgCCCACACGGCCCAGTT) Secreted, RipA catalytic domain constructs were made using the M. smegmatis RipA secretion signal amino acids 1–51—Reverse primer: GAACCTgatatcGACGAGCGTGGCGAG. The RipATB active domain contained amino acids 332–472, while RipASm active domain contained amino acids 357 to 494. Tetracycline inducible strains were created by cloning RipA genes into the Tet On plasmid, pSE100. For time-lapse microscopy, green fluorescent protein (GFP) was cloned downstream of RipASm C408A to create a transcriptional reporter. GFP was also cloned downstream in frame with RipASm to create a translational fusion. These inducible plasmids were then transformed into M. smegmatis in which the plasmid pMC1s, which encodes the tetR gene, had already been integrated at the L5 site. Recombinant gene products were expressed using a published anhydrotetracycline inducible system [40]. Anhydrotetracycline induction was performed with 100 ng/mL of anhydrotetracycline unless otherwise indicated. M. smegmatis with chromosomal ripA under the control of a tetracycline inducible promoter was grown in 7H9 ADC in the presence of 100 ng/mL anhydrotetracycline (aTc) to an OD of 0. 2. The culture was split, the cells pelleted at 5000 rpm for 10 minutes and resuspended in 7H9 ADC with (RipA replete) or without (RipA depleted) aTc for 6 hours. Once depletion in the no inducer culture was confirmed by microscopic examination of the culture for chaining, 10 µg/mL meropenem were added and the cultures and cells grown at 37°C for 6 hours. After 6 hours of meropenem treatment, both RipA replete and depleted cultures were serially diluted and plated onto LB supplemented with hygromycin, kanamycin and 100 ng/mL aTc. Plates were incubated for 3 days at 37°C and colonies were counted. As a control, RipA replete and depleted cells were also treated with 0. 08% SDS (w/vol), 0. 08 µg/mL streptomycin (Sigma) or 500 µg/mL carbenicillin (Sigma). M. smegmatis cells were grown overnight in 7H9 supplemented with dextrose (2 g/L), but not albumin or catalase at 37°C until mid log phase. This media should be made fresh for every experiment. Cells were pelleted and culture supernatants were precipitated with 10% TCA (tricholoroacetic acid) overnight at 4°C. Precipitates were pelleted at 15,000×g for 15 minutes at 4°C and washed once with ice cold acetone. The acetone wash was decanted, and the pellet was air dried at room temperature. Precipitated protein was resuspended in reducing SDS loading buffer at 65°C, for 10 minutes. M. smegmatis cells was fractionated by French press three times. Unbroken cells and insoluble material were pelleted at 1000×g for 10 minutes. The supernatant was collected and insoluble cell wall material pelleted at 27,000×g at 4°C for 40 minutes. The remaining supernatant was centrifuged at 100,000×g for 1 hour at 4°C to pellet the membrane fraction, while te supernatant contains the soluble cytosolic fraction. Rabbit polyclonal antibody was made from an affinity purified using a peptide derived from the Msmeg RipA epitope: NAGRKIPSSQMRRG (Genscript, Piscataway, NJ). Anti-RipA antibody was diluted to 1 mg/mL and used at a dilution of 1∶1000. Anti-FLAG antibody (Sigma, St. Louis, MO) was used according to manufacturer' s instructions. Protein samples were mixed with 4x Laemmli SDS PAGE buffer (Boston BioProducts, Inc, Boston MA) and boiled for 5 minutes. M. bovis BCG and M. tuberculosis protein samples were boiled for 20 minutes. Proteins were separated on 12% Tris-glycine polyacrylamide gels, transferred to PVDF membrane (Pall Corp, Pensacola, FL), probed with anti-sera and developed with SuperSignal chemiluminescent reagent (Thermo, Pittsburg, PA). Densitometry on Western blot signal was performed using Multiguage software (Fujifilm). For TMA-DPH (Invitrogen, Carlsbad, CA) staining, bacteria were centrifuged and media removed. Cell pellets were resuspended in 50 mM TMA-DPH in PBS and incubated in the dark for 10 minutes. Cells were also stained in FM4-64Fx (Invitrogen) at a concentration of 5 µg/mL in PBS for 10 minutes, and then fixed and stored in 4% paraformaldehyde. Samples were imaged using a Nikon TE-200E microscope with a 100x (NA1. 4) objective and captured with an Orca-II ER cooled CCD camera (Hamamatsu, Japan). Shutter and image acquisition were controlled using Metamorph Software (Molecular Devices). Final images were prepared using Adobe Photoshop 7. 0. Four Gene Frames (Fisher Scientific) were stacked onto a glass slide and filled with Middlebrook 7H9 in low melting point agar, supplemented with 50 µg/mL of hygromycin and 100 ng/mL aTc. A glass coverslip was flattened atop the agar to create a smooth surface and then removed after the agar set. The agar pad was sliced into eighths and seven of the pieces were removed to provide an air reservoir. Onto the remaining pad, exponential phase M. smegmatis was pipetted and allowed to adsorb until the surface of the pad appeared dry. Finally, a glass coverslip was applied and the slide was imaged on the microscope in an environmental chamber warmed to 37°C (Applied Precision, Inc.). Time-lapse images were acquired using a DeltaVision epifluorescence microscope with an automated stage enclosed with a 100x oil objective (Plan APO NA1. 40). Cells were imaged every 10 minutes for up to 18 hours using brightfield and fluorescence illumination (461–489 nm; Applied Precision, Inc.) and images recorded with a CoolSnap HQ2 camera (Photometric). Focus was maintained using the software-based autofocus (Applied Precision, Inc). M. smegmatis samples were collected at the indicated growth phases (log, OD600 = 0. 5; early stationary, OD600 = 2; Stationary, 24 hours of OD600>7) and stored in RNA Protect Bacteria Reagent (Qiagen, Valencia, CA) at −80°C. The pellets were then mechanically disrupted by beadbeatting for three 1-minute cycles, and RNA isolated using the RNeasy Mini Kit (Qiagen), with one additional DNAse treatment (Qiagen) on the column before elution and a second DNAse digestion with Turbo DNase according to manufacturer' s instructions (Ambion, Foster City, CA). Reverse transcription was carried out using the High Capacity cDNA Reverse Transcription kit (Applied Biosystem, Foster City, CA). Quantitative PCR reactions were set up in Power SyBr green PCR master mix (Applied Biosystems) and run and analyzed on a Step One Plus real time system (Applied Biosystems). ripA expression was measured using the following intragenic primers: 5′ CAGATCGGTGTGCCCTACTC; 5′ GGCGAACATGTAGAGCATCAG; or against the 3′ UTR region: 5′ GCTCGAGGCCCCTTACAC; 5′ GGAGCGCAAAGTAATCCCATCAG ripA expression was normalized to sigA levels, which utilized the following primers: 5′ AAGACACCGACCTGGAACTC; 5′AGCTTCTTCTTCCTCGTCCTC. | Peptidoglycan (PG) is a major component of the bacterial cell wall, which forms a flexible, but strong mesh around the cell to oppose osmotic pressure and prevent lysis. PG is also dynamically modified, continually being disassembled and polymerized as the cell elongates and divides. It remains poorly understood how cells can titrate enough hydrolysis of the PG to allow bacterial growth without leading to excessive digestion and disruption of cellular integrity. In our work, we have identified two methods by which a critical PG hydrolase, RipA, is carefully controlled in Mycobacterium tuberculosis—protein interactions help prevent lethal RipA dysregulation, while proteolytic cleavage is used as a second step to activate the enzyme in order to separate daughter cells. Our work elaborates multiple post-transcriptional mechanisms for preventing PG hydrolases from becoming lethal autolysins. These different levels of regulation may serve as a more general paradigm for PG remodeling in other bacterial species. | Abstract
Introduction
Results
Discussion
Materials and Methods | bacteriology
microbial physiology
biology
microbiology
microbial growth and development
bacterial pathogens | 2013 | Protein Complexes and Proteolytic Activation of the Cell Wall Hydrolase RipA Regulate Septal Resolution in Mycobacteria | 12,557 | 239 |
Despite worldwide mass drug administration, it is estimated that 68 million individuals are still infected with lymphatic filariasis with 19 million hydrocele and 17 million lymphedema reported cases. Despite the staggering number of pathology cases, the majority of LF-infected individuals do not develop clinical symptoms and present a tightly regulated immune system characterized by higher frequencies of regulatory T cells (Treg), suppressed proliferation and Th2 cytokine responses accompanied with increased secretion of IL-10, TGF-β and infection-specific IgG4. Nevertheless, the filarial-induced modulation of the host`s immune system and especially the role of regulatory immune cells like regulatory B (Breg) and Treg during an ongoing LF infection remains unknown. Thus, we analysed Breg and Treg frequencies in peripheral blood from Ghanaian uninfected endemic normals (EN), lymphedema (LE), asymptomatic patent (CFA+MF+) and latent (CFA+MF-) W. bancrofti-infected individuals as well as individuals who were previously infected with W. bancrofti (PI) but had cleared the infection due to the administration of ivermectin (IVM) and albendazole (ALB). In summary, we observed that IL-10-producing CD19+CD24highCD38dhigh Breg were specifically increased in patently infected (CFA+MF+) individuals. In addition, CD19+CD24highCD5+CD1dhigh and CD19+CD5+CD1dhighIL-10+ Breg as well as CD4+CD127-FOXP3+ Treg frequencies were significantly increased in both W. bancrofti-infected cohorts (CFA+MF+ and CFA+MF-). Interestingly, the PI cohort presented frequency levels of all studied regulatory immune cell populations comparable with the EN group. In conclusion, the results from this study show that an ongoing W. bancrofti infection induces distinct Breg and Treg populations in peripheral blood from Ghanaian volunteers. Those regulatory immune cell populations might contribute to the regulated state of the host immune system and are probably important for the survival and fertility (microfilaria release) of the helminth.
Helminths like filarial nematodes are tropical parasitic worms and the infections that they induce are classified as neglected tropical diseases (NTDs). Filarial infections are vector-borne diseases which are transmitted by blood-feeding insects that are common in tropical and subtropical countries. Although the majority of filarial infections remain in a regulated state, long-term chronic infections can cause overt diseases and individuals suffering from filarial-induced diseases are stigmatized and endure immense social and psychological burdens as well as financial losses which contribute to poverty [1]. For example, lymphatic filariasis (LF) is caused by Wuchereria bancrofti and Brugia spp. and can lead to the development of hydrocele, lymphedema, lymphangitis and elephantiasis causing a major public health problem and an overall elevation in disability-adjusted life years (DALY). Before mass drug administration (MDA) commenced, approximately 120 million people were infected with LF, and 40 million people suffered from disease-related pathologies. Therefore, the World Health Organization launched the Global Programme to Eliminate LF (GPELF) and MDA measures have cured or prevented 96 million new cases of LF over the last 13 years. It is now estimated that 68 million people are still infected and there are 19 million hydrocele and 17 million lymphedema cases [2]. As mentioned above, whereas a portion of humans develop severe forms of disease-related symptoms the majority of individuals retain a homeostatic and regulated state which is essential for the long-term survival of filariae [3–5]. Regulatory immune cells play a crucial role in the regulation of immune responses and indeed higher frequencies of regulatory T cells (Treg) were observed in LF-infected microfilaremic (MF+) and microfilariae negative (MF-) individuals compared to uninfected adolescents and individuals with lymphedema [6,7]. In addition, in vitro stimulation assays revealed that Tregs obtained from MF+ individuals suppressed proliferation and Th2 cytokine responses [8]. Furthermore, it was shown that the modified Th2 responses in MF+ individuals are accompanied with higher frequencies of Treg and alternatively activated macrophages as well as increased secretion of IL-10, TGF-β and infection-specific IgG4: all promoting parasite survival [9,10]. In addition to Treg, regulatory B cells (Breg) have been widely recognized as negative regulators of immune responses controlling autoimmunity and inflammation in suppressing pathological immune responses primarily through the secretion of IL-10 [11]. Indeed, it was shown that helminth infections induce IL-10-producing Breg populations [12–14] but the role of such immune cell subsets during filarial infection remains unclear. Thus, to decipher the role of regulatory immune cell subsets during LF, we analysed Breg and Treg frequencies in peripheral blood from uninfected endemic normals (EN), asymptomatic patent (CFA+MF+) and latent (CFA+MF-) W. bancrofti-infected individuals in Ghana. In addition, to elucidate the prevalence of distinct Breg and Treg subsets in individuals who had cleared the infection but suffer from W. bancrofti-induced clinical symptoms, we also profiled individuals with lymphedema (LE) who were CFA-MF-. Since MDA treatment against LF has been applied in Ghana since 2000 [15], we also analysed peripheral blood from individuals which were previously infected with W. bancrofti (PI) but had cleared the infection due to the administration of ivermectin (IVM) and albendazole (ALB). The composition and inclusion of the different patient groups allowed a detailed analysis of regulatory immune cell subsets in W. bancrofti-affected individuals (CFA+MF+, CFA+MF-, LE, PI) in comparison to EN. We observed that all W. bancrofti-infected individuals had significantly increased CD19+CD24highCD5+CD1dhigh and CD19+CD5+CD1dhighIL-10+ Breg as well as CD4+CD127-FOXP3+ Treg frequencies in the peripheral blood whereas IL-10-producing CD19+CD24highCD38dhigh Bregs were exclusively increased in patently infected (CFA+MF+) individuals. In addition, anti-filarial treatment and clearance of infection (PI group) lead to the reduction of Breg and Treg subsets to levels comparable with those from EN. In summary, the results obtained from this study show that distinct Breg and Treg subsets are induced during an ongoing W. bancrofti infection but return to homeostatic levels upon clearance of infection indicating a potential contribution to the filarial-specific immunity and survival of the parasite.
The studies were approved by the Committee on Human Research, Publications and Ethics at the School of Medical Sciences of the Kwame Nkrumah University of Science and Technology (KNUST), and Komfo Anokye Teaching Hospital, Kumasi, Ghana (CHRPE/AP/022/16), as well as by the Ethics Committee of the University Hospital of Bonn, Germany (018/12). Permission was also obtained from the Nzema East and Ahanta West District Health Directorates, Ghana. Before recruitment and sample collection commenced, meetings were held in the communities to explain in detail the purpose and procedures of the study. Verbal consent to perform the study in the villages was obtained from community leaders, i. e. , chiefs and elders of the selected communities, and written informed consent was obtained from all participants. The study was undertaken according to the principles of the Helsinki Declaration of 1975 (as revised 2008). In 2009, a case control study was conducted in 1774 Ghanaian volunteers within the health districts Nzema East and Ahanta West of the Western region of Ghana to identify genetic biomarkers which are associated with different manifestations of lymphatic filariasis (LF). In 2015, a total of 223 individuals from the initial study agreed to a follow up study and provided peripheral whole blood to characterize regulatory immune cell populations using flow cytometry technique. To assess W. bancrofti infection, night blood was obtained from the participants to determine the presence of MF since the nematode has a nocturnal periodic activity. Finger prick blood test, thick blood film smears and Sedgewick rafter counting technique were all performed. For the thick blood film technique, peripheral whole blood was applied on a glass slide, stained with Giemsa and examined for MF under the microscope at x10 magnification. In addition, 100μl whole blood was mixed with 900μl of 3% acetic acid, poured onto a Sedgewick rafter counting chamber (VWR, Langenfeld, Germany) and MF counts were examined using a microscope at x10 magnification. Furthermore, circulating filarial antigen (CFA) was detected using immunochromatographic card test (ICT) from the BinaxNOW® Filariasis kit (Alere, Cologne, Germany) according to the manufactures description. Lymphedema (LE) individuals were characterized based on the presence of oedema on the upper and lower limb extremities according to the “Basic Lymphedema Management Guidelines” established by Dreyer and colleagues [16]. At the time of sampling, LE individuals tested negative for both CFA and MF parameters, confirming previous studies showing that individuals suffering from lymphedema are usually MF and antigen negative [6,17]. In addition, since no red clay soils derived from volcanic deposits are present in the study region, podoconiosis-induced lymphedema cases were not observed [18,19]. A Malaria Pf Ag rapid test (Guangzhou Wondfo Biotech Co. Ltd, Guangzhou, China) was further applied according to the manufacturer’s instructions to determine Plasmodium infection. Other filarial infections were ruled out via blood smear analysis (e. g. Mansonella perstans) or absence at the study site (Onchocerca volvulus). 100μl whole blood from the participants were plated onto 96-well culture plates (Greiner Bio-One GmbH, Frickenhausen, Germany) and cultivated in 100μl RPMI-1640 medium (Sigma-Aldrich, Munich, Germany) including 10% bovine calf serum (BCS, Sigma-Aldrich). Whole blood cultures were then left un-stimulated or re-stimulated with eBioscience™ cell stimulation cocktail (PMA; Thermo Fisher Scientific, Schwerte, Germany) for 4h at room temperature. Thereafter, regulatory immune cell composition and function was analysed using flow cytometry. To obtain whole blood cells from the in vitro cultures, plates were centrifuged and supernatants removed. Red blood cells were then eliminated from the cultures using a red blood cell lysis buffer (Biolegend, San Diego, USA) and remaining cells were fixed and permeabilized using eBioscience™ fixation/permeabilization concentrate and permeabilization buffer (Thermo Fisher Scientific) according to the manufacture`s description. Thereafter, cells were stained with combinations of fluorophore (FITC, PE, PE-Cy7, APC) -conjugated anti-human CD1d (clone 51. 1), CD4 (clone RPA-T4), CD5 (clone UCHT2), CD19 (clone HIB19), CD24 (clone eBioSN3 (SN3 A5-2H10) ), CD38 (clone HIT2), CD127 (clone eBioRDR5), FOXP3 (clone 236A/E7), HELIOS (clone 22F6), IL-10 (clone JES3-9D7), eBioscience™ monoclonal antibodies from Thermo Fisher Scientific and CD304 (Neuropilin-1, clone 12C2) monoclonal antibody from Biolegend. Stained samples were stored at 4°C and kept in the dark. Within 7 days, the samples were transported to the Kumasi Centre for Collaborative Research in Tropical Medicine (KCCR) in Kumasi, Ghana and acquired using the BD Accuri™ Flow cytometer (BD Bioscience). Afterwards antibody expression levels were analysed using the FlowJo v10 software (FlowJo, LLC, USA). An overview of the analysed regulatory immune cell subsets with their corresponding flow cytometry markers is shown in Table 1. Statistical analyses were performed using the software SPSS (IBM SPSS Statistics 22; Armonk, NY) and the PRISM 5 programme (GraphPad Software, Inc. , La Jolla, USA). Variables did not meet assumptions to allow parametric analysis, therefore to compare more than two groups a Kruskal-Wallis-test was performed and, if significant, followed by a Dunn`s multiple comparison test for a further comparison of the groups. The Spearman`s rank correlation coefficient was applied to analyse rank correlations between two variables. Finally, stepwise multiple logistic regression analysis was performed to decipher possible confounders like gender, age or rounds of MDA on the immunological results. P-values of 0. 05 or less were considered significant.
An initial case control study performed in 2009 in the health districts Nzema East revealed 318 patent (CFA+MF+) and 397 latent (CFA+MF-) W. bancrofti infections as well as 246 lymphedema (LE) and 349 endemic normals (EN). MDA programmes (400mg ALB + 200μg/kg IVM once a year) in these areas by the Ministry of Health were running from 2009. Based on the initial study we re-visited the health districts in 2015 to determine W. bancrofti infections upon implementation of anti-filarial treatment and to analyse the composition and function of regulatory B and T cell subsets using flow cytometry. In total, we obtained 54 EN, 41 CFA+MF-, 13 CFA+MF+, 50 LE and 65 individuals who were previously infected (PI) with W. bancrofti (CFA+MF- or CFA+MF+) in 2009 but were now classified as CFA-MF-. All participants were negative for Plasmodium or other filarial infections. An overview about the characteristics of the study population is depicted in Table 2 and S1 Table. Peripheral whole blood was obtained from the participants (Table 2) and frequencies of regulatory B (Breg) and T cell (Treg) populations were analysed using flow cytometry according to the applied gating strategy (S1 and S3–S5 Figs). The frequencies of CD19+CD24high (Fig 1A) and CD19+CD24highCD5+CD1dhigh (Fig 1B) Breg subsets were significantly increased in latent (CFA+MF-) and patent (CFA+MF+) W. bancrofti-infected individuals when compared to EN and PI. Interestingly, a heamatological study in India reported that whole blood cell counts were increased in individuals presenting filariasis [20]. However, flow cytometry-based analysis of lymphocytes, according to the applied gating strategy here (S1 Fig), showed equal lymphocyte frequencies between the different groups (S2 Fig). Therefore, Breg subsets were induced by W. bancrofti infection and not the result of an overall lymphocyte expansion. Moreover, microfilaremic (CFA+MF+) individuals presented an overall higher frequency of those subsets and further analysis revealed positive correlations between MF counts and measured CD19+CD24high (r = 0. 2017, p = 0. 0025) and CD19+CD24highCD5+CD1dhigh (r = 0. 1801, p = 0. 0070) Breg subsets (Fig 1C and 1D, respectively). These findings show that W. bancrofti infections, especially patent ones, induce distinct Breg subsets and interestingly, in the PI cohort these population had levels comparable to the EN group indicating that they had returned to homeostatic levels. Since W. bancrofti infection induces Breg accumulation in the periphery, we further deciphered the functional role of Breg subsets. Bregs are predominately identified on their ability to produce IL-10 [11] which regulates autoimmunity [21] and suppresses T cell and cytokine responses [22,23]. In mice, IL-10-producing CD19+CD5+CD1dhigh Bregs are called B10 cells which were shown to be induced by LPS or PMA stimulation [24]. Therefore, peripheral whole blood cells were either left untreated (Fig 2A–2C) or stimulated with PMA (Fig 2D–2F) and the frequency of IL-10-producing CD19+CD5+CD1dhigh Bregs were analysed according to the applied gating strategy (S3 Fig). Without ex vivo stimulation the frequencies of CD19+CD5+CD1dhigh Bregs were by tendency increased in W. bancrofti-infected and LE individuals (Fig 2A) when compared to EN and PI groups. IL-10-producing CD19+CD5+CD1dhigh Bregs however, were significantly increased in W. bancrofti-infected groups compared to EN or LE (Fig 2B). In contrast to the significantly decreased frequency of CD19+CD24highCD5+CD1dhigh cells in PI individuals (Fig 1B), IL-10-producing CD19+CD5+CD1dhigh frequencies within un-stimulated peripheral whole blood cells from PI individuals were comparable to W. bancrofti-infected individuals. In addition, in LE individuals, frequencies of IL-10-producing CD19+CD5+CD1dhigh Bregs were significantly reduced when compared to W. bancrofti-infected and PI individuals (Fig 2B), showing that CD19+CD5+CD1dhigh Bregs were functionally impaired. Further analysis revealed no significant correlation between MF counts and un-stimulated CD19+CD5+CD1dhighIL-10+ frequencies (r = 0. 1218, p = 0. 0669; Fig 2C). However, PMA ex vivo stimulation again revealed significantly increased frequencies of CD19+CD5+CD1dhigh Bregs in LE compared to EN and PI individuals (Fig 2D) but no differences could be observed between the different cohorts with regards to IL-10 production (Fig 2E). Again, no significant correlation between MF counts and PMA-stimulated CD19+CD5+CD1dhighIL-10+ frequencies were observed (r = -0. 0288, p = 0. 0669; Fig 2F). Besides CD19+CD5+CD1dhighIL-10+ Breg populations, IL-10-producing immature B cells (CD19+CD24highCD38highIL-10+) are a crucial immunomodulating cell subset in humans since they suppress effector T cell responses as well as Th1 and Th17 differentiation, promote the conversion of CD4+ T cells into regulatory T cells (Treg) and type 1 regulatory T cells (Tr1) and play additional roles during autoimmunity, HIV infection and graft-versus-host disease [21,25–27]. To analyse IL-10-producing immature B cell frequencies in W. bancrofti-infected individuals, peripheral whole blood cells were left either untreated (Fig 3A–3C) or stimulated with PMA (Fig 3D–3F) and frequencies were analysed according to the applied gating strategy (S4 Fig). Whereas no differences in the frequency of CD19+CD24highCD38high Breg subsets could be observed between the groups (Fig 3A), CFA+MF+ had significantly increased CD19+CD24highCD38highIL-10+ frequencies when compared to EN and PI without ex vivo stimulation (Fig 3B). In addition, albeit weak, further analysis revealed a positive correlation between MF counts and un-stimulated CD19+CD24highCD38highIL-10+ frequencies (r = 0. 1801, p = 0. 0070; Fig 3C). Again, upon PMA ex vivo stimulation, no differences of the frequencies could be observed between the different groups (Fig 3D and 3E), although positive correlation between MF counts and PMA-stimulated CD19+CD24highCD38highIL-10+ frequencies were significant (r = 0. 1527, p = 0. 0226; Fig 3F). Overall, these findings suggest that besides CD19+CD5+CD1dhighIL-10+ Bregs, IL-10-producing immature B cells are also induced by W. bancrofti infection and that individuals from the different cohorts have the same potential to produce IL-10 upon ex vivo stimulation. Besides Bregs, Treg (CD4+CD25+ and/or CD4+FOXP3+) were shown to be induced during human filarial infections [8,28,29], but the role of distinct Treg subsets and their precise role during lymphatic filariasis remains unclear. FOXP3+ Tregs can be divided into natural or thymic-derived (tTreg) and peripherally induced Treg (pTreg) [30]. Furthermore, Neuropilin-1 and HELIOS were declared as potential markers for tTreg [31,32] and were used to discriminate FOXP3+ Treg populations here, see applied gating strategy (S5 Fig). Indeed, without ex vivo stimulation, CD4+CD127-FOXP3+ Treg frequencies were increased in the entire W. bancrofti-infected cohort and LE individuals when compared to EN and PI (Fig 4A), confirming that filarial infections promote Treg accumulation [8,28,29]. In addition, further analysis revealed a positive correlation between MF numbers and CD4+CD127-FOXP3+ frequencies (r = 0. 1454, p = 0. 0299; Fig 4B). In regards to the discrimination of tTreg and pTreg, Neuropilin-1 and HELIOS expression on CD4+CD127-FOXP3+ Treg remained unaltered between CFA+MF+, CFA+MF- and LE, but the PI group showed significantly decreased Neuropilin-1 and HELIOS frequencies compared to the EN and CFA+MF- group (Fig 4C and 4D). Since Neuropilin-1 expression was unaltered in the W. bancrofti-infected and LE individuals it suggests that the increased frequency of CD4+CD127-FOXP3+ cells (Fig 4A) are due to the induction of pTreg. The overall data set of the study is shown in S1 Table.
The underlying mechanisms as to why many LF-infected individuals remain in a homeostatic state are still not fully resolved. No doubt multiple subtle triggers and interactions by this nematode on the host contribute to this unique relationship. Although, B cell immunoregulatory mechanisms have been reported in a murine model of Brugia pahangi [31], this study documents primary evidence of functional Breg populations in W. bancrofti infections and reveals how distinct Breg subsets contribute to this overall picture in man: active functional Breg populations that subside upon treatment. In terms of immune-regulation several studies have shown that besides increased levels of IL-10, TGF-β, filarial-specific IgG4 and frequencies of alternatively activated macrophages [9,10], filarial-infected individuals harbour increased Treg frequencies too [7] with higher expression levels of FOXP3, CTLA-4, TGF-β and PD-1 on isolated PBMCs [10]. Expanding on those findings, we show here that CD4+CD127-FOXP3+ Treg frequencies were higher in W. bancrofti-infected individuals and this included the LE group. Moreover, individuals who had cleared the infection due to MDA participation had Treg frequencies comparable to EN indicating that the cells are only required during active infections. In addition, there are two subpopulations of FOXP3+ Treg called tTreg and pTreg [30], but their role in W. bancrofti infections remains uncertain [28]. Therefore, we deciphered the frequencies of these subsets using the markers Neuropilin-1 and HELIOS [32,33] but revealed no differences in Neuropilin-1 and HELIOS expression on CD4+CD127-FOXP3+ Tregs. Interestingly, more recent studies have indicated that HELIOS is not a definite marker for tTreg [34,35] and thus, these findings need to be critically assessed. Nevertheless, since the frequencies of Neuropilin-1 were equal between the CFA+MF+ and CFA+MF- we suggest that the CD4+CD127-FOXP3+ Treg levels in peripheral blood of W. bancrofti-infected individuals depend on the induction of pTreg which indeed were shown to be generated in the periphery upon antigen exposure [30]. And moreover were classified as effective suppressors [36]. In association, CD4+CD25high Tregs obtained from microfilaremic (MF+) Brugia malayi-infected individuals suppressed proliferation and Th2 cytokine responses [8] and a limitation to the current work is the lack of functional suppression assays using the observed regulatory populations. Therefore, future studies should focus on identifying filarial-specific inhibition by those subsets using W. bancrofti or Brugia antigen extracts. Besides the CD4+CD25+FOXP3+CD127- Treg population IL-10-producing regulatory type 1 T cells (Tr1) were detected in filarial-infected individuals [7,37] whereas we recently showed that CD4+α/βTCR+CD49b+LAG3+ Tr1 were decreased in peripheral blood of M. perstans-infected individuals [29]. However, further in depth analysis including Treg markers like CD103, CTLA-4, GITR, ICOS, LAG-3 and TGF-β which were shown to be important for the characterization and function of Treg [30,38–40] need to be performed to decipher the role of pTreg during filarial infection in more detail. Indeed, one limitation of this study was that flow cytometry analysis of peripheral blood was performed in Ghana using the BD Accuri™ Flow cytometer which only allowed 4 colour-based analysis and thus, as mentioned above, characterization of the different regulatory immune cell populations was restricted. Nevertheless, this study indicate that W. bancrofti infection induces pTreg, which were shown to mediate their suppressive function through CTLA-4, GITR, LAG-3, and membrane-bound TGF-β [39–42]. Besides Tregs, the expanding family of Bregs and their role in suppressing pathological immune responses and during helminth infections has recently been recognized [11,14,43,44]. For example, it was shown that various helminth infections induce IL-10-producing Breg populations [12–14,45] which were shown to be antigen-specific during chronic schistosomiasis [46,47], but the distinct role of Breg populations during W. bancrofti infection has remained largely unclear. Recently, we showed that M. perstans-infected individuals harbour high frequencies of CD19+CD24highCD38highCD1dhigh Bregs when compared to uninfected individuals [29] demonstrating that Breg populations are part of the cellular composition that retains a balanced immune reaction to M. perstans infections. Furthermore, we now reveal that W. bancrofti-infected individuals had increased CD19+CD24highCD5+CD1dhigh, IL-10-producing CD19+CD5+CD1dhigh and CD19+CD24highCD38dhigh Breg frequencies in peripheral blood. Interestingly, it was shown that patients with multiple sclerosis (MS), who were co-infected with helminths had increased frequencies of IL-10-producing CD19+CD1dhigh B cells which suppressed T cell proliferation and IFN-γ production leading to a better clinical outcome in regards to MS [13]. With regards to the observed IL-10-producing CD19+CD24highCD38dhigh Breg population, previous studies have shown that immature B cell populations (CD19+CD24highCD38dhigh) can produce high amounts of IL-10 upon CD40 engagement leading to suppression of Th1 and Th17 differentiation and conversion of CD4+ T cells into Treg and Tr1 cells [21,25]. Consequently, this IL-10 producing immature B cell population was shown to influence and modulate immune responses during autoimmunity, HIV infection and graft-versus-host disease [11,21,25–27]. In detail, this Breg subset was shown to maintain tolerance and long term remission during autoimmunity [48,49] and was associated with reduced rejection rates upon kidney transplantation [50] but also may contribute to immune dysfunction in HIV infection through the suppression of HIV-1 specific CD8+ T cell responses [26]. However, here we show that this IL-10-producing immature B cell population is present in peripheral blood of W. bancrofti-infected individuals and was positively correlated with MF release suggesting that the induction of this Breg population promotes fertility and survival of the parasite. With regards to the observed CD19+CD24highCD5+CD1dhigh and IL-10-producing CD19+CD5+CD1dhigh Breg populations, to our knowledge this is the first study that shows the presence of so-called B10 cells in W. bancrofti-infected individuals. Several studies observed that B10 cells can be induced upon LPS or PMA stimulation in mice and have been shown to suppress inflammation [24]. As with other Breg subsets, the suppressive function of B10 cells depend on CD40 engagement [51] and several experimental mouse models have proven the efficacy of B10 cells in dampening autoimmunity [52,53]. However, precursor B10 (B10Pro) and B10 cells were also identified in humans and it is suggested that their development depend on LPS and CpG stimulation and CD40 ligation [23,54]. Since we observed increased B10 cell frequencies in W. bancrofti-infected individuals without ex vivo stimulation we suggest that especially the MF provide the stimuli that drive B10 development in peripheral blood. Indeed, several studies already showed that inflammatory responses by filariae are mediated by TLR-inducing activity from the endosymbiotic Wolbachia bacteria [55–57] which are released from dying MF [58–60]. This study shows that there is an elevation of distinct IL-10-producing Breg subsets during W. bancrofti infection. These Breg subsets could potentially regulate host immunity through the secretion of IL-10 which has been shown to induce immunosuppressive alternatively activated macrophages as well as IgG4, which inhibits the function of various immune cells [9,61–64]. In addition, Brugia pahangi infection experiments in mice revealed that B cell populations and IL-10 secretion play an important role in filarial-driven immunomodulation [31]. Nevertheless, further studies need to decipher whether these distinct Bregs mediate their suppressive function through other molecules like TGF-β that were shown to modulate immune responses during patent filarial infections [28,42]. Although the analysed regulatory immune cell subsets comprised only a small percent of the overall lymphocyte population, we do consider that they are relevant in shaping host immunity during W. bancrofti infection since levels returned to those found in EN after infections were cleared. Indeed, recent studies showed that serial single cell adoptive transfer experiments and even low numbers of CD8+ T cells were effective against Listeria monocytogenes in a murine infection model [65,66], indicating that specificity, education and functional relevance is more critical than cell number. Ghana was one of the first countries in which the MDA against LF was implemented and programmes now cover the whole country [15]. Thus, we were able to analyse whether MDA affects the frequencies of regulatory immune cells in peripheral blood by revisiting our study cohort from 2009. During this unique opportunity we recruited 65 individuals who were previously infected (PI) with W. bancrofti (CFA+MF- or CFA+MF+) but were now classified as CFA-MF-. Indeed, treatment and thus clearance of infection lead to the reduction of Breg and Treg populations in the peripheral blood showing that ongoing W. bancrofti infections and especially MF release, appears to maintain regulatory immune cell development. In addition, sub grouping of the PI cohort into individuals who were CFA+MF- or CFA+MF+ in 2009 did not reveal any differences. Nevertheless, another limitation of this study was the lacking diagnosis of soil transmitted helminths (STHs; ethical clearance did not cover this element) which were shown to induce Breg subsets in previous studies [12–14,43–47]. However, our previous publication on immune profiling in the same region in Ghana showed low STH infection rates (6. 3%) and ruled out co-infections as potential confounders [5]. Thus, we consider that the low prevalence of STH in the study region and the distribution of STH throughout all patient groups argues against STH being a bias for the findings in this study. In addition, a study about B cell subsets and their immune responses in Schistosoma haematobium-infected individuals in Gabon indicated that parasitic co-infections were also not an important confounder [67] Therefore, we suggest that the differences in Treg and Breg frequencies are not due to STHs and can be rather explained by the W. bancrofti infection, especially since Breg and Treg frequencies showed a positive correlation with MF counts. In addition, we performed a stepwise multiple logistic regression analysis and revealed only gender as confounder confirming previous results showing that females are more resistant to infection compared to men [68] and are outperforming their male counterparts in terms of MDA intake and thus compliance [69]. Indeed, rounds of MDA were not revealed as confounder showing that clearance of infection rather than rounds of treatment per se influence infection and thus frequencies of regulatory cell subsets. In conclusion, this study presents initial evidence that IL-10-producing immature and B10 regulatory B cells were induced during an ongoing W. bancrofti infection in man, especially in patently infected individuals. These data contribute to the growing body of evidence about the complex nature of filarial-induced regulatory mechanisms in the host and indicate an important role of IL-10. In addition, MDA diminished the frequencies of regulatory immune cells in peripheral blood, suggesting that only active W. bancrofti-infections induce regulatory B cells in the periphery to shape host immunity. Further studies have to elucidate if re-infections of cured individuals (PI group) lead to enhanced induction of IL-10-producing immature and B10 regulatory B cells and the possible role of memory B cell activation. The capacity of filariae to modulate the host’s immune response is well reported but little is known about the long lasting impact of infection following cure. This study provided a unique opportunity to follow-up on a cohort after several years of MDA and moreover, compares regulatory cell profiles of those individuals with those presenting ongoing infections. The findings also revealed that distinct subsets of Breg populations were elevated during infection and moreover, that these populations had returned to base-line levels in the PI group. These data therefore provide initial evidence that certain filarial-specific cell populations are transient and decline following cure. Whether these are retained within the memory pool requires further investigation but it does underline that besides Treg populations, subsets of regulatory B cells play a crucial role within the complex host-filarial regulatory mechanisms and pathways. | Regulation of the host`s immune system by filarial nematodes is crucial for the fertility and survival of the nematode. Indeed, the majority of W. bancrofti-infected individuals are characterized by a regulated state including increased regulatory T cells (Treg), IL-10, TGF-β and filarial-specific IgG4 and suppressed Th2 cytokine responses. However, the functional role of Treg populations and regulatory B cells (Breg) during filarial infection remains unknown. Thus, in this study we investigated whether W. bancrofti-infected individuals from Ghana harbored distinct Breg and Treg populations which might be important for filarial-specific immunomodulation. Overall, this study shows that W. bancrofti induces distinct Breg populations, especially in patently (microfilaremic) infected individuals who presented significantly increased frequencies of IL-10-producing CD19+CD24highCD38dhigh Breg. Furthermore, clearance of the infection, due to anti-filarial treatment, returned these regulatory immune cells to homeostatic levels showing that an ongoing filarial infection is important for the activation of distinct Breg and Treg subsets. Those regulatory immune cell subsets are a part of a complex system which are induced by filarial nematodes to modulate the host`s immune system and maintain long-term survival. | Abstract
Introduction
Methods
Results
Discussion | blood cells
invertebrates
medicine and health sciences
immune cells
body fluids
pathology and laboratory medicine
immunology
parasitic diseases
animals
signs and symptoms
edema
white blood cells
wuchereria bancrofti
animal cells
wuchereria
t cells
lymphedema
antibody-producing cells
helminth infections
eukaryota
diagnostic medicine
blood
cell biology
b cells
anatomy
physiology
nematoda
biology and life sciences
cellular types
regulatory t cells
organisms | 2019 | Wuchereria bancrofti-infected individuals harbor distinct IL-10-producing regulatory B and T cell subsets which are affected by anti-filarial treatment | 8,397 | 320 |
Pleiotropy has been suggested as a novel mechanism for stabilising cooperation in bacteria and other microbes. The hypothesis is that linking cooperation with a trait that provides a personal (private) benefit can outweigh the cost of cooperation in situations when cooperation would not be favoured by mechanisms such as kin selection. We analysed the theoretical plausibility of this hypothesis, with analytical models and individual-based simulations. We found that (1) pleiotropy does not stabilise cooperation, unless the cooperative and private traits are linked via a genetic architecture that cannot evolve (mutational constraint); (2) if the genetic architecture is constrained in this way, then pleiotropy favours any type of trait and not especially cooperation; (3) if the genetic architecture can evolve, then pleiotropy does not favour cooperation; and (4) there are several alternative explanations for why traits may be linked, and causality can even be predicted in the opposite direction, with cooperation favouring pleiotropy. Our results suggest that pleiotropy could only explain cooperation under restrictive conditions and instead show how social evolution can shape the genetic architecture.
Recent empirical breakthroughs in microbial genetics have suggested a novel mechanism for maintaining cooperation [1–9]. The problem with cooperation is that noncooperative cheats can avoid the cost of cooperating whilst still gaining the benefits, allowing them to outcompete cooperators [10]. Consequently, without some mechanism to favour it, cooperation would not be evolutionarily stable. In microbes, clonal growth means that interactions will often be between close relatives, and so cooperation will be favoured because it is directed towards individuals that share the genes for cooperation [10–16] (kin selection). Experiments on both slime moulds and bacteria have led to the suggestion that pleiotropy provides a novel way to stabilise cooperation when it would not otherwise be favoured by another mechanism, such as kin selection [1–6]. These experiments have discovered cases in which two traits are either controlled by the same gene or coregulated (pleiotropy) and in which these two traits are a cooperative trait and a trait that provides a personal (private) benefit. For example, in the bacteria Pseudomonas aeruginosa, the quorum sensing signalling network controls both: (1) the production of several factors that are excreted from the cell and provide a cooperative benefit to the local population of cells [17–22] (public goods) and (2) private traits such as the production of intracellular enzymes involved in metabolism, and cyanide resistance [3–5]. Consequently, a mutant that did not respond to quorum sensing would avoid the cost of cooperatively producing public goods (cheats) but also pay the cost of not performing the private traits. If this cost of not performing the pleiotropically linked private goods was high enough, this could stabilise cooperation, when it would otherwise be outcompeted by cheats. In different papers, this linkage is referred to as pleiotropy, coregulation, or metabolic constraint, and it has been suggested as a key mechanism for stabilising cooperation, alongside factors such as kin selection and policing [1–9,23,24] (S1 Table). However, these observations of pleiotropy between cooperative and private traits do not necessarily imply that pleiotropy is stabilising cooperation by preventing the invasion of cheats, as alternate explanations are possible. One possibility is that the private and cooperative traits are favoured independently, and it is just more efficient to regulate them together (pleiotropically). Another issue is the extent to which the pleiotropy hypothesis relies on the underlying genetic architecture being relatively fixed, imposing a constraint on evolution. If the genetic architecture could itself evolve, then mutation could produce individuals that still performed the other, relatively essential trait but that did not cooperate. In line with this prediction, the emergence of such cheats has been shown experimentally in the bacteria P. aeruginosa, where private and public goods are linked by quorum sensing [21]. When cooperation was not favoured, individuals rewired the control of the cooperative trait, thereby stopping its production while still producing the private trait. This resulted in a greater competitive ability relative to cooperators [18,21,25]. Another possibility is that the causal link could be in the opposite direction, with cooperation favouring pleiotropy rather than pleiotropy stabilising cooperation. For example, consider a scenario, without pleiotropy, in which cooperation was favoured by a mechanism such as kin selection. If a noncooperative mutant arose, it might initially increase in frequency locally but would eventually be outcompeted by cooperators. Nonetheless, the initial spread of these noncooperators would impose a short-term cost to the cooperators from which they arose and with which they would be competing [26–29]. Pleiotropy between cooperation and another privately beneficial trait would provide a mechanism to prevent the initial spread of cheats and hence could be favoured to reduce or avoid this ‘cheat load’ [28,30]. We theoretically model the conditions required for either pleiotropy to stabilise cooperation or the alternate possibility that cooperation favours pleiotropy. We consider a form of cooperation common in bacteria, in which cells produce a factor that is excreted from the cell and which then provides a benefit to the local population of cells [31]. This can be modelled as a public goods game, in which individuals can pay a cost to perform a task that benefits the local group. We first consider a simple scenario analytically, for which the underlying architecture is fixed and does not evolve. This analysis formalises previous verbal arguments for how pleiotropy could stabilise cooperation. We then use an individual-based simulation approach to test whether allowing the genetic architecture to evolve influences the extent to which pleiotropy is evolutionarily stable. This simulation approach also allows us to test the alternate possibility that cooperation favours pleiotropy.
To provide a baseline, we consider the cooperative production of public goods in a deliberately simple scenario, where there is no pleiotropy with other traits [32–35]. Our aim is to examine what happens in the simplest (null) case without recourse to details that will vary across species. We assume a haploid population subdivided into a large number of local patches, each of size N, where the mean genetic relatedness between individuals is r [36]. Individuals can secrete a public good molecule at personal cost c. Production of this public good provides a benefit b that is shared equally among all patch members, even cheats that do not contribute. We assume that this cost and benefit influences fecundity and that after reproduction, adults die, and all offspring disperse to compete globally for some new patch to start the life cycle again (nonoverlapping generations and no kin competition). The production of the public good molecule will be favoured when rb̃ − c̃ > 0, where b̃ = (N − 1) b/N, −c̃ = b/N − c (Eq 3 of Methods). The b̃ term represents the average benefit that public goods production provides to the other members of the group. The −c̃ term represents the benefit provided to the actor by their own public good production (b/N), minus the cost of producing that public good (c). This standard result is just Hamilton’s rule for a linear public goods game, showing how cooperation can be favoured by kin selection if the indirect benefit to relatives (rb̃) outweigh the direct cost (c̃) [10,34,37–39]. We then examined the scenario in which the cooperative trait is linked pleiotropically to a private trait that is important for reproduction. Pleiotropic links between private and public traits have been discussed in two ways in the microbial literature, either because they are controlled by the same gene, or because they are controlled by two linked genes. For example, in Dictyostelium discoideum, mutants in the single gene dimA both do not receive the signal for the cell to differentiate into a prestalk cell and are excluded from spores [1]. In contrast, in P. aeruginosa, different genes control different traits, but these traits are linked by the quorum sensing system [2]. We consider both types and shall return to the extent to which they represent pleiotropy. In this section, we assume that the cooperation and private traits have to be linked. This could be due to either (1) the two traits being controlled by a single gene, as in the D. discoideum example, or (2) the two traits being controlled by separate genes but the activity of these genes being linked in a way that cannot be broken by mutation (no mutational accessibility). We assume that not performing the private trait reduces an individual’s fecundity by an amount d. We compared the fitness of pleiotropic cooperators with individuals that lacked this pleiotropic trait and therefore neither cooperated nor performed the private trait, as has been done in laboratory experiments [2–4]. In this scenario, switching from defection to pleiotropic cooperation still changes an individual’s inclusive fitness by rb̃ − c̃, but the direct fitness cost c̃ is decreased by d (Eq 4). Hence, the area where cooperation is favoured is larger when cooperation is linked pleiotropically with a private trait and where that private trait provides a relative reproductive advantage (Fig 1 and S1 Fig). An extreme case of this model is when pleiotropy links cooperation to an essential private trait. We considered this specific scenario because, in the empirical examples in which pleiotropy has linked cooperation to a private trait, the fitness of individuals that do not perform this private trait could be effectively zero [2–5]. We investigated this case by assuming that the fitness of noncooperators is zero and that the public good is equally distributed to all remaining pleiotropic cooperators. In this case, pleiotropic cooperation provides a direct fitness benefit (–c̃ > 0), and pleiotropic cooperators always interact exclusively with other pleiotropic cooperators, and so relatedness does not matter (b̃ = 0; Eq 5). Consequently, noncooperators are never able to outcompete pleiotropic cooperators. Both previous verbal arguments and our above results have focused on how pleiotropy can help stabilise cooperative traits that benefit other individuals [1–8] (b > 0). However, our above analysis reveals that pleiotropy can also help favour traits that otherwise provide no benefits (b = 0, c > 0) or even harm (b < 0, c > 0) both their bearer and other individuals on the patch (Fig 1 and S1 Fig). Consequently, rather than pleiotropy stabilising cooperation per se, pleiotropy can help stabilise any type of trait. For example, consider the extreme case in the opposite direction, of a harming trait that reduces the fitness of everyone in the social group, including the individual that produces it (b < 0). This would be analogous to bacterial cells producing an antibiotic or bacteriocin to which they are not resistant. If the genetic architecture is fixed, this harming trait would be favoured if the cost to private-good nonproducers (d) outweighs the sum of the cost of performing the harming trait (c), the cost of harming relatives ([N– 1]b/N), and the cost of harming oneself (b/N). An analogous point has been made in the cooperation in human literature, in which it was pointed out that punishment could help stabilise anything and not just cooperation [40]. Our above analyses have assumed that pleiotropy is a fixed constraint on the genetic architecture, such that individuals can either perform both the private and the public trait or perform neither. This could occur via the two traits being controlled by a single gene or by two linked genes whose link cannot be broken by mutation. However, the genetic architecture can evolve over time [21,41–43], and assuming that it cannot evolve could result in misleading conclusions [44,45]. For example, there is no reason that the private traits that have been observed to be pleiotropically linked to cooperation in bacteria could not in principle be independently regulated [21,46,47]. The key point here is that mutational accessibility, or pathway rewiring, could break the pleiotropic link, making traits independent. Previous verbal arguments and our above model have assumed that neither mutation nor different expression pathways can unlink the cooperation and private traits. We used an individual-based simulation approach to investigate the consequences for cooperation of allowing mutation to unlink the cooperation and private traits and hence allow the genetic architecture to evolve. To approximate a microbial life history, we assumed a haploid population in which subpopulations grow with clonal reproduction on patches before dispersing to compete globally for new patches. Because reproduction is clonal, both traits are perfectly transmitted to daughter cells. The population is subdivided into np = 100 patches. At the beginning of the life cycle, each patch is colonised by NF = 100 founders, which leave N = 10 initial individuals on the patch. In order to vary relatedness across the entire range from r = 0 to r = 1, we vary the likelihood with which these 10 initial individuals are sampled from the same parent founder. This allows us to vary relatedness without varying the number of initial cells, as would occur when different numbers of lineages colonise a patch and can then grow clonally [48]. The cells on a patch potentially produce a public good and grow for k = 10 generations. We assume that individuals have a baseline growth rate of 1 + g so that population size grows. Mutations occur during the growth phase. During the 10 generations of growth, only parents die after each reproduction event, and therefore, there is no population regulation, and generations are nonoverlapping. After these 10 generations, the remaining cells disperse globally and begin the life cycle again. For comparison, we first analysed the scenario in which there is no pleiotropic private trait. Our simulation results for this case were in close agreement with our analytical predictions—cooperation is favoured as predicted by Hamilton’s rule (Fig 2A). Cooperation in our simulation evolves under a slightly larger area than predicted by our analytical model, because relatedness increases during the growth phase, especially at low initial relatedness (see top-left corner of Fig 2A and 2B and S2 Fig). We then considered the scenario in which cooperation is always pleiotropically linked to a private trait and in the specific case when not producing the private good leads to the individual not reproducing (i. e. , their fecundity is 0). In this case, the only possible mutation is to not express both traits, as has been assumed previously [1–9,23]. We focused on this extreme scenario, as a conservative assumption, because it is the case in which pleiotropy could be most effective in stabilising cooperation. Again, our simulation results were in agreement with our analytical results—pleiotropic cooperation is always favoured over noncooperators regardless of relatedness, but the population goes extinct when the average fecundity is lower than 1 (S3 Fig). In the supplementary information, we relaxed the assumption that private nonproducers have fecundity 0 and let them reproduce, albeit with a baseline fitness that is reduced by d. We show that pleiotropic cooperation prevails as long as the cost of not expressing the private trait d is sufficiently high (S4 Fig). We then examined the consequences of allowing the genetic architecture to evolve (Fig 3). We allowed mutation from a genotype in which both the private and public goods are linked pleiotropically (pleiotropic cooperators) to one in which they are independent (nonpleiotropic cooperators) and vice versa. Hence, pleiotropy can be acquired or lost (mutations I and II in Fig 3). Pleiotropic and nonpleiotropic cooperators are otherwise phenotypically similar. As a consequence, pleiotropy is now also in competition with all nonpleiotropic genotypes, for which the private and public good traits are independent (Fig 3). Once traits are independently regulated, a single mutation will influence only one of the traits and not both, as the link between them has been broken. As above, we consider the extreme case in which the private trait is essential, such that individuals that do not produce it cannot reproduce. When we allowed the genetic architecture to evolve, we found that the potential for pleiotropy, where the traits could be linked, did not favour cooperation. Specifically, the production of public goods was only favoured in the same area as when there was no potential for pleiotropy (compare Fig 2A and 2B), which is when the indirect benefits outweighed the direct costs, and Hamilton’s rule was satisfied (rb̃ − c̃ > 0). In the other areas, where public good cooperation was not favoured by kin selection, mutation can create individuals for which the private trait and public goods production are not linked (mutation II in Fig 3). These individuals can then mutate to create a cheat that does not produce the public good but still performs the private trait. Selection favours these cheats, which can then invade and outcompete pleiotropic cooperators. Our result shows that in a simple case, if the underlying genetic architecture can evolve, then pleiotropy does not provide a mechanism to stabilise cooperation. Our aim was to capture the key feature of the ‘pleiotropy stabilises cooperation’ hypothesis, that private and public traits could be linked. Consequently, we assumed that the linkage between traits could either be gained or lost by mutation, which could be applied to a range of specific genetic architectures. More generally, our results illustrate how assuming fixed genetic associations between cooperation and other traits can lead to misleading conclusions. The problem is that assuming certain types of associations can force evolutionary outcomes that would otherwise be unlikely. This has been discussed previously in the context of traits such as punishment, rewarding, kin recognition, and green beards [45,49–52]. Furthermore, our conclusions are supported by a recent experiment in P. aeruginosa that showed the evolution of cheats that circumvent the expression of a cooperative trait but still produce a private trait [21]. We tested the robustness of our results by relaxing our assumption about how mutation can influence the genetic architecture (mutational accessibility) in three ways. In all three cases, we found essentially identical results, that the potential for pleiotropic links between traits did not stabilise cooperation. First, we ran simulations with more realistic mutation rates, in which losing a function was either 10 or 100 times more likely than gaining a function (S5 and S6 Figs). Second, we allowed the genetic architecture to evolve in a number of different ways, by either preventing pleiotropy to be lost, allowing pleiotropic cooperators to generate viable cheats, or allowing all mutations (section S1. 1 of S1 Text and S7 Fig). Third, we explicitly modelled the link between the cooperation and private traits by introducing a regulator for each trait and allowing the private trait’s regulator to become universal (pleiotropic) and hence regulate both traits at the same time (section S1. 2 and S8–S11 Figs). In contrast, Frénoy and colleagues [6] suggested that the potential for genetic links between private and cooperation traits can help populations resist the invasion of cheats. They found that where there was more pleiotropy between cooperative and metabolic traits, cooperation was lost more slowly and in some cases was maintained and that when allowing cooperation to evolve de novo, it evolved with links with metabolic traits. Our models differ in a number of ways, and therefore, we cannot say for sure why our conclusions differ. One possibility is that, because some of the pleiotropic interactions by Frénoy and colleagues [6] were due to gene overlap and/or operon sharing, the pleiotropic link could not be easily broken (personal communication, D. Misevic). The resistance of such pleiotropy to cheat invasion could have been influenced by a number of factors. For example, (1) the values chosen (analogous to a single combination of b and r) may have led to weak selection against cooperation and hence weak selection for unlinked genetic architecture; or (2) maybe longer simulations (>2,000 generations) or larger population sizes (>1,024 individuals) were required for the combination of mutation and selection to break up the links. Targeted simulations of different scenarios would allow the importance of these issues to be determined. Our above results suggest that if the genetic architecture can evolve, pleiotropy will not help stabilise cooperation (Fig 2). Instead, the private and public good traits have to be linked in such a way that the pleiotropy is unavoidable for pleiotropy to stabilise cooperation (Fig 1; S1 and S3 Figs). Considering the empirical examples from microbes, it is not clear why the traits would always have to be linked pleiotropically [1–6,46,47]. For example, the production of intracellular and extracellular factors in bacteria could evolve to be controlled by different circuits [21]. We suggest that unless it is explicitly shown otherwise, the null hypothesis should be that the genetic architecture can evolve, and hence, pleiotropy will not help stabilise cooperation. We hypothesise that causality could even be in the opposite direction, with cooperation favouring pleiotropy. We tested this by comparing the relative proportion of pleiotropic and nonpleiotropic cooperators in the area where Hamilton’s rule was satisfied. We examined the proportion of pleiotropic cooperators relative to nonpleiotropic cooperators—rather than relative to all nonpleiotropic genotypes (including cheats) —for the following reason. When Hamilton’s rule is satisfied, both pleiotropic and nonpleiotropic cooperators are fully favoured (Fig 2B). In this case, the null hypothesis is that they are equally common (i. e. , 0. 5). A proportion of pleiotropic cooperators relative to all cooperative genotypes above 0. 5 would show evidence that pleiotropy is indeed favoured. In support of our hypothesis, we found that, when Hamilton’s rule was satisfied, pleiotropic cooperators were relatively more common than nonpleiotropic cooperators, provided both the public good benefit (b) and relatedness (r) were high (Fig 4A and 4B). Consequently, pleiotropy is being favoured in the conditions under which cooperation is strongly selected for by kin selection (Fig 4A and 4B). We found similar results when the pleiotropic link was modelled more explicitly (S5 and S6 Figs). A similar point that genetic assortment (i. e. , relatedness) could favour pleiotropy was hypothesised by Frénoy and colleagues [6], but they did not test whether increasing assortment, and hence selection for cooperation, leads to more pleiotropy. In our model, cooperation favours pleiotropy because it reduces the mutational ‘cheat load’ locally (Fig 5). If the production of public goods is favoured by kin selection, then any nonproducers that arise by mutation will eventually be outcompeted [32,34]. However, nonproducers will be able to initially invade within a subpopulation [18,19,53–57]. This initial invasion will reduce the fitness of the cooperative lineage from which they evolved [27–29]. Pleiotropic linkage with a private trait prevents the nonproducers from invading and hence provides a fitness advantage to pleiotropic cooperative lineages, relative to cooperative lineages where production of the public good is not linked to a private trait. The key point here is that pleiotropy is only favoured if cooperation was otherwise favoured by a mechanism such as kin selection. Consistent with this cheat load hypothesis, an increase in the mutation rate makes pleiotropy more likely to be favoured (Fig 4A and 4B). To test whether nonpleiotropic cooperative lineages did suffer a cheat load, we ran additional simulations to examine how pleiotropy influenced growth in patches of cooperators. We compared the cheat load at the end of a single growth phase between conditions with and without mutations in patches started with 10 cooperators but where there were NPC pleiotropic cooperators and 10 − NPC nonpleiotropic cooperators. Without mutations, patches can grow maximally and reach maximum average fitness Wmax = 1 + g − c + b/n + (n − 1) b/n = 1 + g − c + b, as no cheat can emerge. With mutations, however, cheats emerge and impair their local patch growth. Cheat load was therefore measured as the difference between the maximum attainable average fitness in the absence of mutations and the observed average fitness with mutations W [28,30]. As predicted, patches with more pleiotropic cooperators at the beginning of the growth phase eventually suffered a lower cheat load and thus had a higher average fitness than patches with more nonpleiotropic cooperators (Fig 4C). The reason is that it takes more mutations to generate viable cheats from a genetic architecture in which private and cooperation traits are pleiotropically linked, as in [6]. In pleiotropic cooperators, the private and cooperation traits first have to be unlinked. This generates nonpleiotropic cooperators, which in turn need to undergo a second mutation to generate viable cheats (Fig 3). Our results emphasise the importance of distinguishing between short-term invasion and long-term evolutionary stability. Pleiotropy does not help stabilise cooperation over evolutionary time—cooperation is only favoured in the region where Hamilton’s rule is satisfied because of indirect fitness benefits (Fig 2). In contrast, pleiotropy does slow down the extent to which cheats will arise and invade in the short term (Fig 4C). The crucial point is that these cheats would not be successful in the long term—as soon as they disperse to a new patch, their fitness would be effectively zero. What matters for evolution is the long-term dynamics. Consequently, although pleiotropy protects cooperators against the short-term invasion of cheats, it does not influence whether cheats can successfully invade and outcompete cooperation on an evolutionary timescale. Another way of thinking about this is that pleiotropy provides an evolutionary advantage against nonpleiotropic cooperators but not against noncooperative cheats (Fig 5). The extent to which pleiotropy is favoured will depend upon a number of factors. Pleiotropic and nonpleiotropic cooperators are phenotypically similar, and the cheat load is generated from rare mutations. Consequently, the evolutionary forces favouring pleiotropic over nonpleiotropic cooperators are actually weak, explaining why pleiotropy does not go to fixation (Fig 4). We found similar results when mutations were 10 times more likely to lead to a loss rather than a gain of function. However, pleiotropy was favoured only with high relatedness and large cooperation benefits, even more so with a mutation bias of 100 (in which case both r and b need to be close to 1 for pleiotropy to be favoured with μ = 0. 005; S6 Fig). More generally, our argument is analogous to how pleiotropic mechanisms could help control the spreads of cancer, if cancer can be conceptualised of as a kind of cheat, within its multicellular host [58,59]. Another issue is that if private traits were not essential, then mutation could generate viable—albeit less fit—cheats that also do not produce private traits. Here, patches with pleiotropic cooperators would still suffer less from cheat load than patches with nonpleiotropic cooperators, where cheats that produce the private would be generated by mutations. In this case, we predict that the cheat load effect would still favour pleiotropy, though to a lesser extent than in the extreme case in which the private trait is essential. In contrast to how pleiotropy can favour any type of trait (Fig 1), our above argument that cooperation favours pleiotropy only works for cooperative traits. A privately beneficial trait would not favour pleiotropy, because the loss of that trait would incur an immediate fitness cost anyway. Also, mutants in either privately beneficial traits do not impose a growth cost locally, because there are no social interactions. There are no cheats if the two traits are private. In contrast, even though they are selected against in the long term, cheats gain a short-term advantage, which is costly to the cooperators they are invading—pleiotropy is favoured to remove this short-term advantage. In confirmation of this, when we examined a privately beneficial trait (b = 0 and c < 0) in our simulations, rather than a cooperative trait, we found that pleiotropy was not favoured (S12 Fig). We further hypothesised that the cheat load problem could be exacerbated in fluctuating environments, where cooperation alternates between being favoured and not favoured. The reason is as follows. If cooperation becomes unfavourable, cheating becomes strongly selected for. Then, local patches with more pleiotropic cooperators, which generate fewer cheats, would produce a higher number of cooperators and hence have a higher fitness when the conditions favouring cooperation returned. To test this idea, we ran simulations in which every 10 generations (i. e. , every growth phase), cooperation alternates between being beneficial (b > 0) and not beneficial (b = 0). We found that fluctuating benefits b decrease the region where cooperation is favoured (S13A and S13C Fig). This is because the benefit b = 0 half the time. As before, we found that cooperation was only favoured in the region where Hamilton’s rule was satisfied and hence favoured by kin selection (rb̃ − c̃ > 0; S13 Fig). Furthermore, and also as before, pleiotropic cooperators were increasingly common when benefits are large and r is high and hence when cooperation was otherwise favoured (S13B and S13D Fig). Comparing simulations with and without pleiotropy, we found that when pleiotropy was allowed to evolve, it decreased the extent to which cheats built up and hence reduced the ‘cheat load’ (S14 Fig). We found qualitatively similar results with fluctuating population structure (alternating between r = 0. 01 and r > 0. 01), although cooperation decreased to a lesser extent (S15 and S16 Figs). This further confirms that if cooperation is first favoured because of kin selection, pleiotropy does not stabilise cooperation after a subsequent decease in relatedness. Overall, these results in fluctuating environments provide further support to how cooperation could favour pleiotropy. In the previous sections, we made the restrictive assumption that mutations cannot occur in pleiotropically regulated genes. For example, when pleiotropy was acquired, no mutations could knock out either the private or public traits: mutants either lost pleiotropy or stopped expressing both genes. In experiments, however, mutants that lose the ability to express the public trait but keep expressing the essential private trait do exist [2]. Such mutations are expected to reduce the advantage of pleiotropy because they generate viable, private good–producing cheats. To test how mutations in pleiotropic genotypes influence the resistance of pleiotropy against cheat load, we ran simulations in which we allowed mutations at both the private and public good loci in pleiotropic individuals, thereby varying mutational accessibility. As before, we found that pleiotropy is favoured when cooperation provides large benefits and relatedness is high, especially when the mutation rate is also high (S17A and S17B Fig). Although these simulations led to a higher frequency of cheating mutants (compare Fig 4C and S17C Fig), patches with more pleiotropic cooperators still produced fewer cheating mutants (S17C Fig). This is because a fraction of the mutations in pleiotropic cooperators generate noncooperative private nonproducers that cannot profit from public goods and hence do not impair their local patch growth. Overall, these results confirm that cooperation promotes pleiotropy rather than the reverse, even under these relaxed conditions. More generally, we have taken a relatively heuristic approach, but there are many possible ways to model genome circuits that could lead to pleiotropy. Genetic details could influence the extent to which pleiotropy reduces the cheat load and hence the extent to which it is favoured, as well as the time required for mutations to break this pleiotropic link. For example, introducing a mutation bias in our simulations led to a considerably reduced advantage of pleiotropy against cheat load. Pleiotropy was favoured only with both very high relatedness and large cooperation benefits (S5 and S6 Figs). Consequently, as we find out more about specificities, it could be useful to model other genetic circuits.
Overall, there are at least four scenarios that could explain the observed instances of pleiotropy between a private and a cooperative trait [1–5]: (1) the association between the private and cooperative traits is a relative coincidence, with no adaptive significance; (2) both traits are favoured under the same conditions, and there is an efficiency benefit to having them coregulated; (3) pleiotropy stabilises cooperation; or (4) cooperation favours pleiotropy. The first hypothesis is our null model. The second hypothesis is analogous to a common explanation for why multiple traits are controlled by quorum sensing, with different traits being favoured at different stages of the growth cycle and at different population densities [25,60–62]. Our theoretical results show how the fourth hypothesis could work and that the third hypothesis requires extremely restrictive assumptions, making it unlikely to be generally important. The critical question for future work is how these different hypotheses could be distinguished empirically. The second hypothesis, that both private and cooperation traits are favoured under the same conditions and that there is an efficiency benefit to having them coregulated, does not require any particularly restrictive assumptions and fits with what we know about traits such as quorum sensing [25,60,61,63]. We therefore suggest that it is the most likely explanation for the observed instances of pleiotropy between a private and a cooperative trait in bacteria [1–5]. The third hypothesis, that pleiotropy stabilises cooperation, requires two patterns, neither of which have been supported by the empirical data. First, we have shown that pleiotropy only helps stabilise cooperation in conditions where cheats would otherwise invade, if there is some constraint that forces the personal and cooperative traits to be linked, such that they could not conceivably be regulated independently (i. e. , the link cannot be broken by mutation). Considering the empirical examples in bacteria, there seems to be no reason that traits such as cooperative extracellular proteases and private adenosine metabolism could not be regulated independently [3]. Cheats that stopped producing the cooperative trait by rewiring its control have actually been shown to emerge and invade cooperative populations [21]. In contrast, in the D. discoideum example, the cooperation and private traits appear to be controlled by the same gene, and so it is more plausible that they have to be linked [1]. The second pattern required to support the pleiotropy favours cooperation hypothesis is that that cooperation was not stabilised by some other factor, such as kin selection. In the slime mould D. discoideum, the average relatedness in fruiting bodies is r = 0. 98 [14], and both experimental evolution and genomic analyses also support a role of kin selection in favouring cooperation [16,64,65]. In the bacterium P. aeruginosa, we do not know relatedness in nature, but clonal growth in spatially structured populations is likely to lead to a high relatedness [11,48,66–68]. Consequently, there is no evidence that the examples of pleiotropy are in species in which cooperation is not explained by kin selection. In contrast, the fourth hypothesis, that cooperation promotes pleiotropy, makes the opposite predictions. It does not matter if the traits could be regulated independently, and pleiotropy is more likely to be observed in cases in which relatedness is high; hence, kin selection favours cooperation. These requirements are consistent with the empirical examples of pleiotropy in both slime moulds and bacteria. The hypotheses could also be distinguished by experimental evolution. If cooperation stabilises pleiotropy, then if selection for cooperation is removed, we would expect both pleiotropy and cooperation to be lost over time. Cooperation could be selected against by imposing low relatedness in cultures [18,54,69,70]. In contrast, if pleiotropy stabilises cooperation, then even under conditions of low relatedness, cooperation will be maintained. In addition, as we have discussed above, the hypothesis that cooperation favours pleiotropy does not necessarily require that the private trait is essential, only that it confers a sufficiently large fitness advantage. More generally, our results illustrate how selection on social traits, such as cooperation, can also shape the genetic architecture [6]. The genetic architecture is not fixed—like any aspect of an organism, it is subject to natural selection and can evolve [21,42,43,71,72]. We have shown that, especially in microbes, if cooperation is favoured, this can select for other traits to be pleiotropically linked with this cooperation. Furthermore, this is only one way in which social interactions could drive genome evolution. Other possibilities include mutualistic cooperation leading to genome degradation or gene transfers between species and selection for cheating leading to genome reduction, genome compartmentalisation, or the sequestering of cooperative traits onto mobile genetic elements [42,43,73–78].
We model an infinitely large population of haploid individuals. The population is subdivided into a large number of patches, each of size N. We assume that individuals interact socially within patches and that generations are nonoverlapping. At this point, we make no assumptions as to how social groups are formed except that individuals within a given patch can potentially be more related to each other than individuals chosen at random from the population. Individuals can produce, at personal cost c, a public good molecule that provides a benefit b/N to each individual on the patch (including the focal). Individuals can also produce a private good that is essential for reproduction: private-good nonproducers have their baseline fecundity reduced by an amount d. We later consider the case in which not producing the private trait leads to the individual’s death. We also assume that individuals produce a very large number of offspring, which all disperse from their natal patch to some new random patch, such that competition is global (no kin competition). Individuals carry two traits, denoted by p ∈ {0,1} and h ∈ {0,1}, coding for the production of the private and public good molecules, respectively. Assuming everyone’s fecundity is always strictly positive, and since competition is global, the relative fitness a focal individual i on patch j is wi, j=Fi, j/F-, where F- is the average fecundity in the population. We first determine when cooperation is favoured in a scenario in which only cooperators and defectors compete with each other. We assume here that both types express the private-good trait. Hence, the expected fecundity of a focal defector FD and that of a focal cooperator FC are given by FD=1+xD (N-1) bN (1) and FC=1-c+bN+xC (N-1) bN, (2) respectively, where xD and xC are the expected frequencies of cooperators among the coplayers of a focal defector and cooperator, respectively. Therefore, both types might potentially have a different social environment, i. e. , if xD ≠ xC. If groups are formed randomly, then xD = xC. Here, xC − xD is relatedness r [38] and is identical to the interpretation of relatedness as a regression coefficient of a partner genotype on the focal player’s genotype [36,37]. Because fecundity is linear in the number of cooperators (the game is of degree 1; [39]), relatedness r is the only necessary genetic association for describing evolutionary change in this game [38,39]. Thus, cooperators will be favoured whenever their relative fitness wC is greater than that of defectors wD, which is if (xC-xD) ︸r (N-1) bN︸b~+ (bN-c) ︸-c~>0. (3) Here, we consider the case in which cooperation is linked pleiotropically to a private trait that is essential for survival and reproduction. Hence, we assume that such private-good nonproducers suffer a fecundity cost d. We let pleiotropic cooperators compete with such private-good nonproducers. Therefore, the fecundity of private-good nonproducers is that of defectors (i. e. , Eq 1) reduced by an amount d, and the fecundity of pleiotropic cooperators remains similar as that of cooperators (Eq 2). Consequently, pleiotropic cooperators will be favoured over private-good nonproducers whenever (xC-xD) ︸r (N-1) bN︸b~+ (bN-c+d) ︸-c~>0. (4) We consider now the scenario in which not producing private goods leads to the individual’s death. Private-good nonproducers can no longer access public goods, and their share is distributed equally among the remaining pleiotropic cooperators. With these assumptions, the fecundity of private-good nonproducers becomes FPN = 0, and that of pleiotropic cooperators, FPC, becomes FPC=1-c+bN+ (N-1) bN =1-c+b (5) As we can see, the fitness of pleiotropic cooperators no longer depends on what others do. Everyone receives the full benefit b. Individuals will be neither more nor less likely than by chance to receive benefits from kin. Thus, the indirect fitness effect in Hamilton’s rule becomes b̃ = 0. Hence, pleiotropic cooperators will be favoured over private-good nonproducers whenever –c̃ > 0, where –c̃ = 1 − c + b. This condition boils down to whether expressing both the private and public traits results in a net gain or loss. However, for a population of pleiotropic cooperators to not go extinct, the average fecundity in the population should not be smaller than 1, which is whenever b ≥ c. We ran individual-based simulations to determine the validity of our analytical findings under a modified, more realistic life cycle in which growth can occur locally on a patch for a certain number of generations before a dispersal event occurs. Specifically, we consider a variant of the haystack model [79]. We model a finite population of haploid individuals subdivided into np = 100 patches. The life cycle is as follows: (1) Each patch is colonized by NF = 100 randomly selected founding individuals. Each founder produces randomly a very large number of juveniles (without mutation), which compete for space on the patch, leaving exactly N = 10 individuals on each patch (the others die). To vary relatedness on a continuous scale, we assume that one randomly selected individual among the NF founders will leave relatively more juveniles on the patch than the others (see below). (2) Individuals interact socially within patches. Social interactions influence their fecundity/growth rate (see above). (3) Unrestricted growth occurs within patches. All offspring survive, and parents die (nonoverlapping generations). (4) Steps 2 and 3 are repeated over k = 10 generations. This growth phase was implemented to simulate periods of local growth between migration events, typical of many microbial species. (v) Individuals disperse globally (every k generations), and exactly NF founders are randomly selected from the completely mixed global pool to colonise each patch. The remaining individuals die. Unless stated otherwise, we start our simulations with all genotypes in equal proportion. Each parameter combination is run over 105 generations and replicated 16 times. | Recent research into microbial communities has revealed that the cooperative secretion of molecules—which are produced by individual cells and benefit neighbouring cells—is linked to the production of privately beneficial intracellular enzymes. This pleiotropic link between commonly and privately beneficial traits has been suggested as a novel way for maintaining cooperation in conditions under which it would otherwise be outcompeted by cheating cells. The reason is that cheats, which do not cooperate, would also lose the benefit of producing the private trait and thus suffer a fitness disadvantage. We test the plausibility of this hypothesis with analytical models and individual-based simulations. We find that cooperation can only be stabilised if one makes restrictive assumptions about the genetic architecture, such that the pleiotropic link with a private trait cannot be broken through further evolution. If the genetic architecture can evolve, then natural selection can favour mutants that do not cooperate but that still perform the private trait, leading to the breakdown of cooperation. We discuss a number of alternative explanations for the observation of linkage between cooperative and private traits and show that causality may even arise in the opposite direction to that previously predicted—when cooperation is favoured, this may select for pleiotropy. Our results suggest a novel explanation for why cooperative and private traits may be linked and show how social evolution can shape the genetic architecture. | Abstract
Introduction
Results
Discussion
Methods | organismal evolution
gene regulation
microbiology
developmental biology
regulator genes
microbial evolution
gene types
population biology
fecundity
kin selection
gene expression
life cycles
evolutionary genetics
population metrics
heredity
genetics
biology and life sciences
evolutionary biology
genetic linkage
evolutionary processes | 2018 | Pleiotropy, cooperation, and the social evolution of genetic architecture | 10,407 | 289 |
Loss-of-function mutations in PINK1 or PARKIN are the most common causes of autosomal recessive Parkinson' s disease. Both gene products, the Ser/Thr kinase PINK1 and the E3 Ubiquitin ligase Parkin, functionally cooperate in a mitochondrial quality control pathway. Upon stress, PINK1 activates Parkin and enables its translocation to and ubiquitination of damaged mitochondria to facilitate their clearance from the cell. Though PINK1-dependent phosphorylation of Ser65 is an important initial step, the molecular mechanisms underlying the activation of Parkin' s enzymatic functions remain unclear. Using molecular modeling, we generated a complete structural model of human Parkin at all atom resolution. At steady state, the Ub ligase is maintained inactive in a closed, auto-inhibited conformation that results from intra-molecular interactions. Evidently, Parkin has to undergo major structural rearrangements in order to unleash its catalytic activity. As a spark, we have modeled PINK1-dependent Ser65 phosphorylation in silico and provide the first molecular dynamics simulation of Parkin conformations along a sequential unfolding pathway that could release its intertwined domains and enable its catalytic activity. We combined free (unbiased) molecular dynamics simulation, Monte Carlo algorithms, and minimal-biasing methods with cell-based high content imaging and biochemical assays. Phosphorylation of Ser65 results in widening of a newly defined cleft and dissociation of the regulatory N-terminal UBL domain. This motion propagates through further opening conformations that allow binding of an Ub-loaded E2 co-enzyme. Subsequent spatial reorientation of the catalytic centers of both enzymes might facilitate the transfer of the Ub moiety to charge Parkin. Our structure-function study provides the basis to elucidate regulatory mechanisms and activity of the neuroprotective Parkin. This may open up new avenues for the development of small molecule Parkin activators through targeted drug design.
Mutations in the PTEN-induced putative kinase 1 (PINK1) and PARKIN genes are the most common causes of autosomal recessive Parkinson' s disease (PD) [1]. Although the molecular mechanism underlying the pathogenesis of PD remain elusive, it has become clear that PINK1 and Parkin protein functionally cooperate in a novel mitochondrial quality control pathway [2]. Upon depolarization of the mitochondrial membrane, the Ser/Thr kinase PINK1 is stabilized on damaged organelles and plays a pivotal role for the activation and recruitment of the E3 Ubiquitin (Ub) ligase Parkin from the cytosol [3]–[6]. Parkin then labels numerous mitochondrial proteins with the small modifier protein Ub [7], [8]. Upon ubiquitination of mitochondria, adaptor proteins such as p97 and p62 are recruited to facilitate clustering of mitochondria around perinuclear regions and the selective degradation of substrates via the proteasome system and of whole organelles via autophagy (mitophagy) [3], [7], [9], [10]. Mutations in both genes, PINK1 and PARKIN, abrogate this presumably neuroprotective pathway through distinct molecular mechanisms and at different steps along the sequential process [3]–[6]. PINK1 has been demonstrated to phosphorylate Parkin at residue Ser65 in its N-terminal Ub-like (UBL) domain [11]–[13]. Moreover, it has been suggested that the activation of Parkin' s enzymatic functions and its mitochondrial translocation are coupled [11], [14], [15]. Very recently, PINK1 has also been shown to phosphorylate the modifier Ub itself at the same conserved Ser65 residue [16]–[19]. Both phosphorylation events appear to be required for full activation of Parkin' s enzymatic functions. Parkin had long been classified as a typical Really-Interesting-New-Gene (RING) -type E3 Ub ligase [20] that bridges the interaction between an E2 Ub-conjugating enzyme and a substrate. However, its Ub transfer mechanism has been challenged lately [21]. In fact, Parkin is a member of the RING-in-between-RING (IBR) -RING (RBR) family of E3 Ub ligases that mediate the transfer of Ub by a novel hybrid mechanism [22]. While Parkin binds the E2 co-enzyme with its RING1 domain (similar to RING ligases), it receives the Ub moiety from the E2 co-enzyme onto its active site (Cys431) in an unstable thioester intermediate (similar to Homologous-to-the-E6-AP-Carboxyl-Terminus (HECT) -type E3 ligases). Ub is then further transferred from Parkin onto a lysine residue of a substrate protein [23]. Consistent with its notoriously weak enzymatic activity, several partial crystal structures of Parkin [24]–[27] show a closed, auto-inhibited conformation. Several intra-molecular interactions between individual domains literally fold back Parkin onto itself. An at least 3-fold inhibition prevents charging of Parkin with Ub that is required for its activation and enzymatic functions [28]. It is evident that in order to gain enzymatic activity, Parkin must undergo major structural rearrangements. Ubiquitination enzymes indeed can perform particularly large conformational changes during their catalytic cycles including the remodeling of domain interfaces [29]. Moreover, phosphorylation-dependent exposure of the RING domain [30], [31] or relief of the auto-inhibited structures has been shown for RING and HECT E3 ligases respectively [32]. Here, we set out to provide structural models for human Parkin that would allow release of its auto-inhibited conformation and consequently activation of its E3 Ub ligase functions. We have applied highly accurate molecular modeling methods to provide for the first time an all-atom resolution of human full-length Parkin. Given the suggested auto-regulatory role of the UBL domain [33] and the importance of PINK1 kinase activity for Parkin activation [11]–[13], we have performed molecular dynamics simulations (MDS) to study conformational changes that might be induced by phosphorylation of Ser65. Strikingly, our simulations and calculations of pSer65 in silico predict a structural rearrangement of the UBL domain that initiates a sequential release of Parkin' s intra-molecular interactions. Along these opening conformations, we have docked the E2-Ub complex, required for Parkin' s activation and enzymatic functions. Importantly, the presented computational predictions are consistent with our cell biological studies. We provide the basis for a better understanding of the molecular mechanisms of auto-inhibition and liberation of Parkin' s catalytic activity.
In order to understand the activation mechanisms of the neuroprotective E3 Ub ligase Parkin (for a schematic view of Parkin and its activation see Figure 1A), i. e. its opening conformations and the release of enzymatic function (s), we performed molecular modeling and dynamics simulations. Our models were generated using the recently resolved X-ray structures of human N-terminal truncated Parkin (PDB IDs: 4BM9 [27] and 4I1H [24]) and of its rat homolog (PDB IDs: 4K7D and 4K95 [26]). The molecular modeling methods have been described previously [34]–[36] and the procedure is outlined in the method section in detail. The generated model for the first time allows a view of the full-length human Parkin protein at an all atom resolution (Figure 1B and Movie S1). Most importantly, it provides the structure of the regulatory N-terminus but also closes smaller gaps across Parkin' s entire length. The UBL domain (residues 1–76) and in particular a flexible linker (residues 77–140) had not been structurally resolved so far for human Parkin. The linker is comprised of two sub-domains: (1) a semi-globular domain from residues 77 to 125 that appears highly dynamic and (2) a tethering loop region from residues 126 to 140 that connects to the RING0 domain. A further examination of the linker' s secondary structural features gives the following: α-helical regions from residues Gly77 to Gly85, Arg89 to Ser108, a helix loop turn from Leu112 to Ser116, and minor beta-roll (anti parallel) from Val117 to Leu123 with the beta-strands from the UBL domain residues Phe4-Asn8 and Leu41-Phe45. From residue Leu123 onward, the linker is random coil through Arg140. The newly modeled N-terminus consequently alters the position of the IBR relative to the UBL domain and RING1. The superposition overlay deviates slightly from the X-ray structures (Figure 1B), however, root mean square (RMS) measurements between the backbone of the improved model and various collected structure remains under 5 Å for residues Ser145 to Val465. The RING domains contained within are rather rigid due to zinc-finger stabilization, but the UBL domain and the adjacent linker are both flexible and capable of large movement with free MDS (Movie S2). Our complete model of human Parkin plus observations for its dynamic motion indicates that the N-terminal UBL-linker region acts like a spring/clamp that tightly holds Parkin in its closed, auto-inhibited conformation. Noteworthy, a prime role for the N-terminus in negatively regulating Parkin' s enzymatic activity had already been established [33]. Consistently, PINK1-dependent phosphorylation of the UBL-located Ser65 appears to play a regulatory role for activation and recruitment of Parkin to depolarized mitochondria [11], [14], [15]. Given this functional link and the particular N-terminal flexibility, we modeled PINK1-dependent phosphorylation of Parkin' s UBL domain at Ser65 in silico to investigate potential opening conformations (Figure 1C). In our Parkin model, Ser65 is buried within a pocket that is formed between the flexible linker region and the UBL domain (Figure 2A/B). We found that phosphorylation of Ser65 (pSer65) resulted in conformational changes compared to unmodified Parkin in the static model (Figure S1) as well as in more flexibility during free MDS (Movie S3). A comparison between the pSer65 Parkin model with existing X-ray structures is given in Figure S2. Locally, pSer65 induced a widening of the surrounding cleft and an increase of water molecules occupying the pocket (Figure 2). We used two center-of-mass metrics to determine the cleft widening. Calculations were carried out first on the edges of the cleft on each side (Figure 2A/C, for orientation see Figure 2B). Ser65 showed a rather narrow gap of about only 7–9 Å, however pSer65 induced a widening of the cleft>12 Å. Figure 2D shows the wide range in flexibility between the linker and the UBL domain as the MDS proceeds. Simulations of greater than 100 ns for Parkin Ser65 and pSer65 gave approximately 8 Å or 14 Å, respectively. To follow up on these significant changes, we studied mutations of Ser65. In functional studies, Serine is often substituted with Alanine for a phospho-dead version (S65A), while mutations to the negatively charged Aspartic (S65D) or Glutamic (S65E) acid are used as phospho-mimic variants. For critical analysis of the intermolecular interactions induced by these mutations (Figure S3), we computed changes in binding energy (ΔΔG), using the Zone Equilibration of Mutants (ZEMu) [37] method, implemented in the MacroMoleculeBuilder (MMB) [38]. However, we could not detect any significant change among the Ser65 substitutions (Figure S3). We next modeled the Ser65 mutations to measure gap distances during free MDS using two correlative center-of-mass sets (Figure 2D/E). Both S65D and S65E showed an initial gap distance of about 10 Å. The cleft further widened quickly over the course of 20 ns and longer using MDS, reaching distances of around 12–13 Å similar to pSer65 (Figure 2E). The phospho-dead S65A variant displayed an initial narrower gap, rather comparable to unmodified Ser65, but showed a pronounced gap opening during simulation of 20 ns and longer (Figure 2E) reaching widths closer to pSer65 and both phospho-mimic mutants. Moreover, we found that hydration of the pocket interfacing UBL domain and linker provokes the opening of Parkin (Movie S4). Based on these findings, we measured the Solvent-Accessible-Surface-Area (SASA-Å2) within the pocket and counted the number of water molecules at common time intervals (from the simulation). The average number of water molecules within the pocket ranges from 9 to 12 for the initial time point (t = 0), 6 to 12 for t = 5 ns, and 12 to 19 for t = 100 ns (Figure 2F). Ser65 allowed the pocket to be filled with a constant number of approximately 12 water molecules during the entire simulation, while pSer65 induced a rapid increase from 12 to 19 water molecules that were maintained over time. In the case of S65A, the pocket initially collapsed from 10 to 6 and then slowly re-hydrated to over 16 water molecules. Both phospho-mimic mutants induced solvation of the pocket surrounding residue 65, more similar to pSer65 Altogether, we provide structural evidence that phosphorylation or mutations of Ser65 perturbs the stability between the UBL domain and the linker region, such that the structure becomes looser more quickly allowing the enclosing cleft to widen. To corroborate our structural predictions how mutations affect opening of the cleft that encloses residue 65, we expressed GFP-Parkin wild type, S65A, S65D, and S65E variants in human HeLa cells. As a functional readout, we used an unbiased high content imaging assay that monitors mitochondrial translocation of Parkin in cells over time [39]. The paradigm relies on chemical uncoupling of mitochondria by treatment with carbonylcyanide m-chlorophenylhydrazone (CCCP), which induces PINK1-dependent phosphorylation of Parkin at Ser65 in the UBL domain, Ub charging of Cys431 and Parkin' s recruitment to mitochondria. Translocation of Parkin is quantified as the ratio of cytoplasmic to nuclear GFP (Parkin) signal for each of the fusion proteins (for an example, see Figure 3A). Un-transfected and low level expressing cells were excluded from the analysis, ensuring comparable GFP intensities across all Parkin variants (Figure S4A). Under basal conditions, Parkin wild type and residue 65 mutants were evenly distributed throughout the cell (GFP ratio ∼1). Upon mitochondrial depolarization (2 h CCCP), the average GFP ratio for Parkin wild type strongly increased, indicating its co-localization with mitochondria (Figure 3B). The catalytically inactive Parkin mutant C431S did not translocate to mitochondria, consistent with previous reports that enzymatic activity and mitochondrial translocation of Parkin are linked [11], [14], [15]. In contrast, Parkin S65A, S65D, or S65E did not abrogate translocation, but showed a clear delay in recruitment at both time points analyzed. We defined translocation positive cells as a GFP ratio of>1. 8, which corresponds to the average value (∼40% of the cells) for Parkin wild type at 2 h CCCP (Figure 3C–E). All Parkin mutations showed a significantly reduced percentage at 2 h CCCP treatment with no significant difference between Ser65 substitutions and the control mutant C431S. Ser65 mutants remained less translocation-positive at 4 h CCCP compared to wild type, but showed significantly more translocation compared to C431S. Representative merged immunofluorescence images are given in Figure 3F (for individual channels, see Figure S4B–D). The delay in Parkin activation and translocation caused by mutations of residue 65 is further indicated by a reduced co-recruitment of the Ub adaptor protein p62 (Figure S4C–D). Upon Parkin-dependent ubiquitinations of mitochondrial substrate proteins, p62 promptly accumulates on mitochondria in order to facilitate organelle clustering and degradation [3], [9]. Noteworthy and consistent with our structural analysis of Ser65 mutations, phospho-dead and phospho-mimic substitutions equally delayed Parkin translocation to mitochondria. Though S65A cannot be phosphorylated by PINK1, it is capable of opening the cleft between UBL domain and linker region, similar to pSer65 or phospho-mimic mutants. This is in agreement with previous reports that phosphorylation of Ser65 plays an important role, but is not absolutely required for the translocation of Parkin [13], [16]. To model putative opening structures, Parkin models were analyzed using both traditional MDS and enhanced sampling MDS techniques to generate a large pool of conformers. We used Targeted Molecular Dynamics or Maxwell' s demon Molecular Dynamics (MdMD) [35], [36], plus mixed torsional sampling with large-scale, low-mode sampling and Monte Carlo Molecular mechanics (e. g. LC-MOD/MCMM) within Schrödinger [40]–[43]. We found several unique conformations that captured the N-terminal flexibility, including some large-scale rearrangements in the positioning of the UBL domain (Movie S3). Retaining only the lowest energy conformers generated from a large ensemble of states, we applied Polak-Ribiere conjugate gradient for structural minimization to obtain the optimal internal coordinates to each structures to arrive at final conformer snapshots. We then dubbed evenly spaced conformers as guideposts spanning the conformational extremes for the UBL domain starting adjacent to the IBR and ending in close proximity to the active site. Using MdMD minimal biasing method, we tested whether or not the Parkin structures could move between the conformers generated with global variable (accessor) based on root mean square deviation (RMSD) to the backbone C-alpha atoms. This revealed a smooth pathway for the movement of the UBL domain (Figure 4 and Movie S5). At the beginning, Parkin showed only minor rearrangements of the UBL domain (Figure 4A), while the movement of the UBL and of the flexible linker is evident at longer times of MDS (Figure 4B–E). We repeated this simulation using a randomizer in our molecular dynamics sprint interval that ensures non-duplicative runs, and show the superposition of replicates alongside (Figure 4A′–E′ and Movie S6). Using the smooth transitions, we have generated>32,000 conformers to study the opening of Parkin. As a consequence of the flexibility in the linker region, we found an array of domain reorientations that connect in a logical fashion the movement of the UBL domain from its initial configuration to a set of states near the active site (Movie S5). We tested each structure generated with YASARA' s What-If and Procheck for backbone dihedrals, rotamers, and packing that support the structures stability during simulations [44]–[46]. The average plot for Phi-Psi space is shown in the Ramachandran plot with Z-axis for dihedral count (Figure S5). The initial MdMD was not applied during the first 1–10 ns, however upon engaging the algorithm, the linearity of the MdMD algorithm measured by RMSD between the replicate and guidepost structure is given (Figure S6). We measured the C-terminal RMSD for residues 145–465 finding three replicates within 2. 75 Å of the initial production run. We used an RMS global variable within MdMD to determine a series of conformers related to UBL motion (Figure S7). In contrast to the zinc-finger stabilized C-terminal core (residues Ser141-Val465), the N-terminal region (residues Met1-Arg140) showed a rapidly growing root mean square fluctuation (RMSF) (Figure S8) and an increase in RMSD (Figure S9) in the MdMD replicates. During the trajectory of the UBL domain across and alongside Parkin, several structural changes were noted that are potentially relevant for E2 co-enzyme binding and Ub charging of Parkin. Following phosphorylation of Ser65 as a trigger, we found that residues within the linker region undergo repeated contacts with the RING1 and RING2 domains during movement of the UBL domain (Movie S5). Our data indicates that several safety belts must be released in order to unleash its E3 ligase activity. First, the UBL-linker region must dissociate from RING1 and IBR domains in order to loosen the entire structure (Figure 5A). Based on our simulations, we measured the release of the inhibitory N-terminus that acts like a spring/clamp. The distance between the UBL and RING1 domains indeed increased from about 20 Å to more than 50 Å over time MDS (Figure 5B). Similarly, the distance between the UBL domain and the IBR region significantly increased from an initial 30 Å to almost 90 Å (Figure 5C). Second, the repressor element of Parkin (REP), a region between IBR and RING2, blocks access of an Ub-loaded E2 co-enzyme (E2∼Ub; the tilde symbol is used to indicate a thioester bond) in RING1 (Figure 5D). This inhibitory interaction is loosened during Parkin' s transitions from stage 1 (Figure 4A) to stages 2/3 (see Figure 4B/C). To measure this release, we calculated RMSD scores. The REP residues considered were all within 5 Å of the supposed RING1 binding region (defined as Cys238, Thr240, Cys241, and Cys263). MdMD showed a steady increase of the RMSD score from 0 Å to>5 Å over the course of 50 ns sampling (Figure 5E). We then measured the distance of the REP element to the E2 binding site in RING1. The bond distance from Tyr391 (REP region) to Cys238 (E2 binding site in RING1) changed over time from under 10 Å to ∼20 Å (Figure 5F). This might allow access of an incoming E2∼Ub complex to the binding site. Interestingly, during continued UBL-linker movement towards the active site (as seen in Figure 4), Tyr391 is eventually able to reposition back into the E2 binding site, possibly indicating a reset mechanism for the next binding event of a (re-) charged E2 enzyme. Third, the RING0 domain buries Parkin' s active site Cys431, making it unavailable to receive Ub from an incoming E2 (Figure 5G). During the UBL-linker movement, RMSD measurements indicate that the RING0 to Cys431 slightly increased (∼3 Å) (Figure 5H), while the distance itself initially increased during the first 10 ns from 15. 8 Å to ∼19 Å and then decreased to under 17 Å (data not shown). SASA calculations indicated that Cys431 initially lost virtually all water interaction surface (first 5 ns), but abruptly began to hydrate thereafter (Figure 5I). This coincides with the binding of the E2 co-enzyme on the other side of Parkin at RING1 that had been blocked by the REP region before. For the remainder of the measurements, Cys431 stayed at an average SASA of 12. 5 Å2 with transient peaks of 20 Å2. This increase in solvent exposure over time is indicative of a conformational reorientation, which could allow the active site Cys431 to receive a thioester-bonded Ub moiety from the E2 enzyme. Taken together, our MDS and subsequent calculations revealed a sequential release of Parkin' s safety belts preventing its activation. The dissociation of the inhibitory N-terminus is triggered by PINK1-dependent phosphorylation of Ser65 in the UBL domain. As a consequence, Parkin' s entire structure is loosened and further perpetuates the liberation of Parkin' s presumed E2 binding region and of its active center as pre-requisite for enzymatic activity. To identify the E2 enzyme binding site in RING1 and to dock an E2∼Ub complex, we scanned across an evenly spaced distribution of ∼50 structures spanning the opening of Parkin using Piper for protein-protein docking [43]. We used the structure of an Ub-loaded E2 enzyme UbcH5a/UBE2D1 (PDB code: 4AP4) that shows an isopeptide amide linkage between the mutant active site of the E2 (Cys85Lys) and Gly76 of Ub (UbcH5-Ub complex) [47]. Members of the UbcH5 family have been shown to charge Parkin with Ub and act as co-enzymes during mitophagy [14], [21], [39], [48]. Each Parkin structure was allowed an attempt to dock with the UbcH5a-Ub complex retaining the best ten conformers from each pairing. The resulting pool was then filtered for lowest energy structures using Schrödinger' s Bioluminate/Piper docking and Molecular Mechanics-Generalized Born Surface Area evaluation [43], [49]–[51]. The optimal bound state of Parkin with the E2 enzyme is shown in Figure 6A. In this structure (state 2/3, see Figure 4B/C), the REP region is pushed back to better reveal residues in RING1 critical for E2 binding. The top performing structures were further studied with unbiased (free) MDS. Following docking of the UbcH5a-Ub complex with Parkin, we completed simulations where the Ub-loaded E2 makes substantial movements toward the active site region of Parkin (Figure 6B–D and Movie S7). The distance from residue Cys431 to the C-terminal Ub residue Gly76 ranges from an initial 60 Å (in the closed conformation) to approximately 30 Å after 200 ns of accelerated MDS [52] using default parameters within the NAnoscale Molecular Dynamics engine [53]. The optimal binding pair kept that distance within 40 Å. It is interesting to note, that the E2-Ub complex rolls around Parkin, thereby moving the Gly76 of the bound Ub moiety into a better position for Parkin' s catalytic center. While the UbcH5a-Ub complex moves across Parkin, it maintains a final average distance of approximately 32 Å from the thiol of Parkin' s Cys431 to Gly76 of Ub. However, without release of the Ub moiety from the E2 (due to the amide linkage in this structure) the co-enzyme stalled in the vicinity of Parkin' s active site, while at the same time the UBL domain moved away from the midpoint configuration (state 3, see Figure 4C) into a new conformation unobserved in MdMD (Figure 6A–C, Movie S7). As a control, we started a simulation with the sub-optimal binding for Ub-loaded E2 and Parkin (higher energy). In this case, we found that the docked UbcH5a-Ub complex tended to stall and gradually drifted from the active site toward the IBR domain, increasing the distance between Ub-Gly76 and Parkin-Cys431 (>65 Å) (Movie S8). In summary, our MDS provide the basis to study major domain rearrangements within Parkin as well as to investigate binding of E2 co-enzymes and Ub charging of Parkin as part of its activation process. The labile nature of the E2∼Ub thioester makes the structural and functional studies of these complexes very challenging. For Parkin, a thioester-bound Ub could not be identified so far [21]. A substitution of the active site cysteine in E1, E2, or E3 ubiquitinations enzymes with a serine residue, results in the formation of a relatively stable oxyester bond to Gly76 of Ub. To confirm the suitability of a C431S substitution as a tool to monitor the E2-dependent Ub charging of Parkin, we used MMB and ZEMu. In contrast to many Zn2+ coordinating cysteines, the catalytic residue Cys431 is not making any intra-molecular interaction (Figure S10). For the C431S substitution we obtained similar ΔΔGs from three Parkin crystals that indicate no major change, corroborating the usefulness of this particular inactive variant. To investigate effects of Ser65 mutations on Parkin C431S-Ub oxyester formation as a surrogate for its activation, we expressed these Parkin variants carrying an additional C431S mutation (Figure 7A). As expected, CCCP treatment of cells expressing the single C431S (S65) mutation induced the formation of an 8 kD shifted Parkin band. The specificity of this band was determined by NaOH treatment that chemically cleaves the C431S-Ub oxyester bond. This is consistent with Parkin' s auto-inhibited, inactive structure before and its activation upon mitochondrial depolarization. Of note, for both phospho-mimic mutations (in particular S65E and to a lesser extent S65D), an 8 kD shifted band was detected even at basal conditions (i. e. without CCCP treatment). Consistently, S65E (and S65D) showed enhanced levels of Ub-charging at early time points (1 h CCCP) while oxyester formation of S65 Parkin became apparent only after 2 h and strongly increased by 4 h of CCCP treatment. In contrast, the phospho-dead S65A mutant showed no discernable Ub-oxyester at 0 h CCCP and compared to Ser65 strongly reduced levels after longer incubation times with CCCP (16 h). These results are consistent with a slight activation of the phospho-mimic Parkin mutants at steady state and a strongly diminished Ub-charging of the phospho-dead variant S65A. To provide further evidence that Ser65 phosphorylation plays a role for Parkin' s enzymatic function, we performed immunoprecipitation coupled to an in vitro ubiquitination assay. We transfected HEK293E cells with FLAG-Parkin and affinity purified Parkin by anti-FLAG from cells that have been treated with CCCP for 1 h or left untreated. Given the importance of PINK1 kinase activity and phosphorylation of Parkin' s Ser65 [11]–[13], we treated some samples with phosphatase to observe its E3 ligase activity with or without this activating posttranslational modification. Immunoprecipitated Parkin was incubated with a complete mix of ATP, recombinant Ub and E1 enzyme as well as different E2 enzymes. While incubation with UbcH5b/UBE2D2 or UbcH7/UBE2L3 resulted in the formation of likely mono- and di-ubiquitinated Parkin species, Ubc13/UBE2N together with its co-factor Uev1a was unable to generate these Ub modifications on Parkin (Figure 7B). Parkin, without any E2 enzymes, was not able to Ub modify itself. No discernable differences in Parkin' s (auto-) ubiquitination were observed between samples that had been left untreated or had been treated with CCCP or phosphatase. CCCP treatment resulted in enhanced poly-Ub levels as judged by streptavidin detection of the biotinylated Ub used. Phosphatase treatment strongly reduced the enzymatic activities of Parkin. It is unclear whether these ubiquitinations were formed on Parkin itself (but are not detectable with a anti-FLAG antibody) or on E2 enzymes or were attached to other co-immunoprecipitated proteins. Taken together, our structural and functional data corroborate an important role of Ser65 phosphorylation and provide a mechanistic insights into unleashing Parkin through unfolding, E2 enzyme binding and Ub charging as well as activation of its enzymatic E3 ligase functions and concomitant recruitment to damaged mitochondria.
Based on several recent but partial X-ray crystals [24], [26], [27] we performed molecular modeling to provide an all atom resolution structure of human Parkin. This neuroprotective RBR-type E3 Ub ligase is inactivated in familial and probably also sporadic forms of PD [1], [54]. Under basal conditions, Parkin is auto-inhibited through several intra-molecular interactions. Our models suggest that the N-terminal UBL domain and the linker region, which has been crystallographically difficult to resolve, act as a spring/clamp that holds Parkin in its closed conformation. Previous studies had shown that upon mitochondrial stress, Parkin' s activation and recruitment is dependent on the upstream kinase PINK1. We describe the first comprehensive molecular dynamics study of Parkin activation upon PINK1-dependent phosphorylation of Ser65 in the UBL domain. We have generated over 30,000 snapshot structures of Parkin that illustrate a sequential release of its intertwined domains along an unfolding pathway to liberate its Ub ligase function (s). Functionally, PINK1-dependent phosphorylation of Parkin' s Ser65 appears as the most upstream event of its activation cascade. We have analyzed a phospho-dead S65A mutation of Parkin as well as the phospho-mimic variants S65D and S65E. Using a high content imaging approach, we found a significant delay in Parkin recruitment upon mitochondrial uncoupling for all Ser65 substitutions, in line with previous studies [13], [55]. In contrast to an earlier report [11], [15], we noted a substantial increase in Ub charging (oxyester formation) of particularly S65E Parkin and to a lesser extent of the S65D variant even in the absence of CCCP. Our findings are in agreement with more recent studies that support the idea of a partially released auto-inhibition for phospho-mimic Parkin mutants [55], [56]. In fact, phospho-mimic Parkin, compared to wild type, showed increased Ub ligase activity at steady state in neuronal cells [55] and Drosophila [56] as evidenced by reduced levels of Parkin substrates under physiological conditions. However, phospho-dead Parkin showed greatly diminished Ub charging at all conditions, consistent with a reduced E2 discharge and E3 ligase activity in vitro [57] and in vivo [56]. Despite the enhanced enzyme activity at base line, neither S65D nor S65E Parkin could be found on mitochondria without depolarization. Although mitochondrial translocation and Ub charging of Parkin appear as interdependent events [11], [14], [15], our data corroborate the hypothesis of a second mechanism that is required for the translocation of Parkin, consistent with the finding that Parkin phosphorylation is not sufficient to trigger its recruitment. Thus, PINK1 phosphorylation of Parkin appears to primarily boost its enzymatic activity, thereby regulating not only mitochondrial function but also activity and survival of dopaminergic neurons [56]. In order to bridge the gap between static structures and enzymatic functions, we performed MDS and used in addition to normal modes, Monte Carlo algorithms and minimal-biasing methods that provided consequent opening conformations of Parkin. Although some of the presented models are hypothetical in nature, it is certain that the N-terminal region has to dissociate from the remaining part of Parkin in order to facilitate further required structural rearrangements. Our full-length Parkin model shows that Ser65 is located within a newly defined cleft that is formed between the UBL domain and the adjacent linker region. Using free MDS, we demonstrate that phosphorylation of Ser65 results in widening and enhanced solvation of this pocket. Similarly, both phospho-mimic and albeit slower also phospho-dead Ser65 mutations allow opening of the cavity as opposed to unmodified Ser65. Our MDS studies further suggest that phosphorylation of Ser65 in Parkin initializes the dissociation of the UBL domain (within ns) and thereby induces larger scale conformational motions over time. It should be noted that the simulation time scale is not an exact match for what is actually occurring in the cell, but is idealized. In the cell, Parkin protein will require more than the calculated time to undergo major structural rearrangements and binding of the E2 co-enzyme in order to become fully active. pSer65-dependent hydration of the surrounding cavity induces the release of the UBL domain, which appears to act as the first safety belt that keeps Parkin' s activity in check under basal conditions. Upon dissociation of the UBL domain, we next detected release of the inhibitory REP region from the E2 binding site in RING1 that could facilitate association of an Ub-charged co-enzyme. As a further consequence we also observed conformational adjustments in RING0 that intervenes between RING1 and RING2 burying Parkin' s active site when unmodified. In its inactive state, the E2 binding site and Parkin' s catalytic center are separated by a distance of more than 50 Å, which would not allow for Ub transfer. However, during opening of Parkin, we measured significant changes in distances, RMSD and SASA values, indicating differences in positioning and solvation of the region surrounding Cys431 that would be important for charging of Parkin. These promising measurements prompted us to dock an Ub-charged E2 enzyme (UbcH5a/UBE2D1) during the opening phase of Parkin. Indeed, we found conformations that were able to accommodate the co-enzyme and allowed us to generate an energetically favorable Parkin: E2-Ub complex. Strikingly, subsequent free MDS, showed a reorientation of the E2 to better position the Ub-Gly76 residue towards Parkin' s active site. This coincides with hydration around Cys431 and results in a significantly reduced distance between the catalytic centers of the E2 and Parkin. The amide linkage in the E2-Ub co-crystal prevented further insights into the transfer of the Ub moiety from the E2 enzyme onto Parkin' s active center, but we noted a repositioning of the REP region back into the E2 binding site in RING1. This putative reset mechanism may allow dissociation of the discharged E2 for the next binding event during consecutive ubiquitinations rounds. During mitophagy, multiple E2 enzymes are utilized by Parkin for Ub charging, mitochondrial translocation and substrate ubiquitinations [14], [39], [48], [57], [58]. We have recently demonstrated that some regulate Parkin' s activation and mitophagy through redundant, cooperative, or antagonistic mechanisms [39]. E2 enzymes usually confer the linkage specificity for RING-type Ub ligases while HECT- and RBR-type E3 ligases are charged with Ub and themselves define the respective Ub linkages that are formed. Parkin appears to catalyze the conjugation of various Ub modifications ranging from (multi-) mono-ubiquitination to poly-Ub chains with different topologies [59], while particularly K27, K48, and K63 linked chains have been observed during mitophagy [3], [7], [55]. In the absence of Parkin co-crystals with co-enzymes and/or substrates, it will be important to dock additional E2 co-enzymes, other co-factors as well as substrates, where structures are available to shed more light onto the catalytic activity (ies) of Parkin. In this context, it is important to note that Parkin has been suggested to self-associate through a PINK1-dependent mechanism upstream of its mitochondrial translocation [14]. Our own findings (unpublished) and other recent studies [19], [56] including the crystal structures (PDB IDs: 4K7D and 4K95) [26] support the dimerization/multimerization capability of Parkin. A small fraction of PINK1-phosphorylated (i. e. activated) Parkin could activate non-phosphorylated Parkin in trans and thereby amplify its E3 ligase activity through an autocatalytic mechanism [19], [56]. The inactive C431S Parkin variant would inhibit this feed-forward loop and consistently is unable to translocate to mitochondria [11], [14], [15], [39] suggesting that the Ub moiety must be passed onto a lysine residue of a substrate that may well include Parkin itself. Of note, the delay in S65A translocation to mitochondria was rescued in the presence of wild type (i. e. pSer65) Parkin [19]. However, S65A Parkin could not be charged by wild type in the C431S-Ub oxyester experiment [56]. Accumulating evidence indicates that pSer65 primarily boosts Parkin' s enzymatic E3 ligase activity, but also suggests additional functions for the phosphorylation of the UBL domain than release of auto-inhibition [57]. A deletion mutant of Parkin lacking the UBL domain was able to translocate to mitochondria [3], [8], but could not ubiquitinate the model substrate Miro1 while retaining robust auto-ubiquitination and E2 discharge comparable to wild type Parkin [57]. A role of the UBL domain in binding to substrates and regulators has been described [60]. Strikingly, it has been shown that PINK1 also phosphorylates the modifier Ub itself at the conserved Ser65 residue [16]–[19], in addition to Parkin' s UBL domain. Of note, Parkin phospho-Ub alone (i. e without CCCP treatment) can activate Parkin wild type, ΔUBL as well as the S65A mutant in vitro [16], [17]. However, optimal activation of Parkin appears to depend on both Ser65 phosphorylation events catalyzed by PINK1 [17]. Consistent with a putative phospho-binding site in Parkin' s RING0 that has been identified through co-crystallization of a sulphate ion [27], Parkin can bind to phospho-mimic Ub which seemed to be dependent on Parkin pSer65 [18]. One might speculate that while phosphorylation of UBL domain could (pre-) activate Parkin' s E3 ligase functions, phosphorylation of Ub that is already attached to a mitochondrial substrate might induce its translocation and full enzymatic activity. Binding of Parkin to phosphorylated Ub moieties (as free or attached monomers or poly-Ub chains) or even its own phosphorylated UBL domain [16]–[19] may also help to maintain an open, active conformation. In summary, our structural and functional data underscore the importance of PINK1-dependent phosphorylation of the UBL domain for the activation of Parkin' s enzymatic functions. In addition to functional studies that are required to entirely elucidate the mechanisms and the sequence of events during activation and translocation of Parkin, molecular dynamics simulations will certainly be useful to provide structural insights. Our models highlights multiple opportunities for analysis of PD mutations and modifications to hopefully open up new avenues for the design of safe small molecule activators of this multipurpose neuroprotective E3 Ub ligase.
The protein sequence of the human E3 Ub ligase Parkin (isoform 1) (Parkin), encoded by the PARK2 gene, was taken from the NCBI Reference Sequence: NP_004553. 2. The following 465 amino acid residues (full-length) were used for modeling: MIVFVRFNSSHGFPVEVDSDTSIFQLKEVVAKRQGVPADQLRVIFAGKELRNDWTVQNCDLDQQSIVHIVQRPWRKGQEMNATGGDDPRNAAGGCEREPQSLTRVDLSSSVLPGDSVGLAVILHTDSRKDSPPAGSPAGRSIYNSFYVYCKGPCQRVQPGKLRVQCSTCRQATLTLTQGPSCWDDVLIPNRMSGECQSPHCPGTSAEFFFKCGAHPTSDKETSVALHLIATNSRNITCITCTDVRSPVLVFQCNSRHVICLDCFHLYCVTRLNDRQFVHDPQLGYSLPCVAGCPNSLIKELHHFRILGEEQYNRYQQYGAEECVLQMGGVLCPRPGCGAGLLPEPDQRKVTCEGGNGLGCGFAFCRECKEAYHEGECSAVFEASGTTTQAYRVDERAAEQARWEAASKETIKKTTKPCPRCHVPVEKNGGCMHMKCPQPQCRLEWCWNCGCEWNRVCMGDHWFDV. Parkin has several conserved domains to be modeled (from the N-terminus to the C-terminus): UBL (residues 1–76), undefined linker region (residues 77–140), RING0 (or UPD) domain (residues 141–216), RING1 domain (residues 217–328), IBR domain (residues 329–378), REP region (residues 379–410), and RING2 domain (residues 411–465). The Parkin sequence was aligned, with each domain modeled as a separate unit built into a composite structure (see Text S1 – Part 1). The modeling was built as a hybrid model from consensus between the programs PRIME (Prime v3. 0, Schrödinger, LLC, New York, NY) [44], [61], YASARA SSP/Homology/PSSM Method [44], [62]–[67], and TASSER [68]–[73]. The variable loops and gaps were filled using knowledge-based homology and knowledge-based potentials with YASARA, or ab initio approach of ORCHESTRAR [74]. Each missing loop was modeled using the Loop Search module implemented in Sybyl 8. 0 or with YASARA loop modeler [44], [45], [63], [75], [76]. Only loops with the highest homology and lowest root mean square deviations were selected for the final models. The side chains and rotamers were adjusted with knowledge-based potentials, simulated annealing with explicit solvent, and small equilibration simulations using YASARA' s refinement protocol and verified by WHAT-IF and PROCHECK [77]. Fragments were divided into overlapping groups between the five templates (see Text S1 – Part 2). Combined fragments were overlaid using in-house superposition algorithms to determine optimal overlay and energies, which left the extraneous overlaid residues to be discarded as unnecessary. Finally TASSER was considered for each fragment and the entire length protein. Refinement of the fragments was completed using YASARA' s refinement module. These refinements started with the Secondary Structure Prediction (SSP) feature of YASARA. Both homology and fold recognition were considered and a final refinement with the entire model was completed using YASARA for 250 ps of MD using knowledge-based force fields. Additionally, YASARA supports an extensive and large loop library for modeling loops and gaps. The sequence and identity of each fragment was reasonable (see Text S1 – Part 2) [44]. The superposition and subsequent refinement yielded an optimal model for the full-length human Parkin protein. Recently released X-ray structures for large portions of Parkin greatly increased the accuracy of the modeling. The final model was subjected to energy optimization with PR conjugate gradient with an R-dependent dielectric. The model conformation was verified with WHAT-IF and PROCHECK and has a valid conformation consistent with good phi-psi space [46], [78], [79]. Atom consistency was checked for all 465 amino acids, verifying correctness of chain name, dihedrals, angles, torsions, non-bonds, electrostatics, atom-typing, and parameters. Each model was exported to the following formats: Maestro (MAE), YASARA (PDB). Model manipulation was done with Maestro (Macromodel, version 9. 8, Schrödinger, LLC, New York, NY, 2010), or Visual Molecular Dynamics (VMD) [80]. MDS was completed on each model for conformational sampling, using methods previously described in the literature [35], [36], [81], [82]. Briefly, each Parkin system was minimized with relaxed restraints using either Steepest Descent or Conjugate Gradient PR, and equilibrated in solvent with physiological salt conditions, as shown in the literature. Following equilibration each system was allowed to run MD calculations between 100–1000 nanoseconds in length. The primary purpose of MD for this study was conformational variability that may occur in the UBL. We also conducted conformationally enhanced sampling with MD biasing methods, like MdMD and MC-based generators. The protocol for refinement include the following steps: (1) Simulated annealing with explicit water molecules and ions, (2) Energy minimization, (3) MDS for 500 ps to relax to the force field (both AMBER03 and YASARA2 were tested). Tables were generated for most optimal conformation. PRIME and YASARA also give output for likely dimerization. FoldX was utilized as a plugin within YASARA to achieve mutant comparisons. In summary, the FoldX algorithm may calculate protein-protein interactions, protein-DNA interactions, or mutations within a protein, whereby FoldX calculates ΔΔG of interaction: ΔΔGab = ΔGab− (ΔGa+ΔGb) + ΔGkon+ΔSsc [83]. Here, ΔGkon reflects the effect of electrostatic interactions on the kon and ΔSsc is the loss of translational and rotational entropy upon making the complex. Charmm27, Amber, and OPLS2005 force fields were tested with the current release of NAnoscale Molecular Dynamics 2 engine. The protein with hydrogens consists of 7,083 atoms. In all cases, we neutralized with counter-ions, and then created a solvent with 150 mM Na+ Cl- to recreate physiological strength. TIP3P water molecules were added around the protein at a depth of 15–18 Å from the edge of the molecule depending upon the side [84]. Our protocol has been previously described in the literature [36]. Solvated protein simulations consist of a box with between 1. 17×105 atoms including proteins, counter-ions, solvent ions, and solvent waters. Simulations were carried out using the particle mesh Ewald technique with repeating boundary conditions with a 9 Å nonbonded cut-off, using SHAKE with a 2-fs timestep. Pre-equilibration was started with 100,000 steps of minimization followed by 10000 ps of heating under MD, with the atomic positions of protein fixed. Then, two cycles of minimization (100000 steps each) and heating (2000 ps) were carried out with restraints of 10 and 5 kcal/ (mol·Å2) applied to all protein atoms. Next, 50000-steps of minimization were performed with solute restraints reduced by 1 kcal/ (mol·Å2). Then, 1000 ps of unrestrained MD were carried out, and the system was slowly heated from 1 to 310 K. The production MD runs were carried out with constant pressure boundary conditions (relaxation time of 1. 0 ps). A constant temperature of 300 K was maintained using the Berendsen weak-coupling algorithm with a time constant of 1. 0 ps. SHAKE constraints were applied to all hydrogens to eliminate X-H vibrations, which yielded a longer simulation time step (2 fs). Our methods for equilibration and production run protocols are in the literature [35], [81], [85], [86]. Equilibration was determined from a flattening of RMSD over time after an interval of>20 ns. Our biasing algorithm, MdMD (MdMD section) expedited conformational searching over timescales inaccessible otherwise. Translational and rotational center-of-mass motions were initially removed. Periodically, simulations were interrupted to have the center-of-mass removed again by a subtraction of velocities to account for the “flying ice-cube” effect [87]. Following the simulation, the individual frames were superposed back to the origin, to remove rotation and translation effects. The MdMD algorithm has been previously described in exhaustive detail for smaller systems [36]. The application of MdMD allows the user to alter the shape of X-ray crystallographic structures to match the cryogenic-electron microscopic (cryo-EM) data, which may present an alternative conformation of the structure. In doing so, the cryo-EM density can drive the MD toward an unknown conformation. By automating this process, human biases and errors are minimized for the model making process. To prevent local wells, the time of sampling, MD sprint varies and Boltzmann velocities can be applied directionally. MdMD algorithm was implemented to collect representative pathway data for Parkin dynamics between LC-MOD/MCMM generated states. Average time for MdMD pathway is between 10 and 20 ns, while the count for discarded states during global variable testing ranges from 16% to 8%. Several replicate runs of MdMD were performed. Using Schrödinger' s Conformational Search Suite, we employed the following scheme. Potential terms were set using the OPLS2005 force field with water solvent and charges designated from the force field. We used an extended cutoff for electrostatic calculations (default: Van der Waals 8. 0 Å, Electrostatic 20. 0 Å, H-bond 4. 0 Å). We utilized constraints fixed on each zinc atom and the associated zinc-finger amino acid residues (i. e. cysteines, histidines, etc), which were determined using harmonic restraints at a force constant of 50 kcal/mol per selected atom. Additionally, torsions of the adjacent amino acid to the zinc-binding amino acid was softly constrained with a 10 kcal/mol force constant. Every other atom of the system was left to freely move under the conditions of the search algorithm, and the constrained pairs were assigned relative to each other, not to coordinate space, such that the zinc-finger could move as a single unit during conformational searches. The Powell-Reeves conjugate gradient energy minimization method was utilized on conformations achieved to return the structure to its local minimum for that particular conformation [88]. Several search schemes were applied to Parkin to look for large global conformational variation. Using the conformational search engine, we examined torsional sampling with MCMM, Mixed-torsional/Low-mode sampling, Large-scale Low-mode sampling, and Mixed torsional/Large-scale low-mode sampling (LC-MOD). When using the LLMOD search, we set the initial convergence criteria to 1. 000. To customize the search we used enhanced torsion sampling options with distance and torsion checking. Maximum number of steps per attempt was 5,000 with 100 steps per rotatable bond, saving 10 structures per search. The default energy window for saving structures is 5. 02 kcal/mol. Probability of a torsion rotation/molecule translation is 0. 5. Minimum distance for a low-mode move is 3. 0 Å and maximum is 7. 5 Å. The search was continued for 100 s of conformers, retaining only lowest energy models and discarding unrealistic structures. Structures from the LCMOD search form guidepost structures for MdMD biasing of the initial model to traverse the landscape of structures generated using Schrödinger' s Large-scale sampling using the initial crude guideposts and various biasing schemes. Final simulations relied on MdMD for smooth pathway transitions. Mutations of amino acids were completed using the Maestro within the Schrödinger suite with the mutate residue feature. Additionally, the build panel in Maestro conveniently allows for placing mutated residues (or growing them) automatically within an existing peptide chain. Also, MacroModel features within Maestro allow for the quick minimization of the structure for local geometry fixes to correct stereochemistry and packing of the amino acids. Modifications for amino acids, such as phosphorylation of Serine-65 (pSer65), was achieved using the 2D sketcher and importing as an new amino acid type which can be parameterized using the Schrödinger force fields. Using OPLS2005 or YASARA2, one can parameterize these modification and import into existing molecular dynamics integration engines, as the parameters for the modification are well documented for YASARA and Schrödinger [42]–[44], [63], [89]. MMB is an internal coordinate mechanics (ICM) [38] code which models 3D the structure and dynamics of macromolecules. It allows the user to control the mobility of all bonds and add constraints and forces [90]. MDS treats all atoms as independent particles in Cartesian coordinates. In ICM all atoms in a molecule are connected to each other, mostly by pin joints, which allow dihedral torsions about the bond axis. However the user may choose to also allow bond stretching and angle bending, or no freedom at all. Specified residues may also be constrained to a fixed relative position and orientation, with respect to each other or to the ground frame. The Amber PARM99 [91] or other force field can be applied, and its bonded and non-bonded terms can be separately scaled by the user [90]. MMB has successfully been used in RNA folding from biochemical data [92], threading [93], [94], and flexible fitting to density maps [95]. ZEMu is implemented in MMB [37], [96]. It consists of first, introducing the mutation and finding an energetically favorable local rearrangement around the mutation site, and second, computing the change in interaction energy (ΔΔG) using a knowledge-based (KB) potential. ZEMu equilibration of the mutation site proceeds as follows: First we specified a small (five residue) flexibility zone centered about the mutation site. The flexibility zone is treated in torsion space, leaving the remainder of the protein rigid and fixed. We then specified a larger, enclosing physics zone inside of which electrostatic and van der Waals forces are active. Due to the lack of solvent or other rigorous treatment of viscosity, it is possible for small chemical groups such as methyl to spin unnaturally quickly, which results in the variable time step integrator taking very small time steps. To deal with this we artificially scale the inertia matrices of such groups by a factor of 11. 0 – empirically found [37] to be sufficient to lengthen time steps without significantly affecting results. This method establishes a flexibility zone and a (generally larger) physics zone in the protein [97], [98]. The physics zone includes all residues within 12 Å of the flexible residues. The flexibility zone includes the mutated residue plus at least two residues on each side, for a total of five residues, in order to model the backbone rearrangements induced by the mutation [37]. For each proposed mutation locus, we equilibrated the corresponding flexibility zone in the wild type complex and evaluated the free energy of unfolding (ΔGwt) with the KB potential FoldX [99]. We repeated the calculation for the mutant complex to obtain the respective free energy of unfolding (ΔGmutant). The difference between these two quantities is our estimate of the experimental change in binding free energy induced by the mutation [100]: ΔΔG = ΔGwt−ΔGmutant. Using MMB and ZEMu with these equilibration and binding energy evaluation required an average of 15 minutes per protein, on a single core of a 3. 00 GHz AMD Opteron 6220 processor. All other calculations in the above MD sections were completed on a Xeon-based cluster with 120 hexa-core processors with 256 GB RAM available. FLAG-Parkin and pEGFP-Myc-Parkin wild type have been described before [3]. Mutant Parkin was cloned using site-directed mutagenesis. Constructs were sequence verified using BigDye Terminator v. 3. 1 and an ABI 3100 Genetic Analyzer (Applied Biosystems). Primer sequences can be obtained upon request. Human epithelial cancer cells (HeLa) were obtained from the ATCC (American Type Culture Collection), human embryonic kidney cells (HEK293E) from Invitrogen. Cells were maintained in DMEM containing 10% FBS at 37°C under humidified conditions and 5% CO2. The Parkin mitochondrial translocation assay has been previously been described [39]. Cells were seeded with 4000 cells/well in 96-well imaging plates (BD Biosciences) and allowed to attach overnight. Cells were transfected with empty vector, GFP-Myc-Parkin wild type or mutants using Lipofectamine2000 (Invitrogen). 48 h after transfection cells were washed 1× in PBS and fixed for 20 min in 4% paraformaldehyde. Nuclei were stained with Hoechst 33342 (1: 5000, Invitrogen) for 10 min and cells washed twice in PBS. Plates were imaged on a BD Pathway 855 with a 20× objective using a 3×3 montage (no gaps) with laser autofocus every second frame. Raw images were processed using the build-in AttoVision V1. 6 software. Regions of interest (ROIs) were defined as nucleus and cytoplasm using the build-in ‘RING - 2 outputs’ segmentation for the Hoechst channel after applying a shading algorithm. As a measure of Parkin translocation, the ratio of GFP signal intensity in the cytosol/nucleus was calculated. To exclude non-transfected cells and to ensure comparable transfection levels among analyzed cells, only ROIs with a GFP signal at least 30% higher than background were taken into consideration. Per Parkin construct we analyzed at least 4 independent experiments with 4 wells each per time point. HeLa cells were plated onto glass coverslips coated with poly-D-lysine (Sigma), fixed with 4% paraformaldehyde and permeabilized with 1% Triton-X-100 in PBS. Cells were incubated with primary anti-TOM20 (1∶2000, ProteinTechGroup 11802-1-AP) and anti-p62 antibodies (1∶500, BD Biosciences 610832) followed by incubation with secondary antibodies anti-mouse IgG Alexa Fluor-647 and anti-rabbit Alexa Fluor-568 (Molecular Probes) diluted 1∶1000. Nuclei were stained with Hoechst 33342 (1∶5000). Coverslips were mounted onto slides using fluorescent mounting medium (Dako). Confocal fluorescent images were taken with an AxioObserver microscope equipped with an ApoTome Imaging System (Zeiss). HeLa cells were transfected with FLAG-Parkin C431S variants using Lipofectamine 2000 according to manufacturer' s protocol and medium was replaced 4 h later. The next day, cells were treated with 10 µM CCCP for 0,1, 2,4, or 16 h. Cells were harvested in preheated (95°C) SDS lysis buffer (50 mM Tris pH 7. 6,150 mM NaCl, 1%SDS). Lysates were homogenized by 10 strokes through a 23G needle. Protein concentration was determined by use of bicinchoninic acid (Pierce Biotechnology). To verify the band shift by oxyester formation, aliquots of lysates were treated with or without NaOH (100 mM final) for 1 h at 37°C. NaOH was neutralized by addition of equal amounts of HCl before samples were run on 8–16% Tris-Glycine gels and transferred onto polyvinylidene fluoride membranes (Millipore). Membranes were incubated with anti-FLAG antibody (1∶100,000, Sigma F3165) overnight at 4°C followed by HRP-conjugated anti-mouse secondary antibodies (1∶15,000; Jackson ImmunoResearch 115-035-003). Bands were visualized with ImmobilonWestern Chemiluminescent HRP Substrate (Millipore) using a LAS-3000 Imager (Fuji). HEK293E cells transfected with FLAG-Parkin wild type were incubated with or without 10 µM CCCP for 1 h prior to lysis. Cells were washed once in cold 1× HBS (20 mM HEPES pH 7. 4,150 mM NaCl) and lysed in IP buffer (50 mM HEPES pH 7. 5,10 mM KCl, 150 mM NaCl, 1 mM EDTA, 0. 5 mM EGTA, 0. 2% NP-40+Complete (Roche) ). PhosStop (Roche) was added to all samples that were not treated with phosphatase. 750 µg protein was immunoprecipitated for 4 h using 20 µl of anti-FLAG EZview beads (Sigma). Beads were washed twice in cold IP buffer and once in 1× Ub buffer (20 mM HEPES pH 7. 4,50 mM NaCl, 5 mM MgCl2) before supernatant was completely removed. Phosphatase treated samples were washed 2× in IP buffer before supernatant was removed and 8 µl H2O, 2 µl 10× buffer and 10 U phosphatase (FastAP, Fermentas) were added to the beads for 30 min at 37°C. Beads were then washed twice with 1× Ub buffer before ubiquitination reaction. Recombinant proteins were purchased from Boston Biochem. Ubiquitination reaction contained 100 nM E1 enzyme (GST-Ube1, #E306), 1 µM E2 enzyme (#E2-622, #E2-640 or #E2-664), 2. 5 mM ATP, 2 mM DTT, 5 µg untagged ubiquitin (#U-100H), 1 µg N-terminally biotinylated ubiquitin (#UB-560) in 1× Ub buffer in a 20 µl reaction. Ubiquitination reaction was carried out for 1. 5 h at 37°C. Laemmli buffer was added and samples boiled for 10 min at 95°C. Samples were run on a 8–16% Tris-glycine gel, blotted and probed with horseradish peroxidase-coupled Streptavidin (1∶100,000; Jackson Immunoresearch 016-030-084) and anti-FLAG antibody (1∶100,000, Sigma A8592). Statistical analysis was performed with one-way ANOVA followed by Tukey' s post-hoc test. Error bars indicate S. E. M. | Parkinson' s disease (PD) is a devastating neurological condition caused by the selective and progressive degeneration of dopaminergic neurons in the brain. Loss-of-function mutations in the PINK1 or PARKIN genes are the most common causes of recessively inherited PD. Together the encoded proteins coordinate a protective cellular quality control pathway that allows elimination of impaired mitochondria in order to prevent further cellular damage and ultimately death. Although it is known that the kinase PINK1 operates upstream and activates the E3 Ubiquitin ligase Parkin, the molecular mechanisms remain elusive. Here, we combined state-of-the art computational and functional biological methods to demonstrate that Parkin is sequentially activated through PINK1-dependent phosphorylation and subsequent structural rearrangement. The induced motions result in release of Parkin' s closed, auto-inhibited conformation to liberate its enzymatic functions. We provide for the first time a complete protein structure of Parkin at an all atom resolution and a comprehensive molecular dynamics simulation of its activation and opening conformations. The generated models will allow uncovering the exact mechanisms of regulation and enzymatic activity of Parkin and potentially the development of novel therapeutics through a structure-function-based drug design. | Abstract
Introduction
Results
Discussion
Materials and Methods | movement disorders
biochemistry
molecular neuroscience
computer and information sciences
cellular neuroscience
medicine and health sciences
computer modeling
cell biology
neurology
neurodegenerative diseases
parkinson disease
biology and life sciences
biochemical activity
computational biology
computerized simulations
cellular structures and organelles
neuroscience | 2014 | Phosphorylation by PINK1 Releases the UBL Domain and Initializes the Conformational Opening of the E3 Ubiquitin Ligase Parkin | 16,271 | 298 |
The series of events that occurs immediately after pathogen entrance into the body is largely speculative. Key aspects of these events are pathogen dissemination and pathogen interactions with the immune response as the invader moves into deeper tissues. We sought to define major events that occur early during infection of a highly virulent pathogen. To this end, we tracked early dissemination of Yersinia pestis, a highly pathogenic bacterium that causes bubonic plague in mammals. Specifically, we addressed two fundamental questions: (1) do the bacteria encounter barriers in disseminating to draining lymph nodes (LN), and (2) what mechanism does this nonmotile bacterium use to reach the LN compartment, as the prevailing model predicts trafficking in association with host cells. Infection was followed through microscopy imaging in addition to assessing bacterial population dynamics during dissemination from the skin. We found and characterized an unexpected bottleneck that severely restricts bacterial dissemination to LNs. The bacteria that do not pass through this bottleneck are confined to the skin, where large numbers of neutrophils arrive and efficiently control bacterial proliferation. Notably, bottleneck formation is route dependent, as it is abrogated after subcutaneous inoculation. Using a combination of approaches, including microscopy imaging, we tested the prevailing model of bacterial dissemination from the skin into LNs and found no evidence of involvement of migrating phagocytes in dissemination. Thus, early stages of infection are defined by a bottleneck that restricts bacterial dissemination and by neutrophil-dependent control of bacterial proliferation in the skin. Furthermore, and as opposed to current models, our data indicate an intracellular stage is not required by Y. pestis to disseminate from the skin to draining LNs. Because our findings address events that occur during early encounters of pathogen with the immune response, this work can inform efforts to prevent or control infection.
Dissemination is key for a pathogen to reach sites where the environment favors survival or the probability of being transmitted to other hosts is higher. As the pathogen invades new tissues, however, the host responds by eliciting immune responses in an effort to eliminate infection. These interactions define the severity of disease and the outcome of infection. Thus, determining how host and pathogen interact during dissemination is key to understanding disease and to designing strategies to control it. Particularly relevant questions include what are the events that follow pathogen entrance into the body (i. e. inoculation) and how do these events define dissemination. The answers to these questions are key not only to deepen our understanding of the biology of infection, but, most importantly, to propose strategies that might interrupt pathogen spread in a clinical setting. Remarkably, for the great majority of pathogens, it is still unknown how dissemination into deeper tissues occurs. This is probably because experiments to study host-pathogen interactions in vivo can be extremely challenging, especially when using infection models that most closely mimic a natural infection (e. g. relevant route of inoculation, use of virulent strain, etc.). The challenges that are associated with the use of animal models are the main reason why most studies have relied on in vitro models to study infection. Notably, most of the current ideas of how host and pathogen interact early during infection derive from these in vitro studies. Yersinia pestis is the causative agent of bubonic plague, a severe bacterial disease characterized by aggressive dissemination within the host. This nonmotile bacterium first disseminates from the inoculation site (IS) into the draining lymph node (LN) after inoculation in the skin [1,2]. Colonization of the LN is then followed by bacterial escape into the bloodstream, resulting in septic shock and death [3]; escape into the bloodstream is a necessary step for ultimate transmission of the bacteria to a new host. The ability of Y. pestis to efficiently disseminate makes it an unparalleled model to study bacterial dissemination in vivo and to understand how a host responds to the threat of severe infection. Successful colonization of the host depends on the expression of bacterial virulence factors (e. g. type III secretion system, pH 6 antigen, F1 antigen) that are upregulated at 37°C and prevent phagocytosis [3–5]. These antiphagocytic factors are predicted to be expressed at low levels during the first hours of infection, a notion that gave rise to the hypothesis that an intracellular stage facilitates trafficking from skin to LN [6,7]. This is partially supported by in vitro experiments showing bacterial survival in macrophages [6]. Whether phagocytic cells are required for Y. pestis dissemination from the skin into the LN is still unknown. The goal of this study was to define what events occur immediately after inoculation of Y. pestis into the skin and how these events affect bacterial dissemination. Specifically, we sought to define the host-pathogen interactions that occur during dissemination. Most importantly, we were interested in testing whether Y. pestis requires phagocytic cells to disseminate from the skin into the draining LN.
Y. pestis survives in multiple tissues during infection. This has been shown extensively through experiments where tissues of infected animals are harvested to obtain bacterial burdens. However, with such an approach it is not possible to make observations of bacteria in the context of the different niches the pathogen interacts with in the host as it disseminates. With the aid of confocal microscopy of whole mounts (i. e. not sections), we visualized Y. pestis at the major anatomical sites the bacteria travels through during infection. An intradermal (ID) model of infection where ∼200 colony forming units (CFU) are injected in the ear pinna was used to mimic the delivery into the dermis that occurs during a flea bite. We chose this dose as it is highly relevant given that 83. 7% of mice inoculated by a flea receive <500 CFU and 93. 5% receive <1000 CFU[8]. While ID models of infection previously have been reported [9–11], our model is based on the use of a particularly small volume of injection (2 μL). A small volume reduces the possibility of confounding effects derived from tissue damage caused by larger volumes disrupting the dermis. After ID inoculation with Y. pestis expressing rfp (RFP-Y. pestis) or gfp (GFP-Y. pestis), whole mounts were fixed and imaged using confocal microscopy. With this approach, we obtained images of bacteria in tissues whose architecture was minimally disrupted without compromising the use of a fully virulent strain of Y. pestis. Bacteria were localized exclusively at two specific sites in the ear. One was the injection site, defined by the small and transient wheal (bubble) that forms in the skin during inoculation (S1 A-C Fig.). Bacteria at this site were found mostly in large clumps or as small groupings of cells (Fig. 1A). The second site where bacteria were found was at the base of the ear in the shape of tube-like structures (presumably lymphatic vessels), noted at 24 hours post inoculation (hpi). The presence of these tubes was infrequent (approximately 1 in 10 mice) but when seen, signal was so strong that it could be seen easily at lower magnifications (Fig. 1B). Bacteria were distributed unevenly in these tubes and appeared as individual or very tight clumps of bacteria (Fig. 1C, D, and E). Staining with DAPI revealed the presence of host cells in very close proximity to bacteria inside of these tubes (Fig. 1F). To determine if bacteria could be detected in the LN and lymphatic vessels that connect ear and LN, we removed the infected ear pinna along with as much adjacent tissue as possible up to the cervical region, including the draining LN and the lymphatic vessels that connect both tissues (S2 Fig.). In LNs at 24 hpi, Y. pestis appeared to be in discrete microcolonies and inside afferent lymphatic vessels that attach to the LN (Fig. 2A). At 48 hpi, very strong signal was detected from the same sites (Fig. 2B). Bacteria inside the lymphatic vessels attached to the LN were found either as single cells and small groups or formed tight clumps (Fig. 2C-E). Regardless of how they were arranged (single cells or clumps), the bacteria appeared to follow linear paths (Fig. 2D and E). These paths could be the result of fluid/sheer stress forces and might reflect the inner ‘structure’ of lymphatic vessels. There also were numerous bacterial associations with host cells, as revealed by DAPI staining (Fig. 2F). Lastly, bacteria were detected in the lymphatic vessels that connect the LN and the ear pinna (Fig. 2G-J), at sites distant from the LN (closer to the ear). While detection of bacteria was infrequent at these sites, when observed, bacteria appeared in tight aggregates that delineated the shape of the lymphatic vessel (Fig. 2 I and J). These are direct observations of Y. pestis from the tissues through which it disseminates. The use of a low-volume ID model of infection and an imaging approach using whole mounts provided images that depict how fully virulent Y. pestis interacts with the host in vivo. From our previous observations we hypothesized that the majority of bacteria we observed in the ear escape the skin to travel to the LN. To test this, we used a dissemination assay and asked whether all the members of the inoculum population could be found beyond the IS. The dissemination assay was based on the use of 10 oligonucleotide-tagged Y. pestis strains. These strains (hereafter referred to as “tagged strains” or simply “strains”) were generated by inserting a unique oligonucleotide tag at a neutral site in the bacterial chromosome [12]. Each tagged strain was found to possess the same level of virulence and growth as un-tagged Y. pestis. The frequency of each tagged strain in a tissue was analyzed with the Kluskal-Wallis test. No difference in the mean frequency of a tagged strain was found in LNs (p = 0. 9020,13 independent experiments) or ears (p = 0. 9739,8 independent experiments). Similar results were obtained with the Mann-Whitney statistical test (S1 and S2 Tables). A mix of 9 tagged strains served as the inoculum in our ID model of infection. The tenth strain (A6) served as a negative control. DNA extracted from bacteria recovered at desired time points post-inoculation was subjected to Southern dot blot analysis to determine which strains disseminated from the IS. DNA from the inoculum and from un-tagged Y. pestis was used as positive and negative controls, respectively (S3A and B Fig.). At early time points, 12 hpi, the average number of tagged strains in the LN was 2. 2 (range from one to seven tagged strains, Fig. 3A). At 48 hpi, colonization of the LN was well established, systemic dissemination had occurred, and mice were close to succumbing to disease. The average number of tagged strains present in the LN at 48 hpi was 2. 8 (range from one to five tagged strains, Fig. 3B). This suggests a bottleneck occurs during the first 12 h of infection, before systemic dissemination takes place. In addition to LNs, we also collected spleens, an organ we use to assess systemic dissemination. All but one mouse (n = 8) had the exact same strain population in spleens and in LNs (Fig. 3B). The mouse that was the exception to this, had strains A3, A4, A5, and B2 in the LN but only A5 and B2 in the spleen. Repetitions of this experiment showed the same trend: either the exact same strains in both organs (majority of cases), or spleens lacking one or more strains that were present in LNs. We never observed a strain present in the spleen that was not present in the LN. This is in agreement with the notion of Y. pestis disseminating from IS to LN and then to the rest of the body. More importantly, this suggests the bottleneck defines the population that is responsible for systemic colonization of the host and that will be (potentially) transmitted to a naïve flea. We then increased the inoculum 10-fold (∼2000 CFU) to test whether increasing the numbers of each tagged strain (∼222 CFU per strain) would alter the number of strains reaching the LN. This inoculum far exceeds the reported average inoculated in mouse skin by a flea (636 CFU, median 82 CFU) [8]. Out of eight mice, four had all 9 tagged strains, one had 8, one had 7, and two had 6 (S3C Fig.). This indicates that even when the number of CFU per tagged strain is considerably higher than what is expected during a natural infection, all the members of an entire strain can still be prevented from reaching the LN. These data indicate colonization of the LN and the rest of the body occurs from only a few bacteria after Y. pestis passes through a strong bottleneck. We also tested whether or not the bottleneck was an ear-specific phenomenon or if it could be replicated from a different anatomical site of the mouse. We performed a dissemination assay comparing mice inoculated in the ear (harvesting the superficial parotid LN) with mice inoculated in the foot (harvesting the popliteal LN). A limited number of tagged strains were observed in the LN (median of 3) and spleens (median of 2) of mice that were inoculated in the foot, indicating that bottleneck formation occurs at anatomical sites other than the ear (S3D Fig.). To determine if the bottleneck was strictly linked to the dermis, we repeated our dissemination assay using a subcutaneous (SC) model of infection [13]. Surprisingly, the median value of tagged strains in LNs from mice inoculated SC was higher (8 tagged strains) than in LNs from mice inoculated ID (3 tagged strains) (Fig. 3C). Furthermore, in half of the mice inoculated SC the bottleneck was completely abrogated. This indicates a key role of the dermis microenvironment in bottleneck formation. Bacteria that do not pass through the bottleneck are confined to the skin. We hypothesized that the source of the bottleneck was bacterial killing in the skin. To test this, we compared the number of strains present in the ear with the number of strains present in the LN of the same mouse. If our hypothesis was correct, the number of strains in the ear should not exceed the number of strains in the LN. Notably, at 12 and 48 hpi, the ear contained a number of strains that ranged between four to nine and that was significantly higher than the number of strains found in the LN (Fig. 3D and S3E Fig.). These results indicate the bottleneck is not due to rapid elimination of bacteria from the skin. These data also indicate bacteria that do not pass through the bottleneck (i. e. do not establish infection in the LN) are confined to the skin throughout infection. The above data suggest that after deposition in the skin, bacteria elicit a change in the microenvironment of this tissue altering their ability to move to the LN. To test this, we inoculated mice in the ear with a Y. pestis strain resistant to carbenicillin and 24 hpi, we inoculated the same animals at the same spot with a Y. pestis strain resistant to kanamycin. LNs were harvested 48 h after the first inoculation and plated on media with either carbenicillin or kanamycin. Bacterial loads in LNs were compared to those from a group of mice injected with PBS at the first inoculation time point (control group). We found lower bacterial loads of kanamycin resistant bacteria in LNs of mice whose ears were previously exposed to bacteria in comparison with those that were exposed to PBS (Fig. 3E). This suggests interactions of Y. pestis with the skin “activate” this tissue, resulting in detrimental effects to bacteria and a decrease in their ability to escape to deeper tissues. Thus, the bacteria that escape to the LNs, most likely do so before this activated stage takes place. Neutrophils control Y. pestis proliferation in the dermis. We thought that “activation” of the skin could derive from recruitment of cells of the innate immune response. Recent work suggested neutrophils are important during Y. pestis infection in the skin [10,14] but whether these cells play a role during subsequent infection steps is unknown. To gain insights into the role of neutrophils during infection, we tracked these cells by microscopy imaging using fluorescently labeled α-Ly6G, an antibody that binds to neutrophils [15]. At 30 minutes post inoculation, very few isolated clusters of neutrophils were observed in the skin (Fig. 4Ai and Aii). These clusters were distant from the bacteria, as bacteria and neutrophils were seen a few ‘fields of view’ apart from each other. This was distinctly different from later time points when bacteria and neutrophils could be visualized in the same field of view. Between 4 and 8 hpi, many bacteria were in close proximity to neutrophils or inside them (as determined by confocal microscopy, S4 Fig.). At 24 hpi a prominent increase in neutrophils was observed in comparison to previous time points. Neutrophils were highly concentrated at the injection site, forming very dense clusters. However, very few bacteria seemed to be associated with neutrophils at this later time point. We and others have speculated that neutrophils could contribute to bacterial trafficking to LNs [10]. However, depletion using a neutrophil specific depletion antibody (α-Ly6G, S5A Fig.) had no impact on bacterial trafficking (Fig. 4B), indicating neutrophils do not contribute significantly to this process. Neutrophil contributions to bottleneck formation were also assessed. Compared to mice injected with PBS, injection of α-Ly6G resulted in an increase in the number of tagged strains in the LN. This increase, while reproducible, was not statistically significant (S5B Fig.), which supports results from CFU data that indicate neutrophils are not needed for trafficking. However, we realized from our microscopy observations that substantial recruitment of neutrophils to the inoculation site occurred, and that bacterial numbers at this site did not seem to change over time. Thus, we hypothesized that neutrophils control bacterial burden in the skin without being able to clear infection. To test this, we compared bacterial burden in the skin (ear pinna) in mice treated with neutrophil depleting antibody or mock treated with PBS. Mice treated with neutrophil depleting antibody showed a highly significant increase in the number of bacteria in the skin when compared to mock treated mice (Fig. 4C). Overall, these experiments suggest Y. pestis interacts with neutrophils in the dermis and that these interactions severely restrict bacterial proliferation in the skin. Dissemination of Y. pestis in a phagocyte-independent manner. It is currently thought that for Y. pestis to move into LNs, an intracellular stage must exist [6,7]. However, results presented here and another recent study [10] suggest neutrophils are not important for bacterial movement to LN. In addition, Shannon, et al. concluded that the interactions of Y. pestis with dendritic cells in vivo are minimal and unlikely to be significant for bacterial dissemination [10]. We explored the notion that bacteria could be transported in a phagocyte-independent manner. In vivo models of bubonic plague infections use Y. pestis grown at 26°C, a temperature consistent with delivery from a flea. At 26°C, the antiphagocytic factors that are crucial for Y. pestis survival in the host are predicted to be weakly expressed. This is supported by studies with cell lines and primary cells that report Y. pestis is significantly less susceptible to phagocytosis by macrophages, dendritic cells, and to a lesser extent by neutrophils, when grown at 37°C than when grown at 26°C [16–18]. Because it has been predicted that Y. pestis travels from the skin to LNs inside of phagocytes, we expected that bacteria grown at 37°C would not disseminate efficiently to LNs. Surprisingly, we found that mice inoculated with comparable numbers of bacteria grown either at 26°C or 37°C showed no difference in bacterial loads in LNs at 12 hpi (S6 Fig.) or in LN and ears at 24 hpi (Fig. 5A). Comparable results were obtained in LNs, spleens, and ears at 48 hpi (Fig. 5B). These data indicate bacteria that are not susceptible to phagocytosis are as efficient in reaching the LN as bacteria that are susceptible to phagocytosis. Before this study, we always harvested Y. pestis from LNs several hours after inoculation. This is mainly because we assumed bacteria would move inside a phagocyte and movement of these host cells is slow, taking a few hours. Given the results presented here we wanted to test whether Y. pestis could reach LNs within minutes after injection, as happens with small molecules such as Evan’s blue (S2 Fig.). After injection into the skin, Evan’s blue can be found in LNs within 30 min as it moves with the flow of lymph, without the need of phagocytic cells. Thus, LNs were harvested at 10,20 and 30 min after bacterial inoculation in the ear. We detected Y. pestis in the LN as early as 10 min after inoculation (Fig. 5C), indicating the bacteria can reach the LN as fast as small molecules that travel through lymphatic vessels with the flow of lymph. Lastly, we wanted to determine whether bacteria that were associated with lymphatic vessels of the ear were also associated with phagocytic cells. To this end, we imaged RFP-Y. pestis during the first 30 min post inoculation, after immunofluorescence labeling of lymphatic vessels of the ear. As determined by DAPI staining, bacteria in close proximity or associated with lymphatic vessels were not associated with any host cells (Fig. 5D and E, S7 Fig. and S1 Video). Together, these experiments indicate that in our infection model, association with phagocytes is not necessary for Y. pestis to disseminate from the skin to the draining LN.
Pathogen dissemination in the host is a crucial and understudied process of infection. Although many studies have addressed bacterial-host interactions of Y. pestis (and other highly virulent pathogens), most of them are limited by the use of in vitro approaches and/or of attenuated strains. While many noteworthy observations have derived from these studies, we think that a complete picture of infection can only be achieved by also examining fully virulent strains in an in vivo context. Herein, we looked at dissemination of fully virulent Y. pestis in mice using three approaches: (a) an ID model of infection with minimal disruption to the skin and a low dose of bacteria to mimic delivery to the dermis that occurs during fleabite, (b) an assay that allowed us to follow dissemination within the host at a population level, and (c) microscopic imaging of the skin and underlying tissue to observe host-pathogen interactions at a cellular level. The combination of these strategies allowed us to make biologically relevant observations of the steps that follow bacterial entrance into the body and that define dissemination. The use of an ID model of infection is of particular relevance as hematophagous insects probe the dermal layer of the skin [19]. Moreover, histological sections of skin probed by infected fleas show Y. pestis is deposited into the dermis and not into the subcutaneous space [11]. Notably, our experiments show Y. pestis escape from the skin is restricted and that neutrophils play a role in controlling bacterial proliferation in this tissue. More importantly, we show bacteria that are not cell-associated enter lymphatic vessels and appear in LNs within minutes after inoculation. Our observations are not consistent with bacterial trafficking within phagocytes. While intracellular trafficking might occur at later stages, our data indicate the first bacteria that arrive in the LN do not need phagocytic cells to reach this compartment. Upon ID injection, the dissemination assay revealed a bottleneck that accounts for a previously unrecognized barrier for Y. pestis to disseminate from the IS. The bottleneck has strong evolutionary implications for Y. pestis, because it defines the population of bacteria that have the potential to be acquired by a naïve flea and thus be transmitted to a new host. More importantly, the bottleneck reveals even highly virulent pathogens, such as Y. pestis, encounter barriers that affect dissemination efficiency. Nearly 10% of our mice showed no bacteria in the LN. This was not due to ineffective inoculation as we confirmed that 100% of the mice had bacteria in the inoculated ear. This suggests efficient bottlenecks form in a fraction of immunocompetent individuals. Efficient bottlenecks in a few individuals could occur in part as a result of inherent variability in the immune responses of a population. It is not known how a natural flea inoculation, in comparison with needle inoculation, would affect bottleneck formation. In Leishmania major infections, a more robust and prevalent immune response is observed during sand fly versus needle inoculations [20]. However, in contrast to what was observed for Leishmania, no difference was observed in host responses to flea compared to needle inoculation of Y. pestis [21]. Therefore, if fleas deposit a relatively low dose of Y. pestis, one would expect the rate of successful infections from fleabites would be low, due to the bottleneck. In agreement with this, studies using fleas estimate the rate of successful plague infections in mice by flea bites to be less than 50% [8]. Interestingly, the bottleneck is abrogated when bacteria are delivered in the SC space beyond the dermis. This observation is relevant as it is very likely fleas deliver Y. pestis into the dermis and not the SC space [11,19,22]. The dermis is particularly proficient in triggering immune responses against invaders. Differences in immunogenicity between layers of the skin also have been shown to exist in cancer research; adenocarcinoma cells in rats develop into tumors only after SC but not ID injections [23]. The SC layer of the skin, on the other hand, is less immunologically competent than the dermis and this might facilitate the passage of more tagged strains into the LN [24]. This might occur due to a delayed influx of immune cells in SC tissue from blood therefore allowing for more local bacterial replication prior to dissemination. Very few studies have used imaging to probe host-pathogen interactions during cutaneous infections in vivo [19]. The use of an ID infection model and fluorescence confocal microscopy of unsectioned tissues provided us with high-resolution observations to reveal bacterial localization and associations in the host without requiring the use of attenuated strains. Microscopy imaging provides qualitative information to understand interactions with the host that is impossible to collect by traditional approaches using bacterial counts from harvested organs. Recent research suggested bacteria could evade a strong neutrophil response in the skin [10,25]. Our observations are in agreement with these reports, as we did not observe bacterial clearance in the skin. However, we also found that neutrophils severely restrict bacterial colonization of the skin, and thus, revealed a role of neutrophils in vivo. The bacterial restrictive properties of the skin are likely to be absent at the onset of infection. However, once present, they affect the bacteria in the skin in such a way that movement of newly inoculated bacteria into LNs is restricted, as shown by our double inoculation experiment. How pathogens disseminate from the IS into deeper tissues is one of the most relevant questions in microbial pathogenesis and one that is very difficult to address using direct approaches. For many pathogens, including Y. pestis, an intracellular stage to reach distant tissues has been proposed [6,7]. In this study, our data suggest association with host cells might not be necessary for Y. pestis to travel to LNs. Very few studies have addressed this question, but similar claims have been made for the highly virulent pathogen Bacillus anthracis [26] and for Salmonella abortusovis [27]. In the latter case, 80% of the bacteria traveling to the LN were found to be free in lymphatic vessels during the first 90 minutes of infection. While bacteria might travel to LNs in multiple ways, our data suggest that an intracellular stage is not required. Finally, we think that a phagocyte-independent mechanism for bacterial movement might explain bottleneck formation. This is because in a phagocyte-independent mechanism, only the very few bacteria that remain in suspension and do not adhere to the skin can be moved to the LN with the flow of lymph. In addition to the inability of all bacteria to be ‘taken’ by the flow of lymph and be conducted to a lymphatic vessel, bottleneck formation could also result from one or the combination of the following: (a) neutrophils might contribute to some extent to the bottleneck if bacteria encounter any of these cells during movement through the lymphatic vessels, and (b) phagocytic cells in the LN may contribute by killing newly arrived bacteria. Further understanding of the cause (s) of the bottleneck is the subject of future studies.
This study was carried out according to the recommendations in the Guide for Care and Use of Laboratory Animals of the National Institutes of Health. All animal studies were approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill, protocol 11–128. All efforts were made to minimize suffering; animals were monitored every 12 h following infections and were euthanized upon exhibiting signs of morbidity. Fully virulent Y. pestis CO92 [28] was used in all experiments. For the dissemination assay Y. pestis was tagged with 10 variants of an oligonucleotide signature tag [12]. The oligonucleotide tag, along with a kanamycin resistance cassette, was inserted at the Tn7 att site in the bacterial chromosome [29]. Each tag contains a unique sequence of ∼80 bp flanked by invariant sequences that were used for amplification. The 10 tagged Y. pestis CO92 strains were tested in our animal models to ensure each had retained the same virulence characteristics of the parent strain. For the dissemination assay, tagged strains were cultured in brain and heart infusion broth (BHI, BD Biosciences, Bedford MA) with kanamycin and incubated at 26°C unless otherwise stated. Standardized liquid cultures (based on optical density at 600 nm) were mixed in a single tube and the mix was serially diluted in phosphate buffered solution (PBS) to obtain the desired inoculum. Methods for detection of the tagged strains are described below. For the double inoculation experiments, the carbenicillin resistant strain also expressed rfp (for easy detection). A mix of the tagged strains was used as the kanamycin resistant strain. For the experiments with bacteria grown at 37°C, liquid cultures were grown in BHI with 2. 5 mM CaCl2 for 6 h at 26°C and then shifted to 37°C for 12. 5 h [30]. Where needed, kanamycin (kan) was added at 25 μg/ml and carbenicillin (carb) at 100 μg/ml. Six-to-eight week-old female C57BL/6J mice (Jackson Laboratory, Bar Harbor, ME) were inoculated under anesthesia (ketamine/xylazine). ID inoculations were done in the dorsal side of the ear pinna or the upper side of the foot. A volume of 2 μL was inoculated with the aid of a Pump11 Elite syringe pump (Harvard Apparatus, Holliston, MA) and a SURFLO winged infusion set with a 27-gauge needle (Terumo, Lakewood, CO). SC inoculations were performed as previously described [13] injecting a volume of 2 μL. Animals were sacrificed by injection with sodium pentobarbital. Organs were harvested at different time points and homogenized in PBS. Homogenates were serially diluted and plated on BHI agar and incubated at 26°C for 48 h to obtain bacterial counts. Mann Whitney or Wilcoxon matched pairs signed rank tests were used for statistical analysis, establishing statistical significance at p < 0. 05 using GraphPad Prism version 4. 0c (GraphPad Software, La Jolla, CA). Mice were inoculated with gfp- or rfp-expressing Y. pestis [31,32] following the procedures described above. After the mice were sacrificed, their ears were separated from the head, gently punctured with scissors (to facilitate fixative diffusion), and submerged in 10% buffered formalin for 24 h. The dermis of the dorsal leaflet was exposed by separating both leaflets and removing the layer of cartilage that separates them. Ears were mounted on glass slides with ProLong Gold antifade reagent with 4′, 6-diamidino-2-phenylindole (DAPI; Molecular Probes, Eugene, OR). Neutrophils were stained by tail-vein injection of fluorescently labeled antibodies against Ly6G (BD, Franklin Lakes, NJ). Lymphatic vessels were stained by ID injection of fluorescently labeled podoplanin antibody clone 8. 1. 1 (BioLegend, San Diego, CA) and LYVE1 antibody (Fitzgerald, Acton, MA) [31,33]. For all antibodies, 5 ng of antibody was used per mouse. Images were taken with an Olympus FV1000 MPE SIM laser scanning confocal microscope and analyzed with the Fiji (ImageJ 1. 48t) software package [15,33,34]. The same procedures were used to image LNs and the lymphatic vessels that connect ears and LNs. These vessels were also visualized by injection of 10% Evan’s blue in the ear. Bacteria that grew on agar plates from undiluted homogenized organs were mixed with 1 mL of PBS using a bacterial cell spreader until a homogeneous suspension was formed. A volume of 20 μL of this suspension was added to BHI broth with kanamycin and incubated at 26°C in a roller drum for 14 h. When less than 15 colonies were present on a plate, individual colonies were picked with a wooden stick and grown under the same conditions described. DNA was extracted from liquid cultures using a Wizard Genomic Purification Kit (Promega, Madison, WI). DNA concentrations were measured and standardized at 100 ng/μL. The oligonucleotide tag sequence was amplified by PCR using primers P2 (5’-TAC CTA CAA CCT CAA GCT-3’) and P4 (5’-TAC CCA TTC TAA CCA AGC-3’), which hybridize to the invariable region of all oligonucleotide tags. PCR reactions were prepared under standard conditions except for (a) the use of a mix of dideoxynucleotides (ddNTPs) with a ddATP: ddCTP: ddGTP: ddTTP ratio of 5: 1: 5: 5 and (b) ddCTP labeled with 32P was incorporated into the reaction. PCRs were cleaned using a MiniElute kit (Promega, Madison, WI). In this manner, this 32P-labeled amplicon consisted of a mix of DNA from any of the tagged strains that was present in a sample. Southern dot blot assays were conducted using the obtained labeled amplicon as probe. The probe was hybridized to positively charged nylon membranes (Roche, Manheim, Germany) previously crosslinked with 200 ng of plasmid DNA (pCIITN7K-a) containing the tags, as previously described [12]. The DNA crosslinked to the nylon membrane was arranged in 10 separate spots, each one containing DNA of each of the unique oligonucleotide tags used in the study. Because 10 of the oligonucleotides were used in the membranes and 9 were used to inoculate the mice, one of the oligonucleotides (A6) spotted on the membranes served as a negative control. Photographic film was exposed to the membranes with hybridized DNA and the developed film was scanned to obtain a digital image. The digital image was processed using the Fiji (ImageJ 1. 48t) software package [33] to invert the image (‘invert’ function) and to calculate the integrated density value per dot. An integrated density value six times higher than that of the negative control (A6), after background subtraction (‘subtract background’ function), was scored as positive. Ly6G+ cells were depleted after tail vein injection of 100 μL (0. 2 mg/mL) of low endotoxin Ly6G antibody clone 1A8 (BioLegend, San Diego, CA) [15,34]. Antibody to Ly6G was injected 24 h before inoculation of bacteria. Depletion of targeted cells was assessed by flow cytometry defining neutrophils as Ly6G+ cells; monocytes as F4/80−, CD11b+, Ly6G− cells; macrophages as F4/80+ cells; and dendritic cells as F4/80−, CD11b+ cells. | The earliest stage of any infection takes place when a pathogen enters the body (inoculation) at an initial site of contact. From this point, the pathogen can spread into deeper tissues where the pathogen itself and the immune responses against it cause disease. Very little is known about the events that follow inoculation and how pathogens move from the initial site of contact into deeper tissues. A better understanding of this process can potentially result in strategies to control or prevent disease. We studied the highly infectious bacterium that causes bubonic plague (Yersinia pestis) and how it spreads inside the body, from the skin into lymph nodes. We found that movement from the skin is highly restricted as only a small fraction of the bacteria that are deposited into this tissue are found in lymph nodes. While it is currently thought that Y. pestis spreads from the skin inside trafficking cells of the innate immune response, our work suggests these cells are not required for the bacteria to move into lymph nodes. Our findings can influence vaccine development efforts as these strategies are based on the study of early pathogen interactions with cells of the immune response. | Abstract
Introduction
Results
Discussion
Materials and Methods | 2015 | Dissemination of a Highly Virulent Pathogen: Tracking The Early Events That Define Infection | 9,087 | 249 |
|
In metabolism research, thermodynamics is usually used to determine the directionality of a reaction or the feasibility of a pathway. However, the relationship between thermodynamic potentials and fluxes is not limited to questions of directionality: thermodynamics also affects the kinetics of reactions through the flux-force relationship, which states that the logarithm of the ratio between the forward and reverse fluxes is directly proportional to the change in Gibbs energy due to a reaction (ΔrG′). Accordingly, if an enzyme catalyzes a reaction with a ΔrG′ of -5. 7 kJ/mol then the forward flux will be roughly ten times the reverse flux. As ΔrG′ approaches equilibrium (ΔrG′ = 0 kJ/mol), exponentially more enzyme counterproductively catalyzes the reverse reaction, reducing the net rate at which the reaction proceeds. Thus, the enzyme level required to achieve a given flux increases dramatically near equilibrium. Here, we develop a framework for quantifying the degree to which pathways suffer these thermodynamic limitations on flux. For each pathway, we calculate a single thermodynamically-derived metric (the Max-min Driving Force, MDF), which enables objective ranking of pathways by the degree to which their flux is constrained by low thermodynamic driving force. Our framework accounts for the effect of pH, ionic strength and metabolite concentration ranges and allows us to quantify how alterations to the pathway structure affect the pathway' s thermodynamics. Applying this methodology to pathways of central metabolism sheds light on some of their features, including metabolic bypasses (e. g. , fermentation pathways bypassing substrate-level phosphorylation), substrate channeling (e. g. , of oxaloacetate from malate dehydrogenase to citrate synthase), and use of alternative cofactors (e. g. , quinone as an electron acceptor instead of NAD). The methods presented here place another arrow in metabolic engineers' quiver, providing a simple means of evaluating the thermodynamic and kinetic quality of different pathway chemistries that produce the same molecules.
A primary scientific goal of metabolic research is to develop an understanding of the evolutionary, chemical and physical forces that shape the structure of cellular metabolism. Specifically, to what extent are present-day metabolic pathways the result of evolutionary optimization rather than fossilized accidents? In recent years various aspects of central metabolism have been explained on the basis of specific selection pressures and constraints imposed during evolution [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13]. Among the various constraints that shape the structure of metabolic pathways, thermodynamics features prominently, linking fundamental physical properties to pathway architecture [1], [9], [10], [11], [14], [15], [16]. Thermodynamic profiling also plays a central role in synthetic pathway design by identifying the most promising candidate pathways and discarding infeasible ones [14], [17], [18], [19], [20], [21]. Thermodynamic analysis is typically applied to determine whether a reaction direction or pathway is feasible in physiological conditions [9], [22]. Although not widely appreciated, thermodynamic potentials also constrain the kinetics of biochemical reactions and pathways [13], [23], [24], [25]. Specifically, the Gibbs energy dissipated by a reaction, ΔrG′, affects the net reaction rate through the flux-force relationship [23]: ΔrG′ = −RTln (J+/J−), R being the gas constant, T the temperature, J+ the forward flux and J− the backward flux. Consequently, an enzyme catalyzing a reaction that is far from equilibrium (ΔrG′<<0) carries almost no backwards flux (Figure 1a) while an enzyme catalyzing a near-equilibrium reaction (ΔrG′∼0) “wastes” many enzyme units catalyzing substantial flux through the reverse reaction. According to the flux-force relationship, as a reaction shifts towards equilibrium we would see an exponential increase in the number of enzyme units required to catalyze a single unit of flux. For example, a ΔrG′ of −7. 3 kJ/mol implies that about 5% of the enzymatic flux is in the reverse direction. Alternatively, a rather close-to-equilibrium ΔrG′ of −1 kJ/mol implies that about 40% of the enzymatic flux is in the reverse direction and the reaction rate is only ∼20% of the rate that would be achieved if all enzyme units catalyzed the forward reaction. Metabolic Control Analysis (MCA) is commonly used to describe the control that enzymes exert on metabolic fluxes [26], [27], [28], [29], [30]. This methodology starts with a given steady-state and mathematically describes how changes in enzyme abundance affect the pathway flux. Application of MCA requires enzyme kinetic properties which are laborious to measure and differ between organisms and isozymes. Here, we describe a complementary approach that requires no kinetic data and is not dependent on a particular initial steady state. We aim to identify pathways that, due to their thermodynamics, likely require higher enzyme levels to catalyze a unit of flux. Further, we pinpoint the particular pathway reactions responsible for these thermodynamic limitations. The flux-force relationship is instrumental in these analyses as it expresses a relationship between the Gibbs energy dissipated during a reaction (ΔrG′) and the amount of enzyme required to sustain a particular flux through that reaction. Therefore, the protein burden imposed by a pathway is directly related to the thermodynamic landscape of that pathway. We develop a quantitative framework to analyze the thermodynamic landscape of metabolic pathways. Our framework identifies those reactions within a pathway whose rates are constrained by low thermodynamic driving force. These enzymes will constrain the activity of the pathway unless they are present at high concentrations or are much faster-than-average catalysts [31]. Using this methodology it is straightforward to compare different pathways that achieve similar metabolic goals (unlike MCA, which assumes a particular steady-state for each pathway, making comparison difficult). To demonstrate our method, we apply it to pathways of central metabolism, including fermentation pathways (e. g. , Embden-Meyerhof-Parnas (EMP) glycolysis) and oxidative pathways (e. g. , TCA cycle). We compare various alternative pathways by their thermodynamic landscapes and identify the reactions supported by a low thermodynamic driving force and, hence, requiring a high enzyme expression level.
We used the Component Contribution method [32] for estimating the standard Gibbs energies of reactions, ΔrG′o (reactant concentrations of 1 M). This method produces a consistent set of ΔrG′o values by integrating three sources of information in decreasing priority: (1) ∼200 Gibbs energies of formation (ΔfGo) collected and published by Alberty [33], [34]; (2) apparent equilibrium constants (K′) available in the NIST database of enzyme-catalyzed reactions [35], [36]; and (3) the pseudoisomeric group contribution method as described in detail in ref. [22]. All ΔrG′o values were transformed to a pH of 7. 5 and an ionic strength of 0. 2 M [22], representing typical cellular conditions [37]. We use these conditions in all analyses presented in this paper, unless stated otherwise. The Gibbs energy dissipated by a reaction can be calculated from the standard Gibbs energy of the reaction (ΔrG′o) and the reactant concentrations. It is given by ΔrG′ = ΔrG′o+RT·ln (Q), where Q is the reaction quotient (also known as the mass action ratio). Because of its more intuitive sign we will often refer to a reaction' s Driving Force, defined as −ΔrG′ [38]. When analyzing pathways containing multiple reactions, it is convenient to use matrix notation [39]. We define S as the stoichiometric matrix, with rows corresponding to compounds and columns to reactions; i. e. , Sij is the stoichiometric coefficient of compound i in reaction j (positive for products and negative for substrates). Go denotes a column vector of reaction energies, i. e. , Goj is the standard change in Gibbs energy (ΔrG′o) of reaction j. Finally, let x be the column vector of the log-concentrations, so that xi is the natural logarithm of the concentration of compound i in molar units. Then the vector of reaction driving forces (−ΔrG′) is given by − (Go+RT·ST·x). We use a convention where all reactions in S are written such that the forward direction is the direction of the net flux in the pathway, and the stoichiometric coefficients represent the actual molecularities in the enzyme' s reaction center [38] (i. e. , the number of reactant molecular entities that are involved in the ‘microscopic chemical event’ constituting an elementary reaction). This convention obviates the situation existing in more general stoichiometric models [40], where scaling the flux of a reaction by a scalar and dividing the stoichiometric coefficients of that reaction by the same factor results in an equivalent system. Using the actual molecularities is vital for our analysis, as the flux-force relationship cannot accept an arbitrary definition of stoichiometry. Considering Go to be constant and allowing x to vary, a pathway is feasible if and only if there is at least one solution to the linear system defined by the constraints ln (Cmin) ≤x≤ln (Cmax) and −ΔrG′>0; i. e. , there must exist a set of metabolite concentrations within the predefined range (Cmin to Cmax) such that all reactions have a positive driving force. The concentration bounds (Cmin and Cmax) may be the same for all compounds or can be defined individually. Even if a pathway is composed of thermodynamically-feasible reactions, the pathway may be thermodynamically infeasible as a whole. That is, no solution for x exists within the chosen concentration range such that −ΔrG′>0 and so the pathway reactions cannot all be made feasible simultaneously [41]. As described in the Introduction, the driving force of a reaction constrains its rate, with near-equilibrium reactions requiring exponentially more enzyme to sustain the same rate as reactions far from equilibrium. We define the Flux-Force Efficacy – a unitless measure between 0 and 1 – as the ratio between the net flux () and the total flux (), which – according to the flux-force relationship – is related to the change in Gibbs free energy by. Hence, the higher the driving force of a reaction, the higher its Flux-Force Efficacy. Because of this interdependence between thermodynamic potentials and flux, pathways operating near equilibrium will incur a kinetic penalty due to backwards flux. We therefore attempt to quantify a pathway' s tendency to operate near-equilibrium. We seek a set of reactant concentrations that will maximize the driving forces of all reactions in the pathway. To achieve this, we use the minimum over all reaction driving forces (−ΔrG′ values of pathway reactions) as an optimization goal and maximize it – within metabolite concentration bounds – using linear programming. This can be formalized as a linear problem: where B represents a tight lower bound (i. e. , the minimum) on the driving force of all reactions. Maximizing B yields a solution where all reactions are as far from equilibrium as possible given the defined concentration ranges. The maximal value of B is denoted as the Max-min Driving-Force (MDF) of the pathway and is measured in units of kJ/mol [42]. If a pathway has an MDF of 7. 3 kJ/mol then a set of metabolite concentrations exists, within the pre-defined concentration range, such that all pathway reactions dissipate at least 7. 3 kJ/mol. A ΔrG′ = −7. 3 kJ/mol corresponds to a J+/J− ratio of exp (7. 3/RT) = 19, which in turn suggest that 95% of the overall flux is in the forward direction and 95/19 = 5% is in the backward direction. Therefore, the Flux-Force Efficacy is 95−5 = 90% (= 0. 9). Since the MDF is a tight bound, it is impossible to find a set of concentrations within the specified range for which all reaction driving forces are larger than, say, 7. 4 kJ/mol. Notably, the MDF solution is also equivalent to minimizing the total enzyme mass in a linear pathway, assuming that all enzymes have the same specific activities (see Text S1 for a mathematical proof). In order to quantify the extent to which a reaction or a metabolite affects the value of the MDF, we use the concept of shadow prices [42], [43]. Every primal linear optimization problem has a complementary dual problem. The variables of the dual problem – called shadow prices – quantify how much the value of the primal objective – i. e. , the MDF – will increase when a single constraint is relaxed by a unit amount. There are three types of constraints in the MDF linear problem: the lower bound (B) of the reaction driving forces (−Go−RT· ST · x≥B), the upper bound of metabolite concentrations (x≤ln (Cmax) ), and the lower bound of metabolite concentrations (ln (Cmin) ≤x). We therefore define the Reaction Shadow Price as the shadow price associated with the constraint on the driving force of that reaction, representing how much a change in Go will affect the MDF. The Metabolite Shadow Price is the maximum of the absolute values of the two shadow prices associated with the constraints (lower and upper bound) on that metabolite' s concentration. The shadow prices are the solution for the variables of the dual problem w, umax, umin: According to the definitions above, w are the reaction shadow prices and max (|umax|, |umin|) are the metabolite shadow prices. See Text S1 for a full derivation of the dual problem. The shadow prices represent a scaling between a change in the constraint and the resulting change in the MDF. For example, a reaction shadow price of 0. 25 indicates that a 4 kJ/mol decrease in ΔrG′o would increase the pathway MDF by 1 kJ/mol (assuming that this reaction still limits the pathway MDF). Similarly, a metabolite shadow price of 0. 5, associated with the upper bound constraint, implies that raising the upper bound concentration of this metabolite by 10 fold will result in the MDF increasing by 0. 5*RTln (10) ≈3 kJ/mol. Shadow prices are 0 for reactions whose ΔrG′o does not constrain the MDF, and likewise for metabolites whose concentrations do not constrain the MDF. Throughout our analyses we used metabolite concentration bounds characteristic of cellular physiology, a lower bound Cmin = 1 µM and an upper bound Cmax = 10 mM [9], [44], [45]. An exception was made for cofactors, whose concentrations were fixed to those characteristic of E. coli' s cytoplasm. Cofactors participate in tens or even hundreds of reactions and so their concentrations are considerably more constrained by the endogenous metabolic network than common reaction intermediates [46]. Fixing the concentrations of cofactors allows us to encode these constraints imposed by the wider metabolic network on individual pathways. Wherever possible, we constrained the cofactor ratios rather than their absolute concentrations, since the ratios are more conserved in many cases. The co-factor constraints we use are as follows: [ATP]/[ADP] = 10 [45], [47], [48], [49], [ADP]/[AMP] = 1 [45], [49], [NADH]/[NAD+] = 0. 1 [45], [50], [NADPH]/[NADP+] = 10 [45], [51], [Ferredoxinreduced]/[Ferredoxinoxidized] = 1 (corresponds to a reduction potential of −400 mV [52]), [orthophosphate] = 10 mM [53], , [pyrophosphate] = 1 mM [56], [57], [CoA] = 1 mM [45], [CO2 (aq) ] = 10 µM (ambient conditions). Enzyme abundances control the steady-state fluxes within a pathway [26], [27], [28], [29], [30]. When an enzyme is upregulated, the rate of the reaction it catalyzes increases instantaneously, but the rates of other reactions in the pathway do not change at first. This state, however, cannot be maintained for long as the concentration of the enzyme' s substrates decrease, while its products accumulate. Therefore, other reactions in the pathway are affected, and eventually, after the system settles in a new steady state, all fluxes might be altered. The term “control” describes such indirect, global effects. Potentially, all enzymes exert control on all fluxes within the pathway, but to different extents. Metabolic Control Analysis describes this control mathematically: if we consider small changes to a given steady-state, the effect of an enzyme' s abundance on the pathway flux, can be quantified by the scaled flux control coefficient [26], [58]: where J is the steady-state flux, vi is the rate of reaction i and Ei is the abundance of the enzyme catalyzing reaction i. In the general case, control coefficients depend on the pathway structure, enzymatic parameters, and allosteric regulation. Since all control coefficients for a flux J always sum to 1 [58], [59], control can only be redistributed among the different pathway enzymes. The flux control coefficients are related to the thermodynamic driving forces, as derived in Text S1 and was shown in ref. [26]. This relationship is easily derived for linear pathways whose enzymes follow the reversible Michaelis-Menten rate laws [26], [60], [61]. In Text S1, we derive the relationship for two specific cases: (1) all enzymes are completely substrate (but not product) saturated. Importantly, because all enzymes are essentially reversible, the rates of all reactions are sensitive to the concentrations of the substrates and products, even if all enzymes are substrate saturated. A full analysis of this case is given ref. [61]; (2) the substrates and products of all enzymes are well below their KM (enzymes operate in the linear regime). We show that in both cases the control coefficients are completely determined by reaction driving forces, such that the two are always correlated: the higher the driving force of a reaction, the higher the control it exerts on the pathway flux. Specifically, reactions with low driving forces have very low control coefficients. For the first case – all enzymes are substrate-saturated – we find that. For the second case – when all enzymes are substrate and product sub-saturated – we find that: , in agreement with the derivation of [26]. In both cases the scaling factor α is identical for all reactions and is determined by. The software for all the analyses presented in this paper is open source (MIT license) and stored in an online repository (https: //code. google. com/p/milo-lab/). We use IBM' s ILOG CPLEX Optimization Studio 12. 5 to solve the MDF primal and dual problems. The code which predicts the standard Gibbs energies [32] also depends on OpenBabel (http: //openbabel. org) [62], Calculator Plugins from Marvin (version 5. 10. 1) by ChemAxon, and the KEGG database (http: //www. genome. jp/kegg/) [63].
The kinetics of a reaction can be linked to three main factors: (1) maximal velocities and saturation levels, related to the enzyme kinetic parameters and to the concentrations of substrates and products; (2) enzyme abundances and (3) reverse flux though reactions [13]. Thermodynamics determines this last factor through the flux-force relationship: the reaction driving force (equivalent to the minus of the reaction change in Gibbs energy, i. e. , −ΔrG′) equals RTln (J+/J−), as presented above. For low enough driving forces (roughly 3 kJ/mol or less) the effect of ΔrG′ on the reaction flux is similar to that of kcat (or VMAX) in the sense that a fold change in either of them will change the reaction rate by the same fold change. The dependence of the reaction rate on −ΔrG′ decreases as −ΔrG′ increases above ∼3 kJ/mol. When −ΔrG′ exceeds 10 kJ/mol, the reaction rate is effectively insensitive to thermodynamic effects, reflecting the fact that there is negligible flux in the reverse direction. To address the effect of driving force on reaction flux more systematically, we define the Flux-Force Efficacy as the ratio between reaction net flux and total flux, (Methods). The Flux-Force Efficacy can be interpreted as the ratio between the actual reaction rate and the rate expected if backward flux was insignificant (assuming maximal velocities, saturation levels and enzyme concentrations are kept constant). Figure 1A shows how this ratio scales with the reaction driving force. We use the term “efficacy”, instead of “efficiency”, to distinguish between Flux-Force Efficacy and thermodynamic efficiency, as the two are antagonistic. For example, a reaction operating close to equilibrium is often considered to be thermodynamically efficient since it dissipates almost no Gibbs energy. Yet, such a reaction is characterized by a low Flux-Force Efficacy as J+≈J−. In contrast, a reaction with a high Flux-Force Efficacy, characterized by J+>>J−, is thermodynamically inefficient, dissipating a considerable amount of Gibbs energy. For a fixed enzyme level, a high Flux-Force Efficacy implies a high net reaction rate, as backward flux is negligible. On the other hand, a low Flux-Force Efficacy indicates considerable backward flux, leading to a decreased reaction rate. In a complementary manner, we can use this relationship to estimate the amount of enzyme required to sustain a particular flux through the reaction [13]. Figure 1B demonstrates this effect schematically using the energetic profiles of two putative pathways. Both pathways start and end with the same compounds, employ five enzymes and carry the same net flux. The kinetic parameters of all enzymes in both pathways, as well as enzyme and metabolite concentrations, are assumed to be identical. All reactions in the green pathway have the same, moderate driving force, which translates to a small backward flux. Hence, a small amount of each enzyme suffices. In the blue pathway, the driving force of the first two reactions is large while the last three reactions are near equilibrium. These final three reactions, therefore, require a lot more catalyst in order to sustain the same flux as the first two reactions in the blue pathway. When analyzing an entire pathway, it is essential to consider the interplay between the driving forces of all participating reactions: varying the concentration of a given metabolite can modulate the driving force of multiple reactions. We developed a method for finding metabolite concentrations – within an allowed range, see Methods – that maximize the driving force, and hence the Flux-Force Efficacy, of pathway reactions. Specifically, our computational tool uses the minimum over all reaction driving forces as an optimization function and maximizes it (Methods). This minimum, representing the smallest driving force among all pathway reactions, is defined as the Max-min Driving-Force (MDF). Figure 1C illustrates the application of the optimization approach to EMP glycolysis. The grey dashed line represents the ΔrG′o values of the different pathway reactions, while the magenta line represents the ΔrG′ values for each of the reactions after optimizing reactant concentrations to maximize the MDF. After this optimization, all reactions in the pathway have a positive driving force (i. e. , a negative slope) and so it is clear that the EMP pathway is thermodynamically feasible. Presuming the ΔrG′o values used are accurate and that our concentration bounds reflect cellular concentrations, pathways are thermodynamically feasible if and only if they have a positive MDF. Moreover, the value of the MDF indicates the degree to which a pathway is expected to be kinetically constrained by backward flux. A pathway with a high MDF can achieve a steady-state with very low backward flux as all of its constituent reactions can achieve high driving forces simultaneously. On the other hand, a pathway characterized by a low MDF contains reactions that are expected to have low driving force in physiological conditions. Due to the flux-force relationship, these reactions must either sustain low flux or be catalyzed by an abundant enzyme. For example, the green pathway shown in Figure 1B operates at high MDF which results in a high pathway flux and/or a low enzyme requirement. On the other hand, the blue pathway operates at low MDF, containing near-equilibrium reactions which reduce pathway flux and/or require higher enzyme levels. The shadow prices determine whether a specific reaction or metabolite constrains the pathway MDF, as described in detail in the Methods section. A decrease in the ΔrG′o value of a reaction with a positive shadow price would lead to an increase in the MDF. Similarly, if the concentration of a metabolite with a positive shadow price is permitted to violate the allowed concentration range (becomes too high or too low), the MDF increases (Methods). According to our model, enzymes catalyzing reactions with positive shadow prices are expected to be present at higher concentrations or have higher-than-average kcat values. While it is tempting to test this hypothesis systematically, it is, unfortunately, very challenging using available experimental data. Specifically, the MDF analysis requires that the magnitudes of all fluxes are precisely determined (taking into account that the flux through enzymes participating in the same pathway might differ due to an overlap with other metabolic routes). However, even for the relatively simple case of E. coli' s central metabolism, flux and metabolite concentration measurements from different groups vary significantly, even when performed under similar conditions (e. g. , [64], [65]). In addition, proteomic measurements related to lowly-expressed proteins are still quite noisy. Finally, most kinetic parameters reported in the literature were measured in vitro, which can differ considerably from those experienced in vivo [66], [67]. These issues limit our ability to perform a comprehensive systematic analysis of the relationship between the thermodynamic parameters, the measured kcat values and enzyme levels. Our current contribution suggests specific predictions to be tested when the needed experimental technologies mature. In the sections below we demonstrate our methodology by applying it to well-known central metabolic pathways. Our analysis, although not systematic, provides several examples of thermodynamic properties affecting pathway flux and suggests thermodynamic based explanations for key biochemical phenomena. In most organisms, the TCA cycle is the pathway responsible for the catabolic oxidation of organic compounds to CO2 (Figure 2A). Figure 2B presents the MDF of the TCA cycle (solid blue line) as a function of pH. We chose to vary pH, rather than other factors that affect the MDF, because cellular pH can differ considerably between organisms [37] and because the thermodynamics of many biochemical reactions producing or consuming protons is greatly affected by changes in pH. Figure 2B shows that the TCA cycle has a low MDF. In fact, it seems infeasible at pH≤7, which contradicts the observation that numerous organisms operate the TCA cycle at low cytosolic pH values [37]. To understand this puzzling finding, we asked which reaction (s) are responsible for constraining the pathway' s MDF – i. e. , which reactions have a positive shadow price. We find that, at non-alkaline conditions, the only reaction with a positive shadow price is malate dehydrogenase. The oxidation of malate to oxaloacetate using NAD as an electron acceptor (marked in red in Figure 2A) is characterized by a large positive ΔrG′o (>30 kJ/mol at pH≤7). How, then, are cells able to sustain high flux through the TCA cycle? The MDF framework enables us to suggest solutions for this apparent paradox. First, a high turnover number can compensate for operating at a low MDF: if the maximal activity of an enzyme is high enough, it will be able to operate sufficiently fast even at a low Flux-Force Efficacy. For example, an enzyme having a kcat of 100 s−1 and catalyzing a reaction with a driving force of only 0. 3 kJ/mol (Flux-Force Efficacy ∼6%) is equivalent to an enzyme having a kcat of 10 s−1 (the average kcat [31]) but catalyzing a reaction with a driving of 3 kJ/mol (Flux-Force Efficacy >50%). Thus, the high turnover number of malate dehydrogenase (well above 1000 s−1 [31]) might have evolved to compensate for its low driving force. However, this compensation effect does not answer how the cycle can carry flux at pH≤7, when malate oxidation is expected to become infeasible. Another possible explanation is that the concentration of oxaloacetate – having a positive shadow price (Figure 2A) – is lower than 1 µM, the lower-bound concentration assumed in our analysis. As oxaloacetate is an unstable compound [68], it is tempting to suggest that it is indeed found at a sub-micromolar concentration in-vivo. As shown in Figure 2B, allowing oxaloacetate concentrations beneath 1 µM increases the pathway' s MDF and the pH range in which it is thermodynamically feasible. However, keeping the concentration of oxaloacetate so low might be deleterious, as it would limit the rate of reactions which utilize this metabolite, e. g. , citrate synthase, aspartate transaminase and PEP carboxykinase. In fact, the relatively high affinity of citrate synthase towards oxaloacetate – KM being on the order of 1 µM [31], [69] – can be interpreted as representing an adaptation towards a low oxaloacetate concentration. A further possibility, also supported by experimental studies, is that oxaloacetate is channeled between malate dehydrogenase and citrate synthase [70], [71], [72], [73]. If channeling indeed takes place, the cellular concentration of oxaloacetate can be extremely low without compromising the rate of the enzymes utilizing it. From a thermodynamic point of view, malate dehydrogenase and citrate synthase can then be treated as a single reaction [9], [74]. This unified reaction does not represent any thermodynamic difficulty as its ΔrG′o is lower than −20 kJ/mol. As shown in Figure 2B, such a scenario increases the pathway MDF and makes it feasible in any physiological pH. Following the logic that substrate channeling can alleviate thermodynamic constraints, we expect that metabolites with positive shadow prices (i. e. , whose concentration constrains the pathway MDF) will have a higher propensity to be channeled between enzymes, therefore potentially guiding experimental efforts to such locations in search of evidence for substrate channeling. When high throughput methods for identifying channeling are developed, it will be possible to test this hypothesis systematically. Another solution to this thermodynamic puzzle might be the use of electron acceptors with a higher reduction potential than that of NAD. For example, various organisms operate a malate: quinone oxidoreductase enzyme [75], [76], [77], [78], [79]. In many of these organisms, this enzyme replaces more common NAD-dependent enzymes as the major route of malate oxidation [76], [77], [78], [79]. As shown in Figure 2B, using malate: quinone oxidoreductase enables the TCA cycle to operate at high MDF regardless of the cytoplasmatic pH. The downside of this approach is that less ATP can be produced via oxidative phosphorylation when using a quinone as an electron carrier instead of NAD. Finally, it is important to note that the TCA cycle is not actually a cycle in many organisms and under various conditions (e. g. , [80]). Instead, it often operates in a forked-mode, where malate dehydrogenase catalyzes the favorable direction (i. e. , oxaloacetate reduction), eliminating the thermodynamic constraints due to malate oxidation. Remarkably, it was recently suggested that E. coli uses a forked TCA cycle even during aerobic growth, despite the low ATP yield associated with this mode [65]. Several natural alternatives to the TCA cycle are also known to support the complete oxidation of organic compounds to CO2 [64], [81]. The structures of these pathways are given in Figure S1. Figure 2C compares these metabolic alternatives on the basis of their MDF and ATP yield per glucose. ATP is assumed to be produced from substrate-level phosphorylation and from NAD (P) H through oxidative phosphorylation. The P/O ratio – measuring how many ATP molecules are produced per one oxygen atom being reduced – was taken to be 1. 5, the representative value for E. coli [82]. Figure 2C suggests that the TCA cycle represents a combination of high ATP yield and high MDF which is better than most of its counterparts – especially if assuming substrate channeling of oxaloacetate (‘TCA channel’) or the usage of quinone instead of NAD (‘TCA MQO’). The oxidative pentose phosphate pathway (‘OxPP’), while producing less ATP molecules than the TCA cycle, supports the highest MDF among all oxidative pathways. EMP glycolysis (Figure 3A) is the most investigated fermentation pathway [10]. Substrate-level phosphorylation – coupled to glyceraldehyde 3-phosphate oxidation – is the process responsible for de novo ATP synthesis in the pathway (the downstream pyruvate kinase only recoups the ATP invested at the beginning of the pathway) [10]. Nevertheless, some organisms bypass substrate-level phosphorylation altogether such that glyceraldehyde 3-phosphate is directly oxidized to glycerate 3-phosphate, without producing ATP (Figure 3B) [83], [84]. Using the MDF methodology we can offer some insight as to why this may be. Figure 3C displays the MDF of the EMP pathway – with and without substrate-level phosphorylation – as a function of cellular pH. While the EMP pathway has a very low MDF and seems to be infeasible at pH<6. 5, the pathway variant which bypasses substrate-level phosphorylation is characterized by a far higher MDF. This suggests that organisms that do not depend on the degradation of organic compounds for energy conservation – like phototrophs or obligatory respiratory prokaryotes – can profit considerably by skipping substrate-level phosphorylation and operating at a much higher MDF, which can be translated into higher flux or, alternatively, to a lower protein investment required to sustain a given rate [13]. Notably, the trend shown in Figure 3C does not mean that the two substrate-level phosphorylation reactions (glyceraldehyde phosphate dehydrogenase and phosphoglycerate kinase) are the only ones that constrain the pathway MDF. The five reactions marked in red in Figure 3A are those with positive shadow price, showing that multiple reactions constrain the driving force of the EMP pathway. In fact, if any reaction from fructose bisphosphate aldolase to phosphoglycerate mutase had a more favorable ΔrG′o value then the MDF of the entire pathway would increase. Fructose bisphosphate is one of the two non-cofactor metabolites with a positive shadow prices (Figure 3A), and the only one whose concentration upper bound (10 mM) limits the pathway MDF (Methods). Interestingly, the concentration of fructose bisphosphate has been measured to be 15 mM [45], the only glycolytic metabolite whose concentration is higher than 10 mM. This 50% higher concentration adds ∼1 kJ/mol to the driving force of the thermodynamically-constrained reactions, increasing their rather low Flux-Force Efficacies. This example demonstrates how the methodology presented here can be used to rationalize why certain compounds attain higher (or lower) concentrations than others in cells. This further suggests a systematic study of whether an energetic analysis, as the one outlined here, can predict metabolite concentrations on a large scale. However, the measurement of metabolite concentrations using current technologies remains quite noisy, as evident by the dramatic discrepancies between different quantification methods (e. g. , [46]). As measurement technology matures, the generality of the connection between the range of metabolite concentrations and the thermodynamically-constrained reactions could be evaluated systematically. Several glycolytic variants are known to exist in nature and their structures and shown in Figure S2. Figure 3D plots the MDF (at pH 7. 5) of each of these pathways against the number of ATP molecules it produces per glucose molecule metabolized. As shown in the figure, there is a clear tradeoff between the MDF and ATP yield, with high- MDF pathways conserving less energy as ATP than pathways with lower MDF. Specifically, the methylglyoxal pathway (‘MGX’) – converting dihydroxyacetone phosphate into the highly reactive compound methylglyoxal [85], [86], [87] – and the non-phosphorylative Entner-Doudoroff (ED) pathway (‘EDNP’) – used by hyperthermophilic archaea [84], [88], [89] – seem to be promising choices for fermenting glucose if ATP production is not important but a high glycolytic flux is required [13]. Within the general trend shown in Figure 3B, some pathways seem better than others. The non-phsophorylative ED pathway, the phosphoketolase pathway (‘PKT’) – using the pentose phosphate pathway and cleaving xylulose-phosphate to glyceraldehyde-phosphate and acetyl-phosphate [90], [91] – and the pyruvate formate lyase pathway (‘EMP PFL’) – cleaving pyruvate to acetyl-CoA and formate and performing substrate-level phosphorylation on acetyl-phosphate – lie on the Pareto front [92], i. e. , no other pathway has both a higher MDF and a higher ATP yield. Notably, despite their prevalence in nature, neither the EMP nor the ED pathways are on the Pareto front, which suggests that thermodynamic properties alone are insufficient to explain the structure of central metabolism pathways, as we previously analyzed in detail (e. g. , [10]). Specifically, the phosphoketolase and pyruvate formate lyase pathways have higher MDF values than the EMP and ED pathways and yield at least as much ATP. However, it is known that other factors constrain the operation of these pathways in nature. The pyruvate formate lyase enzyme (EC 2. 3. 1. 54) employs an oxygen-sensitive radical mechanism and so can only be used in anaerobic or microaerobic environments [93], [94], [95]. This limitation may explain why the EMP-PFL pathway is less abundant in nature than MDF analysis would lead us to expect.
We introduce a quantitative framework for analyzing the thermodynamic profile of metabolic pathways and identifying reactions that limit metabolic flux within feasible pathways (i. e. , require high enzyme levels to sustain a specific rate). While near-equilibrium reactions can significantly increase the protein burden of a pathway, they may have certain advantages. For example, if the direction of a reaction must change quickly in response to some stimulus, operating near equilibrium (and at high enzyme level) is a good strategy: a small change in reactant concentrations can reverse the reaction direction while maintaining a similar absolute flux. This may be particularly important for glycolysis, where some carbon sources require glycolytic flux (e. g. , glucose and fructose) and others require flux in the direction of gluconeogenesis (e. g. , acetate and succinate). Therefore, fast environmental fluctuations in the availability of carbon sources may require speedy reversal of most glycolytic reactions, which is consistent with recent measurements indicating that reactions in glycolysis mostly operate with low driving-force in E. coli [45], [96]. Other functional advantages of working near equilibrium were recently suggested [97], [98]. Our methodology takes into account the physiological conditions, including pH, ionic strength, metabolite concentration ranges and cofactor concentrations. This feature is useful when comparing different organisms hosting the same pathway in different conditions. At the same time, the exact values of some of these parameters are not known with high certainty. In particular, the definition of the metabolite concentration range used in the optimization is challenging, as especially high (>10 mM) and especially low (<1 µM) metabolite concentrations have been measured (e. g. , [45]). Furthermore, the physicochemical properties of the metabolites affect their cellular concentrations [44], suggesting that the concentration ranges should be individually tailored to each metabolite. It is important to remember that the MDF methodology assumes that metabolite concentrations are optimized to achieve the most favorable thermodynamics. These optima are calculated using thermodynamic and stoichiometric data with respect to a single pathway and ignoring the rest of the endogenous metabolic network. Yet, in-vivo metabolite concentrations are constrained by many other factors, including their stability, permeability and their participation in other metabolic routes, and so cellular concentrations are unlikely to match these optima precisely. Hence, many of the pathways we analyzed might be more thermodynamically constrained than suggested by the MDF analysis Our analysis is sensitive to the definition of reactions, i. e. , what counts as independent metabolic steps. Merging reactions into a single metabolic step or splitting them into several steps can considerably affect the MDF of a pathway and the Flux-Force Efficacies of its reactions. For example, consider a reaction dissipating 2 kJ/mol and hence operating at a Flux-Force Efficacy of ≈40%. If this reaction is split into two steps, each of these will optimally dissipate 1 kJ/mol and its Flux-Force Efficacy will be only ≈20%. Hence, dividing a pathway into more steps results in lower MDF and Flux-Force Efficacies. Yet, the definition of metabolic steps is not arbitrary. A reaction should be treated as an independent metabolic step if all of its substrates and products are soluble. On the other hand, if two reactions involve a common reactant which remains bound to the enzyme (s), they can be treated as a single metabolic step [9], [74], as was suggested for channeling of oxaloacetate between malate dehydrogenase and citrate synthase. Notably, the MDF analysis is insensitive to the kinetic parameters of the enzymes participating in the pathway. In reality, the net reaction flux is determined both by the internal and external reaction energetics. Internal reaction energetics refers to the thermodynamic landscape associated with (i) the binding and release of the reactants from the enzyme' s active site; (ii) the different reaction intermediates formed during catalysis; and (iii) the activation energies of converting one reaction intermediate to another [99]. The internal reaction energetics determines the apparent kinetic parameters of the enzyme catalyzing the reaction (i. e. , kcat, KM) [99]. On the other hand, the external reaction energetics refers to the driving force of the net reaction, which depends on the concentrations of the substrates and products, as analyzed in this manuscript. Figure 4 schematically demonstrates the interplay between the net reaction flux and the internal and external energetic profiles. A reaction with a low kcat/KM should be compensated by a high driving force (Figure 4A), because otherwise the net flux will be low (Figure 4B). On the other hand, a reaction having a high kcat/KM can operate closer to equilibrium (i. e. , at a low driving force) and still sustain a high net flux (Figure 4C). Finally, a high driving force and a low internal thermodynamic barrier result in a very high net flux (Figure 4D). Interestingly, the thermodynamic driving forces of reactions can be directly connected to their control coefficients [27], [30], [100], [101], [102]: for a reaction that has a low thermodynamic driving force, the forward and backward fluxes are considerably larger than the net flux. For a near-equilibrium reaction, then, increasing the enzyme concentration will increase the forward and reverse fluxes to a comparable degree, bringing reactant concentrations even closer to equilibrium. This, in turn, lowers the already-low driving force of the reaction and neutralizes the effect of increasing the enzyme concentration. In brief, increasing the abundance of an enzyme catalyzing a near-equilibrium reaction will have only a modest effect on pathway flux. On the other hand, a reaction with a high driving force will exert high control on the pathway flux. For such a reaction the net flux roughly equals the forward flux, which is much larger than the reverse flux. In this case, increasing the enzyme abundance will mostly increase the forward flux (in absolute terms). Even if the driving force decreases somewhat, the flux-force efficacy will remain high (see Fig. 1A). Hence, an increase in enzyme abundance will not be compensated and will have a considerable effect on reaction rate. In the Methods and Text S1, we detail the direct mathematical relationship between reaction driving forces and flux control coefficients, which shows that upregulation of enzymes catalyzing reactions with high driving force has a large effect on pathway flux. Our ultimate goal is to establish a single framework that integrates pathway thermodynamics and enzyme kinetics. We believe it should be possible to reformulate measured kinetic data as thermodynamic potentials and analyze pathways in purely energetic terms. By considering the chemical potential of reaction intermediates and integrating these data with the concentrations of soluble pathway intermediates, one can arrive at a more complete analysis of pathway activity. It remains for future research to develop such an integrated framework. | Given data about enzyme kinetics and reaction thermodynamics, traditional metabolic control analysis (MCA) can pinpoint the enzymes whose expression will have the largest effect on steady-state flux through the pathway. These analyses can aid experimentalists in tuning enzyme expression levels along a metabolic pathway. In this work, we offer a framework that is complementary to MCA. Rather than focusing on the relationship between enzyme levels and pathway flux, we examine a pathway' s stoichiometry and thermodynamics and ask whether it is likely to support high flux in cellular conditions. Our framework calculates a single thermodynamically-derived metric (the MDF) for each pathway, which is convenient for selecting the promising pathways from a large collection. This approach has several advantages. First, enzyme kinetic properties are laborious to measure and differ between organisms and isozymes, but no kinetic data is required to calculate the MDF. Second, as our framework accounts for pH, ionic strength and allowed concentration ranges, it is simple to model the effect of these parameters on the MDF. Finally, as it can be difficult to control the exact expression level of enzymes within cells, the MDF helps identify alternative pathways that are less sensitive to the levels of their constituent enzymes. | Abstract
Introduction
Methods
Results
Discussion | systems biology
biochemistry
enzymes
metabolic pathways
biology
enzyme kinetics
metabolism
computational biology | 2014 | Pathway Thermodynamics Highlights Kinetic Obstacles in Central Metabolism | 11,201 | 278 |
Self-incompatibility (SI) is the flowering plant reproductive system in which self pollen tube growth is inhibited, thereby preventing self-fertilization. SI has evolved independently in several different flowering plant lineages. In all Brassicaceae species in which the molecular basis of SI has been investigated in detail, the product of the S-locus receptor kinase (SRK) gene functions as receptor in the initial step of the self pollen-rejection pathway, while that of the S-locus cysteine-rich (SCR) gene functions as ligand. Here we examine the hypothesis that the S locus in the Brassicaceae genus Leavenworthia is paralogous with the S locus previously characterized in other members of the family. We also test the hypothesis that self-compatibility in this group is based on disruption of the pollen ligand-producing gene. Sequence analysis of the S-locus genes in Leavenworthia, phylogeny of S alleles, gene expression patterns, and comparative genomics analyses provide support for both hypotheses. Of special interest are two genes located in a non-S locus genomic region of Arabidopsis lyrata that exhibit domain structures, sequences, and phylogenetic histories similar to those of the S-locus genes in Leavenworthia, and that also share synteny with these genes. These A. lyrata genes resemble those comprising the A. lyrata S locus, but they do not function in self-recognition. Moreover, they appear to belong to a lineage that diverged from the ancestral Brassicaceae S-locus genes before allelic diversification at the S locus. We hypothesize that there has been neo-functionalization of these S-locus-like genes in the Leavenworthia lineage, resulting in evolution of a separate ligand-receptor system of SI. Our results also provide support for theoretical models that predict that the least constrained pathway to the evolution of self-compatibility is one involving loss of pollen gene function.
Self-incompatibility (SI) is a widespread plant reproductive system that prevents inbreeding by facilitating the rejection of self-pollen. It is a major evolutionary feature of the flowering plants [1]. SI is a complex phenotype whose functioning requires co-evolution among several interacting components [2]. It has been proposed that SI evolved several times in the angiosperms [3], a hypothesis supported by molecular investigations that have also helped pinpoint the genes that control pollen specificity, pollen recognition, and the downstream reactions that mediate cessation of pollen tube growth [4]. The evolutionary loss of SI leading to self-compatibility (SC) and the potential for the shift to self-fertilization is often stated to be irreversible [5], [6]. Despite increasing knowledge of the mechanisms that underlie SI, the question remains as to how such a complex system could have evolved independently in many different angiosperm lineages. One answer may lie in the phenomenon of neo-functionalization of genes. It has been noted that the mechanisms that underlie SI share a number of features with another important plant function, namely pathogen recognition and rejection [7]. Moreover, it has become increasingly clear that evolution can reshuffle and reshape functions through exon recruitment and domain swapping [8], and so it is conceivable that SI could have evolved by co-opting genes with receptor and signaling roles that initially functioned in plant defense. Neo-functionalization of genes has been shown to be most likely when there are strong selection pressures [9]. The avoidance of inbreeding and its negative fitness consequences provide one such selective context [10]. In the sporophytic type of self-incompatibility (SSI), the pollen and stigma SI phenotypes (or “specificities”) are controlled by the diploid genotype of the parent (the sporophyte) [11]. SSI is known from 10 families of flowering plants [12]. It has been best characterized in the Brassicaceae family. In Arabidopsis and Brassica (and several other closely related Brassicaceae), the SI locus (S locus) contains two tightly linked genes that have been shown to be principally responsible for the SI phenotype [2], [11], [13], [14]. One of these genes, the S-locus receptor kinase (SRK), produces a transmembrane receptor expressed in the stigma. The extracellular domain of this protein can bind to the secreted protein ligand produced by the other S-locus gene, the S-locus cysteine-rich gene (SCR, also known as SP11), which is expressed in the tapetum of anthers, coating pollen with the protein product [15], [16]. When self-pollen recognition occurs, it initiates a signaling cascade that prevents self-pollen hydration and growth of the pollen tube [17], [18]. Though not included in the initial studies of the molecular basis of SSI in the Brassicaceae, the genus Leavenworthia has played an important role in evolutionary studies of plant mating systems. Detailed biosystematic work in the genus [19] documenting both inter- and intraspecific variation in the presence/absence of SI in a geographically localized region of the southern United States led to many subsequent investigations that focused especially on the ecology and population genetics of the group [20]–[24]. More recently, application of molecular genetic tools to the study of Leavenworthia uncovered a locus that co-segregates with the SI reaction, exhibits high levels of polymorphism, forms an allele phylogeny characterized by long terminal branches, and exhibits high effective rates of migration, and trans-specific polymorphism of alleles [25]–[28], all expected features for the S locus. The portion of the Leavenworthia S locus sequenced in earlier studies contains a number of characteristics also reported for SRK in other Brassicaceae, in particular an exon sequence that is similar to that of the SRK extracellular domain (S-domain), which contains several hypervariable regions thought to be involved in pollen recognition [25]. This gene was referred to as Lal2. Despite published evidence that Lal2 functions as SRK in Leavenworthia, the full sequence of the gene (i. e. , the expected seven exons coding for the entire extracellular S-domain, transmembrane domain, and kinase domain) could not be PCR-amplified using primers anchored in conserved regions of the SRK coding sequence, and no SCR gene (which is expected to be present in the genome close to SRK) was detected using PCR-based approaches. Moreover, the bulk of Lal2 alleles do not cluster phylogenetically with the SRK alleles of Arabidopsis, Brassica, and other Brassicaceae species. Two putative S alleles exhibiting sequence similarity to the S-domain of Arabidopsis lyrata SRK have been observed, but these represent fewer than 3% of the Lal2 alleles characterized to date [25], and in a series of five separate diallel crosses involving 20 plants, Lal2 allele sequences in each of 19 plants correctly predicted compatibility relationships, further indicating that it is unlikely that our investigations have failed to uncover the bulk of Leavenworthia S-locus haplotypes. The phylogenetic relationships of Leavenworthia S alleles to others in the Brassicaceae family is unexpected, especially given that biosystematic studies place the genus Leavenworthia in the tribe Cardamineae, which is more closely related to Arabidopsis and Capsella than to Brassica [29]. In this report we present new data on the Leavenworthia S locus gleaned from fosmid cloning, sequencing, expression analysis, comparative genomic, and crossing studies. While sequence characteristics and tissue expression pattern of both the pollen and stigma genes strongly support the hypothesis that the previously described Lal2 gene forms a portion of the Leavenworthia S locus, comparative synteny studies, along with closer examination of sequence variation at this locus, suggest that the Arabidopsis S-locus ortholog was lost in Leavenworthia following the divergence of the group from the common ancestor with other members of the Cardamineae. In addition, phylogenetic analysis of Lal2, SRK, and other gene family members suggests that SI in this genus is based on genes that have diversified separately and are thus likely paralogous to Arabidopsis SRK and SCR. We also show that two separate losses of SI in one species of Leavenworthia (L. alabamica) are likely due to independent mutations in the SCR-like gene coding sequence and/or its promoter. Together these results portray SI as a reproductive system that is more evolutionarily plastic than previously believed.
Leavenworthia alabamica includes several races that differ in floral characteristics and mating system [20]. The L. alabamica populations studied here belong to three races. The a1 race consists of SI plants with large, strongly scented flowers, and outwardly dehiscing anthers. Plants of race a2 are SC, with large but weakly scented flowers, and partially inward dehiscing anthers, while a4 plants are also SC, but with small flowers lacking scent, and fully inward dehiscing anthers. To better characterize the Leavenworthia alabamica Lal2 (LaLal2) gene and gain knowledge about its genomic context, fosmid libraries were constructed from single individuals of all three races. Clones containing LaLal2 were isolated after screening the libraries by PCR, and their sequences were obtained using 454 sequencing technology. The a1 race plant was heterozygous at LaLal2, whereas the a2 and a4 race plants were each homozygous for different LaLal2 alleles (whose S-domain sequences match those previously reported in these races [25]). One LaLal2-containing clone was obtained from each of the a1 race and a2 race libraries (35,750 bp and 39,236 bp, respectively). From the a4 race library, two overlapping clones were isolated; these assembled into one long contig of 64,895 bp. The assembled sequences from the different L. alabamica races cover a similar genomic region, and they share a number of structural features characteristic of other Brassicaceae SRK/SCR S loci. We therefore refer to them below as Leavenworthia S haplotypes. Also included in our analysis are partial sequences, obtained by PCR amplification, of an additional S haplotype found in a population of fully SI plants belonging to the a1 race. This S haplotype contains a LaLal2 S-domain sequence identical to that of the SC race a2. To distinguish between the a1 haplotype from the a1 fosmid clone and this second a1 haplotype, they are referred to below as a1-1 and a1-2, respectively. Previous sequence information available for LaLal2 was limited to the portion of the sequence corresponding to the extracellular domain of members of the S-domain 1 (SD-1) receptor-like kinase (RLK) gene family to which SRK belongs [25]. Analysis of the fosmid clones sequences allowed the full-length genomic sequence of LaLal2 to be determined. Homology of the full-length genomic LaLal2 sequence extends over the entire length expected for genes belonging to the SD-1 receptor kinase family. After excluding other Leavenworthia sequences, the highest match obtained from our BLASTn searches with the genomic LaLal2 sequence was NCBI Gene ID 9305017 from Arabidopsis lyrata (coverage 41%, E value 2e-106), which has no characterized function (Table S1). For brevity the NCBI Gene ID 9305017 will be referred to as the Arabidopsis lyrata Lal2 (AlLal2) gene. Other, lower similarity matches were to Brassicaceae SRK sequences. We determined the LaLal2 coding regions by combining data obtained from RT-PCR and 5′/3′ RACE sequences, which show that the gene has seven exons (Figure S1A), as observed in SRK [30]. The predicted amino acid sequences of LaLal2 and AlLal2 have signal peptide and transmembrane domain signature sequences, as expected for a transmembrane receptor coding sequence (Figures 1 and S1B). Domain organization of LaLal2 and AlLal2 proteins predicted by the SMART/Pfam online program [31] is as follows: two overlapping B-Lectin domains, an S_locus_glycoprotein domain and a PAN_APPLE domain in their extracellular domain, and an intracellular catalytic kinase domain, the latter being made up of the 11 subdomains described for protein kinases (Figures 1 and S1B) [32]. In addition to these domains, most of the known SRK alleles as well as their most closely related SD-1 RLK gene family members, ARK1 and ARK3, also possess DUF3660 and DUF3403 domains (Figure 1) [33]. Alignment of amino acid sequences of LaLal2 and AlLal2 to those of Brassicaceae SRK alleles (e. g. , AlSRK14, BoSRK12, and AhSRK43) as well as to those of A. thaliana ARK1 and ARK3 produced gaps in Lal2 sequences in regions corresponding to the DUF3660 and DUF3403 domains. Although A. lyrata and A. halleri SRK sequences belonging to the class B SRK alleles [34] also lack these two predicted domains (e. g. , AlSRK14 and AhSRK28), their sequences cluster phylogenetically within the clade of SRK alleles and not with the Lal2 sequences (Figures 1, S2, and 2). Moreover, upon closer examination of the regions around the deletions of DUF3660 and DUF3403 in class B SRK alleles (around residues 535 and 870, respectively), the amino acid residues flanking the deletions are seen to be more similar to SRK and ARK than to Lal2 (Figure S2). There are also a number of alignment gaps that were found to be specific to all LaLal2 and AlLal2 sequences (Figures 1 and S2). Altogether, LaLal2 and AlLaL2 appear to be gene orthologs that code for a type of SD-1 receptor kinase that is closely related to but distinct from SRK sequences. Lal2-like sequences were found in Brassica rapa (Bra010990) and Capsella rubella (Carubv10025960), though in genomic regions not syntenic with Leavenworthia and A. lyrata Lal2. Phylogenetic analysis of the full-length coding sequence of LaLal2 alleles, AlLaL2, and these Lal2-like sequences from C. rubella and B. rapa, together with that of SRK and the SRK-related sequences (e. g. , ARK2 and ARK3) of other Brassicaceae species, showed that the Lal2 group and the SRK-ARK group form two separate clades, which appear to have diverged before the onset of the strong allelic diversification of SRK (Figure 2A). Lal2-like sequences from C. rubella and B. rapa also form part of the Lal2 clade, and show the topological relationship in the tree expected from species relationships, as do the ARK3 sequences within the SRK-ARK clade [29]. Similar results were obtained when phylogenetic analysis is based only on the S-domain portion of the sequence, or on the transmembrane and kinase domain portions (Figure S3A and S3B), which suggests that the phylogenetic pattern of separate diversification of Lal2 is unlikely to be due to a domain-swapping event that may have modified a hypothetical duplicate of SRK. Synonymous and nonsynonymous substitutions differentiating LaLal2 and SRK sequences do not appear to be concentrated in any one portion of the gene (Table S2). We applied the branch-site model test [35] to detect positive selection at individual codon sites in LaLal2 sequences following their divergence from the most closely related sequences in the phylogeny (Figure 2B). The test rejects the null hypothesis of no selection and indicates that at least one codon (located in the hypervariable region of the S-domain described in [25]) has undergone positive selection (likelihood ratio test statistic = 8. 426, p<0. 005) following divergence from the other sequences. It has been noted that the SCR gene in previously characterized Brassicaceae S-locus haplotypes has the structure of a plant defensin [36]. In the three fosmid clones we sequenced, a gene exhibiting characteristics of a plant defensin was found ca. 2,000–10,000 bp upstream of LaLal2. This gene is referred to below as SCR-like (SCRL). The LaSCRL alleles of the a1-1 and a1-2 haplotypes contain full open reading frames and were used for further sequence analysis of the gene. Based on their cDNA sequences, we established that the SCRL gene consists of two exons, a characteristic common to the majority of plant defensin encoding genes [37]. Analysis with the SignalP online tool [38] predicts that the coding sequences of a1-1 and a1-2 LaSCRL translate into preproteins composed of an N-terminal signal peptide, required for protein secretion, and a small hydrophilic mature protein (Figure 3). The cleavage site of the signal peptide is predicted to be located after amino acid 25 in both a1-1 and a1-2 LaSCRL, generating mature proteins of 67 amino acids (aa) and 70 aa, respectively. While the signal peptide sequences of a1-1 and a1-2 LaSCRL are partially conserved (72% aa identity), the mature protein sequences are highly variable (32% identity), though like SCR, they contain eight cysteine residues (although their positions are not well conserved in the two sequences). Protein structure prediction using the modeling packages I-TASSER and DiANNA [39], [40] suggests that the LaSCRL product has a compact tertiary structure formed by disulfide bridges between a number of the cysteine residues, as seen in the SCRs of other Brassicaceae. BLAST searches with the cDNA sequence or the amino acid sequence of a1-1 LaSCRL found only a limited number of significant hits. As with LaLal2, however, the genes with highest similarity are found in A. lyrata: genes NCBI Gene ID 9302985 and NCBI Gene ID 9305018 (Table S3), neither of which has known functions. Sequence similarity with the two A. lyrata genes is mainly restricted to exon 1 of SCRL, which corresponds to most of the signal peptide sequence. NCBI Gene ID 9302985 and NCBI Gene ID 9305018 (Figure 3) are predicted to also encode mature proteins containing eight cysteine residues and that show low sequence identity with LaSCRL. Phylogenetic analysis was not possible with SCRL and SCR sequences due to difficulties in aligning the regions. Alignment of the three fosmid sequences together with sequence similarity searches in the A. thaliana genome database revealed that the diversity pattern in this Leavenworthia genomic region resembles the SRK/SCR S-locus region of other characterized Brassicaceae species [41]. The LaLal2 and LaSCRL genes themselves have high sequence diversity, but are flanked (at least on the right of LaLal2) by highly conserved regions (Figure 4A). If we define the core S locus as being the region of low sequence similarity between the three haplotypes and comprising LaLal2 and LaSCRL, the size of the S locus is 14 kb in the a4 haplotype, the only one for which sequence information on both sides of the S locus is available. Because the upstream sequences of the core S locus of the a1-1 and a2 haplotypes are currently undetermined, their sizes remain unknown, but are at least 15. 3 kb in the a1-1 haplotype and 11. 4 kb in the a2 haplotype. In all three Leavenworthia haplotypes, the LaLal2 and LaSCRL transcription units are arranged tail-to-tail and the gene order is the same. Annotation of the fosmid sequences using the A. thaliana reference genome revealed that the conserved regions on each side of the Leavenworthia core S locus are syntenic with an A. thaliana chromosome 4 region (Figure 4B). This region contains genes annotated as At4g37820 to At4g37910 on one side of the Leavenworthia core S locus, and genes At4g40050 to At4g39880 on the other side, but none with sequence homology to LaLal2 or LaSCRL. Moreover, there are no reports of an S locus in this region in other Brassicaceae species that have been examined to date, including A. lyrata. Therefore the existence of an S locus in this genomic region in Leavenworthia appears to be novel. As noted above, however, LaLal2 and LaSCRL do show sequence homology to annotated but uncharacterized genes in A. lyrata, with highest homology to, respectively, NCBI Gene ID numbers 9305017 (called here AlLaL2), and NCBI Gene ID numbers 9302985 and 9305018. All three genes are located in close proximity on A. lyrata scaffold 7, and notably, AlLaL2 and NCBI Gene ID 9305018 are positioned only 9. 8 kb apart, and are in a tail-to-tail configuration, like LaLal2 and LaSCRL in Leavenworthia (Figure 5). We refer below to the NCBI Gene ID 9305018 of A. lyrata as AlSCRL. Annotation of the surrounding genomic sequence using the A. thaliana reference genome revealed that this A. lyrata scaffold 7 region (between positions 852,500 bp and 1,060,200 bp) contains genes with annotations identical to all the genes found in the Leavenworthia a4 haplotype fosmid clone sequence. Most are homologous to genes on A. thaliana chromosome 4. However, a gene homologous to At1g26290 located on A. thaliana chromosome 1 was found in all three Leavenworthia haplotypes (between LaLal2 and the Leavenworthia At4g40050 homolog), as well as in the A. lyrata syntenic genomic region (Figures 4 and 5). In addition to the region homologous to the Leavenworthia Lal2/SCRL S-locus region, A. lyrata chromosome 7 also carries the SRK/SCR S locus, the latter being located at positions 9,335,860 bp (NCBI gene ID 9303924/ARK3) to 9,377,892 bp (NCBI gene ID 9305963/PUB8). The A. thaliana region carrying the SRK/SCR S-locus orthologous genes is also located between genes At4g21350 (PUB8) and At4g21380 (ARK3), in the homologous chromosome 4 region. Although the A. lyrata region with the homologs of the Leavenworthia LaLal2 region genes is also on chromosome 7, it is more than 8 Mb away from the S-locus region. Conversely, we were able to identify the Leavenworthia genomic region carrying the homologs of the Arabidopsis SRK/SCR S-locus genes from data obtained in an ongoing project to sequence the Leavenworthia alabamica race a4 plant genome (http: //biology. mcgill. ca/vegi/index. html). This Leavenworthia genomic scaffold is syntenic to genomic blocks found in the SRK/SCR S-locus region of A. thaliana (Figure 6A). Of special interest is the observation that the genomic block located between PUB8 and ARK3, which contains the SRK and SCR genes in Arabidopsis species, is highly reduced in length in L. alabamica, which is 1. 1 kb from the stop codon of the ARK3 ortholog to the start codon of the PUB8 ortholog (versus 4231 bp in the shortest A. lyrata S locus sequenced to date [41]), and neither SRK or SCR is present. PCR amplification and sequencing of the ARK3-PUB8 region in an a1-1 S haplotype homozygote plant confirmed the absence of SRK and SCR orthologs in that region in a SI individual as well (Figure S4). This result is consistent with earlier crossing studies that showed that Lal8, the putative Leavenworthia ARK3 ortholog, does not co-segregate with SI reactions [25]. Other PUB8 and ARK3 orthologs were not found in any other Leavenworthia genomic region. It is informative to compare S locus locations in different Brassicaceae species for which data are available. To date, S loci have been reported in three different synteny blocks. As part of the genome sequencing project mentioned above, we were also able to determine that Sisymbrium irio has a putative SRK ortholog with an apparently intact open reading frame (despite the fact that this species is self-compatible), with a location similar to that of Arabidopsis SRK gene (Figure S5). In Capsella rubella [42], the S locus also occupies a genomic region syntenic to the Arabidopsis SRK/SCR S locus [on scaffold 7, between positions 7,520,515 bp (Carubv10007030m/ARK3) and 7,563,814 bp (Carubv10005064m/PUB8) ]. In Brassica, the S locus genomic location is different, lying between orthologs of A. thaliana At1g66680 and At1g66690 [on chromosome 1 of Brassica rapa, between positions 17,225,424 bp (Bra004178/At1g66680) and 17,282,231 bp (Bra4183/At1g66690) ] [43]–[45]. The S locus locations and phylogenetic relationships of these genera are summarized in Figure 6B, which suggests that the Arabidopsis SRK/SCR S locus location is ancestral. Given the conservation of sequence and synteny described above for LaLal2 and LaSCRL versus AlLal2 and AlSCRL, we conducted an expression pattern study by RT-PCR of the two genes in a Leavenworthia plant homozygous for the a1-1 S haplotype and a A. lyrata SI individual in an effort to determine whether they could play a role in SI, or may have played such a role earlier in the evolutionary history of A. lyrata. It was shown previously that the SRK gene is more highly expressed in stigmas [44], [46] and that the SCR gene is expressed in anthers [13], [44] in Brassica and Arabidopsis, which is concordant with their respective roles in the SI mechanism. In Leavenworthia, LaLal2 expression was detected at similar levels in leaves, roots, and anthers and at higher levels in stigmas at the different stages of flower development (Figure 7A). In A. lyrata, AlLal2 expression was detected in anthers and stigmas at the different stages of flower development but not in leaves and roots (Figure 7B). As for the SCRL gene, its expression in Leavenworthia was detected in anthers, most strongly 2 d or 1 d before anthesis, and at lower levels in anthers at flower opening (stage 0), and in stigmas at the different stages of flower development (Figure 7A). LaSCRL expression could not be detected in leaves and roots. A similar expression pattern was observed for AlSCRL in A. lyrata (Figure 7B). Although the expression of LaLal2 is not specific to stigmas and the expression of LaSCRL is not specific to anthers (was also found in stigmas, which was also shown for SCR/SP11 in Brassica when using RT-PCR [43]), their expression in stigmas and in anthers, respectively, in higher levels than in other tissues is in accordance with their involvement in the SI mechanism. To compare the relative expression levels of AlLal2 versus AlSRK and AlSCRL versus AlSCR in A. lyrata, we also analyzed RNAseq data obtained from flower buds (stage 12) of the MN47 strain. Our analysis indicated that AlLal2 exhibits less than 8% of the expression level compared with that of AlSRK, and that AlSCRL exhibits less than 5% of the expression level compared with that of AlSCR (Table S4). We examined whether the A. lyrata Lal2 and SCRL genes exhibit a pattern of high polymorphism that would be expected if they play a role in SI. We amplified the S-domain of AlLal2 and the majority of the sequence of AlSCRL from 10 individuals in a single SI population (Population IND) located in Indiana [47]. PCR products were visualized on SSCP gels. Banding patterns across 10 individuals were identical for both genes, suggesting monomorphism in the population (Figure S6). We sequenced the single-stranded products for each gene, and these results show the presence of only one allele at each locus. This is in contrast to the observed high levels of polymorphism exhibited in the same population where the synonymous polymorphism for genes unlinked to SRK is πs = 0. 013 [48], suggesting that there is no evidence for a genome-wide population bottleneck in this population. The sequences of the a2 and a4 S haplotypes were obtained with the goal of determining the nature of loss of SI in these Leavenworthia SC races, particularly by analyzing sequences and expression of LaLal2 and LaSCRL in plants homozygous for the a1-1, a2, or a4 haplotypes. We included in these analyses the a1-2 haplotype found in SI plants of the a1 race. The a1-2 LaLal2 allele encodes an S-domain sequence identical to that of the a2 allele (Figure S7), and these two alleles should therefore have the same SCRL pollen specificity. None of the LaLal2 allele sequences includes any mutations disrupting the coding sequence (Figure S1B). Using stigmas of flower buds 2 d before anthesis, we found that LaLal2 is expressed at similar levels in plants homozygous for each of the S-locus haplotypes described in this study (Figure 8A). In contrast, analysis of LaSCRL sequences and expression revealed that the a2 and a4 alleles, from the SC races, have various disruptive mutations. In our race a4 plant, no LaSCRL expression could be detected in anthers 2 d before anthesis (Figure 8B), a development stage at which the a1-1 LaSCRL allele is highly expressed (Figure 7A). The coding region of the a4 LaSCRL allele deduced from the genomic DNA sequence contains a premature stop codon and the cleavage site of the signal peptide appears to be defective compared to that of the a1-1 and a1-2 LaSCRL alleles (Figure 3). Expression of the a2 LaSCRL allele was detected in anthers 2 d before anthesis (Figure 8B), but its translated sequence differs from that of a1-2 by one amino acid residue, and there is a premature stop codon after amino acid residue 45 (Figure 3). We crossed plants homozygous for the a1-2 haplotype or the a2 haplotype, to determine whether their incompatibility reactions fit those expected based on the sequence differences outlined above. The plant with the a1-2 haplotype appears to be compatible as a pollen recipient when a2 plants are used as pollen donors (89% of nine crosses produced fruit or had germinated pollen tubes). In contrast, the reciprocal crosses (a2 recipient plants and a1-2 pollen donors) appear to be incompatible with only 10% of 20 crosses that produced a fruit or had germinated pollen tubes. These proportions are significantly different (Z = 4. 135, p<0. 001) and support the hypothesis that SC in the a2 race is due to a mutation in SCRL (a1-2 pollen was shown to produce offspring when used in crosses with other pollen recipients). These results suggest that, as in other Brassicaceae, Leavenworthia possesses an S locus, which when disrupted leads to SC. Loss of SI in Leavenworthia a2 and a4 races is probably not due to loss of LaLal2 function, but to mutations in the male function SCRL gene. It is not known whether putative downstream genes in the SI pathway (e. g. , ARC1, MLPK) [49]–[51] are functional or not in all race a4 plants, though ARC1 appears to be deleted in a plant obtained from one a4 race (self-compatible) population [52].
We have characterized the Leavenworthia S locus in detail and have shown that it comprises two closely linked genes located in a genomic region of low sequence conservation among Leavenworthia haplotypes, as is also the case for the SRK/SCR S locus in other Brassicaceae members [41]. The two Leavenworthia S-locus genes, LaLal2 and LaSCRL, resemble the S-locus genes SRK and SCR in their sequence and expression pattern, but unlike their orthologs in populations of Arabidopsis lyrata, they are highly polymorphic. Phylogenetic trees constructed from Leavenworthia Lal2 alleles [25]–[28] show a pattern of long terminal branches similar to that observed at SRK/SCR S loci [53], [54]. While our previous studies indicated the existence of a functional S locus in the SI Leavenworthia races, the results reported here suggest that the genes comprising the Leavenworthia Lal2/SCRL S locus are unlike those of other Brassicaceae S loci that have been characterized to date. First, in Leavenworthia, SRK and SCR are absent from the syntenic block in which they occur in Arabidopsis and its close relatives, a genomic position that appears to be ancestral in the Brassicaceae. This is true in the case of the Brassica S locus as well, where it has been suggested that translocation of the entire S locus may have occurred [44]. However, the Brassica SRK sequences fall within the same clade as those of Arabidopsis and its relatives, despite the significantly greater phylogenetic distance between the genera as compared to Leavenworthia and Arabidopsis. By contrast, the Leavenworthia Lal2 sequences and their sequence homologs in other Brassicaceae taxa form a distinct clade, which appears to have diverged from the SRK-ARK clade before allelic diversification at SRK that presumably occurred at the onset of the ancestral SI system of Brassicaceae. As well, the Lal2 amino acid sequences have distinct deletions compared with those of Arabidopsis and Brassica SRKs. Finally, although the SCR-like gene in Leavenworthia shares several features in common with SCR, including high sequence diversity, a coding sequence with eight cysteine residues, and a defensin-like protein predicted to form a compact tertiary structure held together by disulfide bridges, they align too poorly with those of SCRs to be orthologous. Instead, the LaLal2 and LaSCRL sequences of Leavenworthia resemble SD-1 receptor kinase and defensin-like gene family members, respectively, found in a conserved syntenic block in A. lyrata, on the same chromosome as the SRK/SCR S locus but distant from it. Below we propose several possible explanations that could account for the distinct characteristics of the Leavenworthia S locus noted above. First we address the question of the time of the duplication event that gave rise to the separate SRK and Lal2 lineages, and second we address the question of the time of acquisition of pollen-pistil recognition function by Lal2/SCRL. Regarding the first issue, focusing on the phylogenetic relationships of the Lal2 and SRK sequences as shown in Figure 2, we note that these two groups of sequences form separate clades, and that the Lal2 group belongs to a lineage that apparently diverged from the SRK group before SRK became involved in self-pollen recognition and underwent allelic diversification. The alternative hypothesis—that there was a duplication of SRK that gave rise directly to Lal2 and occurred while SRK was already functioning in SI and thus still undergoing allelic diversification, but before the divergence of genera Arabidopsis, Capsella, Leavenworthia, and Brassica—is unlikely for the following reasons: (1) it is at odds with the structure of the gene tree and with the high level of divergence of Lal2 from SRK throughout the entire Lal2 sequence (Table S2); (2) under this hypothesis one would expect to find a gene tree with Lal2 and SRK sequences interspersed at the branch tips; and (3) if Lal2 functioned as a pollen protein-receptor this early in the evolution of SI, one would expect the level of polymorphism at Lal2 to be high. In earlier work we showed that there is a relatively low level of polymorphism at LaLal2 compared with SRK, and we found evidence of strong positive selection in hypervariable regions of the S-domain thought to be involved in recognition, both in our earlier studies [28] and in the PAML branch-site model analysis described above. Strong positive selection is thought to provide an indicator of recent diversification of the S locus, since negative-frequency-dependent selection for new S-allele specificities is expected to be most pronounced when S allele numbers are low, as expected following recent evolution of an S locus, or a population bottleneck [55]. Moreover, we have shown that the A. lyrata Lal2 and SCRL genes do not exhibit polymorphism. Regarding the issue of the time of acquisition of pollen-pistil recognition function by Lal2/SCRL, we propose two alternative scenarios. In both cases we assume that divergence of SRK and Lal2 predates the origin of SI in the Brassicaceae, and moreover, at the time of origin of SI in the family, these two genes were paralogous, with distinct functions and genomic locations. We assume that the lineage leading to SRK then acquired a role in SI and subsequently diversified leading to a large clade of SRK alleles that exhibit transgeneric polymorphism. It also likely gave rise to related genes (that do not have a function in SI) through duplication and translocation to new genomic locations unlinked to the S locus (e. g. , ARK1). According to the first scenario (Scenario I), the ancestral S locus (i. e. , with SRK/SCR) was lost at some point in the lineage leading to Leavenworthia, and so functional SI was lost as well (Figure 9). Pollen-pistil recognition then re-evolved based on a receptor-ligand system using the LaLal2 and LaSCRL genes, with a burst of diversification. Although this scenario involves a shift in the genes involved in pollen-pistil recognition in the SI system in the Leavenworthia lineage, it is possible that the genes involved in the signaling cascade leading to inhibition of pollen germination in the incompatibility reaction have remained the same as in the other lineages. Alternatively (Scenario II) the evolution of a new S locus in Leavenworthia could have been a two-step process, one in which SI was never completely lost (Figure 9). This could have occurred if one gene of the new S locus (e. g. , LaLal2) evolved pollen-protein recognition function, followed by evolution of a role as a protein ligand in SI for the second gene (LaSCRL), a series of events that could have been favored under high inbreeding depression if the ancestral system was “leaky” and allowed some selfing. Then, the original SRK/SCR S locus could have later been lost in Leavenworthia (perhaps following polyploidization). These two scenarios both fit the pattern of earlier divergence of Lal2 seen in the gene phylogeny (Figure 2), and are compatible with the evidence of relatively low diversity of Lalal2 alleles, and detection of strong selection in hypervariable regions of LaLal2 [28]. The data from this study are insufficient to know whether SI was lost in the lineage leading to Leavenworthia (Scenario I), or whether it was retained without interruption of the SI response (Scenario II), but there are several reasons to consider that SI may have been lost in the Leavenworthia lineage before being regained. First, the loss of SI is indeed common in the flowering plants and in the Brassicaceae—it has been estimated that half the species in the family are self-compatible [56], [57], and thus, the possible loss of SI within Leavenworthia cannot be considered as an atypical event. Second, Leavenworthia has recently been shown to be a paleopolyploid species (M. Lysak, A. Haudry, M. Blanchette, personal communication). As is the case in other such taxa, the evolutionary history of Leavenworthia likely involved interspecific hybridization followed by polyploidization. Hybridization and polyploidization in an individual possessing SI may lead to loss of fertility due to the absence of mates with gametes capable of producing viable offspring, which in turn could have led to selection for the loss of SI. That is, self-fertilization (as brought about by the loss of SI) may have increased the ability of an ancestral plant to form viable offspring [58]—this is not to say that polyploidy must necessarily have led to the immediate breakdown of SI [59], [60] but rather that polyploidization could have provided a “selective filter” that favored its loss. Clearly, Scenario I challenges the widely held notion that SI once lost is not easily regained [5], [6]. SI is, however, known to have evolved several times in the angiosperms, and so it is conceivable that it could re-evolve within the same family following loss of its pollen-pistil recognition system. It has been noted that the Brassicaceae is enriched for S-receptor kinase genes and these often occur near SCR-like genes [33]. Given the role that these genes play in recognition [7], it is possible that they could have formed the basis for the evolution of the pollen-pistil recognition system in SI in this family more than once. As well, we note that, though not specific, the expression of Lal2 and SCRL in stigmas and anthers, respectively, in both A. lyrata and Leavenworthia suggest the presence of regulatory elements necessary to bring about a new S locus in the lineage leading to Leavenworthia. It has been suggested that the loss of adaptations for outcrossing and transition to a high self-fertilization rate represent an evolutionary dead end, either because selfing lineages have higher extinction rates than outcrossing ones (due to accumulation of deleterious mutations), because of loss of adaptability, or because once lost, the purging of the genetic load leads to reduced inbreeding depression, so that outcrossing mechanisms cannot be easily regained via selection [57], [61]–[63]. If the Lal2/SCRL S locus arose following the loss of SI, the re-evolution of SI would require that the selective pressure, inbreeding depression, be retained. Theory suggests that if inbreeding depression is largely due to mutations with low selective coefficients, and if moderate levels of outcrossing persist following loss of SI, inbreeding depression may not necessarily be purged [64]. Scenario II is also interesting to consider. It would likely entail a period of evolutionary history in the Leavenworthia lineage in which two separate S loci could have co-existed within the same genome. SI systems with two unlinked recognition loci are known in the grasses [65]. We found different disabling mutations at the SCR-like gene in different SC populations of L. alabamica, suggesting independent loss of SI in these populations. The same conclusion was also inferred based on phylogenetic relationships among the SI and SC populations of this species [26]. The finding that mutations in the pollen gene are involved in each case where SI has been lost in L. alabamica parallels recent reports in Arabidopsis thaliana and A. kamchatica [60], [66] and also lends support to a prediction from population genetic theory that mutations disabling the pollen gene (as opposed to those disabling the stigma gene) should more easily spread in populations [67]. Moreover, the loss of SI in L. alabamica was probably recent, as LaLal2 genes in the SC populations are apparently still intact and expressed, and at least one of the SC L. alabamica populations studied here (the a2 race population) exhibits mixed selfing and outcrossing. Had the loss of SI and breakdown of SCR-like genes in these populations occurred in the more distant evolutionary past, it would presumably have rendered the LaLal2 gene selectively neutral and subject to mutational decay, and we would have expected to find a signature of such decay or neutrality in LaLal2 sequences. However, we cannot rule out the possibility that this gene also serves an additional unknown function, as suggested by the expression of LaLal2 in tissues other than stigmas. For example, a dual function has been found for an SRK gene in Arabidopsis [68]. The results of this investigation suggest that S locus evolution in Brassicaceae is more complex than initially thought. The vast majority of molecular-level studies of SI have been conducted with a limited number of model plant systems or their close relatives [4]. The work we present here, on a non-model organism, underscores the importance of looking outside these systems to understand more broadly the evolution of SI. It will be important to examine the genetic basis of SI in more distantly related Brassicaceae species to determine whether there are other taxa with SI systems that appear not to be based on SRK and SCR. Apart from the evidence that we have presented and discussed above, there are other types of information that could be useful in determining with greater certainty whether the S locus in Leavenworthia could have evolved as a duplication of the SRK/SCR S locus, rather than as a result of neo-functionalization, as we have proposed here. One potentially useful piece of information pertains to the role of Lal2 and SCRL in other Brassicaceae taxa. Even though apparent Lal2 sequence orthologs exist in other Brassicaceae species, there is no information available to test whether pollen recognition in SI is based on Lal2 alleles in any of these taxa (apart from what we have presented for Arabidopsis lyrata, suggesting that it is not). To further rule out the possibility that Lal2/SCRL-based SI exists in other Brassicaceae genera, it would be valuable to explore the levels of polymorphism of Lal2 and SCRL orthologs in other taxa and determine whether they are characteristic of an S locus. In addition, crosses could be conducted to reveal whether these genes co-segregate with SI reactions, as has been done in earlier studies that focused on the role of SRK and SCR in SI. The existence of a few rare S allele sequences in Leavenworthia with some similarity to those of A. lyrata and A. halleri SRKs (as noted above) could be interpreted as support for the duplication (as opposed to neo-functionalization) hypothesis. But such evidence is premature. These sequences could simply be pseudogenes that are linked to the Leavenworthia S locus, and therefore show cosegregation with SI. It would be useful to determine the genomic location of these sequences in the few plants where they occur, and whether they play any active role in SI. Finally, in future research directed at understanding the evolution of the Leavenworthia S locus, it would also be interesting to attempt transformation of SC species of Arabidopsis or Brassica with Leavenworthia SCRL and LaLal2 genes from the same haplotype, to determine whether these genes function within the context of the same downstream signaling pathway (s) as SRK/SCR [49]–[51].
Leavenworthia alabamica seed was sown in a 1∶1 mixture of PRO-MIX BX (Quebec, Canada) and sand. Plants used for expression analyses, genome sequencing, and fosmid cloning were grown in a Conviron PGW36 growth chamber under 14-h days at 22°C with a nighttime temperature of 18°C. Plants used for crossing were grown in a greenhouse at a minimum daytime temperature of 20°C and 18°C at night. Supplemental lighting was provided as needed to achieve a minimum day length of 12 h. When generating plants for expression analyses and crossing, plants homozygous for functional S-locus haplotypes (a1-1 and a1-2) were generated through self-pollination using a saline treatment modified from [69]. The stigma of the plant to be selfed was hydrated with 0. 5 M NaCl. After 1 h the stigma was then pollinated with self-pollen, either from an anther from the same flower or from another open flower of the same plant. The resulting progeny were screened for homozygosity for the allele of interest. Plants from the a2 and a4 races of L. alabamica are homozygous for the a2 and a4 LaLal2 S haplotypes, respectively. Crosses and pollen tube staining were conducted according to previously published methods [25]. Pollinations were considered compatible when more than five pollen tubes were visible in the style of the maternal parent or >1 seed was produced in the mature silique. The Arabidopsis lyrata plant used for AlLal2 and AlSCRL expression analysis was obtained from a seed collected in [70] and was grown in a Conviron PGW36 growth chamber under 16-h days at 22°C with a nighttime temperature of 18°C. Genomic DNA samples of plants of the a1-1, a2, or a4 S haplotypes used in fosmid library construction were extracted from purified nuclei. Nuclei were purified from fresh or frozen plant tissues. Tissues were grinded in liquid nitrogen using a mortar and pestle. Powdered tissues were added to freshly made and ice-cold nuclei extraction buffer [10 mM Tris HCl (pH 9. 5); 10 mM EDTA (pH 8. 0); 100 mM KCl; 500 mM sucrose; 4 mM spermidine; 1 mM spermine; 0. 1% β-mercaptoethanol] in a ratio of 20 ml of buffer per gram of tissue. Solution with added tissue was stirred using a magnetic stir bar for 10 min and then filtered through two layers of cheesecloth combined to one layer of Miracloth into a clean beaker. Cold lysis buffer (nuclei extraction buffer with 10% Triton X-100) was added at a ratio of 2 ml per 20 ml of nuclei extraction buffer. Solution was stirred for 2 min, poured into cold 50 ml polyethylene tubes, and centrifuged at 2,000 g for 10 min at 4°C to pellet nuclei. Supernatant was poured off, and the remaining supernatant was removed with a micropipette after a quick-spin. DNA was extracted from purified nuclei using Genomic-tips 20/G and the Genomic DNA Buffer Set (Qiagen). Instructions given in the Qiagen Genomic DNA Handbook (August 2001) for Yeast starting at p. 37, step 8 were used except for this following modification: at step 9, Proteinase K was added and incubation was carried overnight with gentle shaking at 50 rpm on a MixMate Plate and Tube Mixer (Eppendorf) to lyse the nuclei. Genomic DNA samples used in standard DNA analysis were extracted with the DNeasy Plant Mini Kit (Qiagen). Fosmid libraries were constructed using the CopyControl HTP Fosmid Library Production Kit (Epicentre Biotechnologies) as specified by the manufacturer' s instructions with the following modifications and specifications. Genomic DNA was sheared by passing gDNA samples 35 times through a Gastight 10 µl Hamilton syringe (model 1701). Sheared DNA was end-repaired and submitted to size separation by migration in a 1% low melting point agarose gel for 36 h at 35 V in 0. 5× TBE buffer. Insert DNA ranging from 23 to 40 kb was recovered from the gel matrix using GELase. We used 250 µg of purified DNA for ligation into the pCC2FOS Vector. After titering the packaged fosmid clones, cells were grown overnight at 37°C in liquid gel pools [71], [72] in 96-deep-well plates at a density of either 100 or 250 cfu per pool [200 µl of LB SeaPrep Agarose (Lonza Rockland Inc.) supplemented with 12. 5 µg/ml chloramphenicol (Cam) ]. Clones containing the Lalal2 gene were isolated by doing successive rounds of PCR screening on library pools of decreasing number of clones. In the first round, an aliquot of several library pools were combined to create superpools. Cells were pelleted by centrifugation and resuspended in sterile water. An aliquot of 0. 5 µl each of resuspended cells was used in standard PCR reactions. In the second round, pools from the obtained positive superpools were screened. In the third round, positive pools were plated on LB agar plates supplemented with 12. 5 µg/ml Cam to get isolated colonies. Colonies were individually picked and combined into pools of 10 colonies for PCR screening. Final screening round was carried on individual colonies grown on LB agar+12. 5 µg/ml Cam plates from positive pools of 10. To increase sensitivity of the screening, each round of screening consisted of two successive rounds of PCR reaction (primary and secondary). Primary PCR reactions were carried with primer pair Lal-Sdomain5′-F and Lal-Sdomain3′-R. Secondary PCR reaction used nested primer pair LalGenF and LalRcon. See Table S5 for primer sequences. Total RNA samples were extracted from plant tissues by using the RNeasy Plant Mini Kit (Qiagen). RNA samples were purified from DNA contamination by carrying an on-column treatment with DNase as specified in the manufacturer' s instruction manual. For expression analysis of Lal2 and SCRL by RT-PCR, 1 ug of total RNA was used in reverse transcription reactions using SuperScript II Reverse Transcriptase (Invitrogen, Burlington, ON) and Oligo (dT) 12–18. The 5′/3′ RACE reactions were carried with the FirstChoice RLM-RACE Kit (Invitrogen) using 2 ug of total RNA. The 5′ adapter-ligated RNA was reverse transcribed with the M-MLV Reverse transcriptase provided with the kit and using either random decamers or the 3′ RACE adapter as primers. PCR amplifications on reverse-transcribed products were carried using the following conditions: 1 µl RT products, 1× PCR buffer, 0. 2 mM dNTP mix, 2 mM MgCl2,0. 4 µM forward primer, 0. 4 µM reverse primer, and 0. 75 U Taq Polymerase (Invitrogen), in a final volume of 20 µl. PCR cycling was done in a C1000 thermal cycler (Bio-Rad) using the following program: initial denaturation at 94°C, 5 min followed by 35 cycles at 94°C, 30 s; 58°C, 30 s. ; 72°C, 1 min; and a final elongation step at 72°C, 5 min. See Table S5 for primer sequences. Illumina RNAseq reads from A. lyrata seedlings, roots, and stage 12 flowerbuds obtained courtesy of Dr. Richard Clark and Joshua Steffen were obtained using methods described in [73]. RNAseq reads were aligned to the A. lyrata reference genome (strain MN47: JGI) using both novoalign (Novocraft) and spliceMap (PMID: 20371516). Novoalign was used in read quality re-calibration mode with a low level of mismatch permitted (t = 50) between read and reference. Independently spliceMap was used to map reads spanning exon junctions. For each gene model, an expression level was determined by adjusting the read-count per gene by the exon-length and total reads in the respective sequencing libraries. Sanger, Illumina, and 454 sequencing were performed at the McGill University and Genome Quebec Innovation Centre. The genomes of Leavenworthia alabamica (a4 race), Sisymbrium irio, and the Leavenworthia short read data were gathered as part of an ongoing comparative genomics investigation involving these and other Brassicaceae species (Blanchette et al. , unpublished data). The sequences of the a1-1, a2, and a4 fosmid clones were also assembled from 454 data. In the case of the genomes, reads were generated in accordance with the Illumina protocols, with special attention paid to gentle shearing of mate-pair circular DNA to ensure >500 nt fragments, thereby reducing the probability of a read fragment-join chimera. Paired end (2×105, nominal 64 nt gap) Illumina reads were generated to a depth of 80× for each genome, trimmed for quality (3′ trimming where Q<32) and assembled with the Ray assembler [74] using automatic coverage depth profiling and a Kmer of 31. Scaffolding of Ray contigs was then undertaken with the SOAPdeNovo (BGI) assembler using a combination of 5 and 10 KBase mate pair reads (Blanchette et al. , unpublished data). Assembly of the fosmid sequences was undertaken in batches of pooled barcoded libraries covered by 1/8 of a flowcell of 454 sequencing (200× coverage). After stripping vector contaminants Newbler (Roche) was used to assemble the reads into ∼40 Kbase contigs using essentially default assembly parameters. Comparison of targeted fosmid assemblies (454) and short read whole genome assemblies (Illumina-Ray) from L. alabamica of the a4 race demonstrated high levels of concordance. Standard sequence analyses were done using the Geneious v. 5. 4. 6 software (Auckland, New Zealand) [75]. Amino acid and nucleotide sequences were aligned with MUSCLE [76]. Fosmid sequences were aligned using VISTA [77]. Annotation of fosmid sequences was done by sequence blast against the Arabidopsis thaliana genome. Because of the high sequence diversity of LaSCRL, this gene could not be detected by blast search but was found by eye examination of short ORFs obtained from different translation frames for the presence of eight cysteines. The Mauve Genome Alignment software v. 2. 2. 0 [78] was used to compare the S locus of A. thaliana with syntenic genome region of Leavenworthia and the S locus of Leavenworthia with syntenic genome region of A. lyrata. Protein domains were determined by submitting gene amino acid sequences to the SMART/Pfam prediction tools [31]. In addition to the a1-1, a2, and a4 LaLal2 sequences, we selected full-length coding SRK, and the closely related receptor-like kinase genes ARK1, ARK2, and ARK3 sequences from several Brassicaceae taxa. We included the coding sequence of AlLal2 (NCBI gene ID 9305017), the A. lyrata gene showing apparent orthology to LaLal2 as based on sequence similarity and conserved synteny (see above). Sequences homologous to Lal2 were identified in Capsella rubella (Carubv10025960m) and Brassica rapa (Bra010990). This was done as follows. First, pairwise alignments were generated between A. lyrata and L. alabamica, C. rubella, and Brassica rapa genomes, using lastz [79] in gapped, gfextend mode. These alignments were then chained [80] to generate extended sets of alignments split by gaps of less than 100 KBase. Low scoring chains were rejected and a subset of the highest scoring chains were annotated as candidate orthologous alignments between pairs of genomes. For the L. alabamica and B. rapa genomes, up to three orthologous chains were permitted for each region of the A. lyrata genome to represent orthology between the diploid and hexaploid contexts. The remaining chains were annotated as candidate homologous alignments. These alignment chains were used to identify candidate orthologs and homologs. The AlLal2 (NCBI gene ID 9305017), Carubv10025960, and Bra010990 predicted coding sequences were edited by sequence alignment of their genomic sequences with the Leavenworthia and A. lyrata Lal2 cDNA sequences obtained by sequencing. The outgroup for the analysis was selected from the sequences on the basis of closeness in evolutionary distance to the ingroup sequences as suggested by [81], from the Brassicaceae family RLK sequences examined in [33]. The sequences were aligned using the default settings in Clustal Omega v. 1. 1. 0 [82], and the best-fit nucleotide substitution model for the alignment was determined by the Aikake Information Criterion as implemented in jModeltest v. 0. 1. 1 [83], [84]. MrBayes v. 3. 1. 2 [85] was used to carry out Bayesian phylogenetic inference under the GTR+I+Γ substitution model. All parameters were estimated during two independent runs of six Markov Monte Carlo chains, both of which were run for 4,000,000 generations (longer runs gave identical results). Phylogenetic trees were sampled every 4,000th generation, and a consensus phylogeny was built from the 751 trees remaining after the first 250 were discarded as burn-in. Nexus formatted alignments including the commands used in MrBayes are available from the Dryad Digital Repository: http//dx. doi. org/10. 5061/dryad. mq5ct [86]. The branch-site model test for positive selection at codon sites was carried out using the CODEML program in the PAML 4. 4 package [87]. The tree (Figure 2B) was obtained using the PHYML [83] with default settings as implemented in Geneious v. 5. 4. 6 [75]. Foreground branches for the branch-site model were assumed to be those in which LaLal2 evolved separately from related sequences in Figure 2B. To determine whether sequence evolution of Lal2 associated with S locus evolution in this group was concentrated into particular protein domains, we compared the sequence of the a1-1 haplotype with that of the phylogenetically closest SRK sequence (allele SRK15 from Arabidopsis halleri). Estimates of synonymous and nonsynonymous substitution and their ratios were obtained by maximum likelihood using the program CODEML in the PAML package [87]. Estimated parameters for each major protein domain were compared by constraining them to be equal and carrying out the log likelihood ratio test. We amplified portions of AlLal2 and AlSCR from 10 individuals from the IND population of A. lyrata (material obtained courtesy of Dr. Barbara Mable) [47]. Polymorphism data of genes unlinked to the S locus were obtained from [48]. PCR primers are reported in Table S5, and PCR reaction protocols were identical to those reported above for RT-PCR. Amplicons were run on single-strand conformational polymorphism (SSCP) gels, as described in [28], [88]. Bands corresponding to single-stranded products of AlLal2 and AlSCRL were cut from the gel, re-amplified, and sent for Sanger sequencing at the McGill University and Génome Québec Innovation Centre (Montreal, Canada). Sequence trace files were edited by eye in Geneious v. 5. 4. 6 [75] and aligned to the reference copies of AlLal2 (100% identity) and AlSCRL (99. 8% identity). Sequences unique to this study were deposited in GenBank. | Self-incompatibility (SI) is a pollen recognition system that enables plants to avoid the inbreeding caused by self-pollination. It involves a pair of tightly linked genes known as the S locus. The product of one of these genes acts as the receptor and recognizes the pollen protein produced by the same plant, while the product of the other gene is the pollen protein that is recognized by the receptor. In this study, we have analyzed the gene sequence, genome organization, and gene evolutionary history of S loci in members of the Brassicaceae family, which includes plants of the genus Leavenworthia. From our analyses, we conclude that both genes that comprise the ancestral S locus in the Brassicaceae were lost in Leavenworthia. We show, however, that plants of this genus possess two other linked genes that exhibit patterns of polymorphism and expression that are characteristic of an S locus. These genes occupy the same genomic position in Leavenworthia as do two non-S-locus genes in the related species Arabidopsis lyrata, genes that are not known to function in self-recognition in this species. We suggest that these genes have evolved to assume the function of the pollen recognition system of SI in Leavenworthia—that is, that there has been de novo emergence of a distinct Brassicaceae S locus in this genus. We also present evidence that the breakdown of the SI system in two Leavenworthia races is due to independent mutations in the S-locus pollen gene, in accordance with theoretical predictions for the spread of S-locus disrupting mutations. | Abstract
Introduction
Results
Discussion
Materials and Methods | genome evolution
gene function
genome sequencing
plant science
plant genomics
plants
flowering plants
plant genetics
comparative genomics
biology
evolutionary genetics
flowers
angiosperms
plant evolution
plant phylogenetics
genetics
genomics
evolutionary biology
genomic evolution | 2013 | Secondary Evolution of a Self-Incompatibility Locus in the Brassicaceae Genus Leavenworthia | 15,799 | 376 |
Tumor necrosis factor (TNF) is critical for controlling many intracellular infections, but can also contribute to inflammation. It can promote the destruction of important cell populations and trigger dramatic tissue remodeling following establishment of chronic disease. Therefore, a better understanding of TNF regulation is needed to allow pathogen control without causing or exacerbating disease. IL-10 is an important regulatory cytokine with broad activities, including the suppression of inflammation. IL-10 is produced by different immune cells; however, its regulation and function appears to be cell-specific and context-dependent. Recently, IL-10 produced by Th1 (Tr1) cells was shown to protect host tissues from inflammation induced following infection. Here, we identify a novel pathway of TNF regulation by IL-10 from Tr1 cells during parasitic infection. We report elevated Blimp-1 mRNA levels in CD4+ T cells from visceral leishmaniasis (VL) patients, and demonstrate IL-12 was essential for Blimp-1 expression and Tr1 cell development in experimental VL. Critically, we show Blimp-1-dependent IL-10 production by Tr1 cells prevents tissue damage caused by IFNγ-dependent TNF production. Therefore, we identify Blimp-1-dependent IL-10 produced by Tr1 cells as a key regulator of TNF-mediated pathology and identify Tr1 cells as potential therapeutic tools to control inflammation.
TNF is a key pro-inflammatory cytokine required to control intracellular pathogens and kill tumours [1]. However, excessive TNF production can cause diseases such as rheumatoid arthritis, inflammatory bowel disease, psoriasis, ankylosing spondylitis, graft-versus-host disease and sepsis [2,3]. As such, TNF is a major target for the prevention of inflammatory diseases, and inhibitors of TNF activity are widely used in the clinic [3,4]. An important drawback to this approach is that it can increase susceptibility to infection, especially intracellular pathogens [5,6]. Therefore, a better understanding of how TNF is regulated during inflammation is needed to identify more selective ways to control disease while minimizing risk of infection. CD4+ T cells play critical roles in coordinating immune responses by helping B cells produce high affinity antibodies, CD8+ T cells to kill infected and transformed cells and innate immune cells to recognize and control pathogens and tumour cells [7,8]. Many diseases caused by protozoan parasites require the generation of IFNγ- and TNF-producing CD4+ T (Th1) cells for the activation of macrophages and dendritic cells to kill captured or resident pathogens [9,10]. However, these potent pro-inflammatory cytokines, along with other T cell-derived cytokines such as IL-17, can also damage tissues, and as such, CD4+ T cell responses need to be tightly regulated so they themselves do not cause disease [11]. IL-10 is a major regulatory cytokine, and its secretion by conventional CD4+ T cells can suppress inflammation by directly inhibiting T cell functions, as well as upstream activities initiated by antigen presenting cells (APC’s) [12]. Initially, IL-10 production was identified in Th2 cells [13], but has since been described in Th1 [14–16], FoxP3-expressing regulatory T (Treg) [17,18] and IL-17-producing CD4+ T (Th17) [19] cell populations. Thus, CD4+ T cell-derived IL-10 production is emerging as an important mechanism to prevent immune pathology. In mice infected with protozoan parasites, Th1 cells are an important source of IL-10 that can promote parasite survival, but also limit pathology [20–28]. These IL-10-producing Th1 (Tr1) cells have also been identified in humans with visceral leishmaniasis (VL) caused by Leishmania donovani [29] and African children with Plasmodium falciparum malaria [30–32]. Tr1 cells are increasingly recognized as a critical regulatory CD4+ T cell subset that prevent immune pathology during disease and protect tissue from damage caused by excessive inflammation [12,33–35]. Despite these protective functions, Tr1 cells may also promote the establishment of infection [34] and suppress Th1-mediated, tumour-specific immunity [36]. However, it is not clear how much of this activity can be attributed to Tr1 cells or other IL-10-producing cell types. Therefore, a better understanding of IL-10 regulation by different cell types is required for the development of new therapeutic approaches targeting this cytokine. Lymphoid tissue remodelling occurs in many chronic inflammatory settings associated with infectious, autoimmune and metabolic diseases [37–39]. This includes parasitic diseases such as malaria and VL that are associated with pronounced splenomegaly and disruption of lymphoid follicles [40,41]. In experimental models, this is accompanied by extensive vascular remodelling, white pulp atrophy and increased numbers of tissue macrophages [42–46], features also reported in human [47,48] and canine [49] VL. This remodelling results in dramatic changes to leukocyte movements in the spleen, and despite identifying excessive TNF production as a major contributor to these alterations [50,51], the immunoregulatory networks that fail are unknown. Here, we identify a novel pathway of IL-10-dependent control of tissue pathology during parasitic infection. We show that IL-10 produced by Tr1 cells protects against IFNγ-dependent, TNF-mediated tissue damage, but limited the control of parasites that cause malaria and VL. This pathway is critically dependent on the transcriptional regulator B lymphocyte-induced maturation protein 1 (Blimp-1), which promotes IL-10 production by Tr1 cells. These findings provide new insights into the regulation and function of Tr1-derived IL-10, and thus reveal new opportunities to harness the therapeutic potential of these cells to protect against TNF-mediated diseases.
The transcriptional regulator Blimp-1 (encoded by the Prdm1 gene) has recently been implicated in the generation of IL-10-producing Tr1 cells [52,53]. To explore the relationship between Blimp-1 and IL-10 production in T cells during protozoan infections, we made use of Prdm1fl/fl x Lck-Cre C57BL/6 (Prdm1ΔT) mice [54]. Strikingly, mice lacking Blimp-1 expression in T cells controlled non-lethal, P. chabaudi AS growth more efficiently than Cre-negative (Prdm1fl/fl) litter mate controls (Fig 1A). This corresponded with an increased frequency and number of activated CD4+ T cells and Th1 cells, but severely impaired development of Tr1 cells in the spleen (Fig 1B–1E). A similar, but much smaller effect (50 fold less) was also seen in CD8+ T cells (S1 Fig). The pattern of immune response observed in control Prdm1fl/fl mice was similar to what we (S2 Fig) and others [23] observe in wild type C57BL/6 mice, suggesting the presence of the flox transgene is having minimal influence over immune responses. In addition, the changes described above in mice lacking Blimp-1 expression in T cells were observed as early as 4 days p. i. , and clearly evident at day 7 p. i. (S3 Fig). When mice lacking Blimp-1 expression in T cells were infected with lethal P. berghei ANKA, they had reduced parasite burdens (S4A Fig), but despite delayed onset of severe disease and a small survival advantage, all mice ultimately succumbed with severe neurological symptoms (S4B Fig). Again, the reduced parasite burden in mice lacking Blimp-1 expression in T cells was associated with an increased frequency and number of Th1 cells, but impaired development of Tr1 cells (S4C and S4D Fig). Thus, Blimp-1 expression in T cells enhanced parasite growth and was critical for the generation of IL-10-producing Tr1 cells during experimental malaria. We next examined Blimp-1 expression in CD4+ T cells using transgenic Blimp-1/GFP reporter mice [55] infected with P. chabaudi AS. Blimp-1 expression in CD4+ T cells was highest in IL-10-producing cells, lowest in TNF-producing cells, while those producing IFNγ expressed intermediate levels of Blimp-1 (Fig 2A and 2B). A similar pattern of association between CD4+ T cell cytokine production and Blimp-1 expression was found in mice with experimental VL caused by infection with the human protozoan parasite L. donovani (Fig 2C). Interestingly, there was no difference in Blimp-1 expression between CD4+ T cells expressing IL-10 alone and Tr1 cells in both infections (Fig 2B and 2C). Consistent with the finding that IL-12 is an important driver of Blimp-1-dependent Tr1 cell differentiation in autoimmunity [52], we found that Blimp-1 and IL-10 expression by IFNγ-producing CD4+ T cells from L. donovani-infected mice required IL-12 (Fig 2D and 2E). This reliance on IL-12 was most apparent for CD4+ T cells because although IL-12 blockade caused a small, but significant, reduction in IL-10 and IFNγ double-producing CD8+ T cell frequency, this was not accompanied by a significant reduction in Blimp-1 levels. In addition, we found no significant reduction in IFNγ-producing CD8+ T cells, although we did measure a small, but significant reduction in Blimp-1 levels (Fig 2E). Importantly, we found increased accumulation of PRDM1 mRNA in CD4+ T cells isolated from the blood of VL patients, compared with the equivalent cell population in the same individuals after completion of drug treatment (Fig 3A). This was associated with elevated IL-10 mRNA levels in the same cells (Fig 3B), as well as increased plasma IL-10 (Fig 3C), as previously reported [29,56]. VL caused by L. donovani in mice is characterized by a chronic infection of macrophages in the spleen, but acute infection of macrophages in the liver [57]. Strikingly, and in contrast to littermate controls, mice lacking Blimp-1 expression in T cells controlled infection in the spleen effectively (Fig 4A). This improved control of parasite growth was again associated with an increased frequency of activated CD4+ T cells and Th1 cells, but restricted development of Tr1 cells (Fig 4B). However, despite improved control of parasite growth in mice with Blimp-1 deficient T cells, these mice presented with significantly larger spleens, associated with an increased frequency of TNF-producing CD4+ T cells (Fig 4C). Similar effects were also observed in the liver (Fig 4D–4F), and these were also associated with elevated serum TNF and IFNγ levels (Fig 4G). Enhanced CD4+ T cell responses were antigen-specific, as shown by increased IFNγ and TNF production in response to stimulation with parasite antigen (Fig 4H and 4I). Although IL-10 was not detected in serum, IL-10 production was measured in response to antigen re-stimulation, and consistent with the above Tr1 data, was significantly reduced in cells from mice lacking Blimp-1 expression in T cells, compared with littermate controls (Fig 4I). Blimp-1 has previously been shown to restrain CD4+ T cell IL-17 production [58], but we found no significant changes in serum IL-17 levels (S5A Fig) or parasite-specific IL-17 production (S5B Fig) in mice with Blimp1-deficient T cells. In addition, L. donovani-infected mice lacking Blimp-1 expression in FoxP3+ T (Treg) cells (Prdm1ΔTreg mice) showed none of the above changes (Fig 5). Thus, lack of Blimp-1 expression by conventional CD4+ T cells dramatically improved anti-parasitic immunity, but also promoted tissue pathology. We also observed relatively minor changes in the myeloid cell compartments of mice with Blimp-1-deficient T cells, relative to littermate controls (S6 Fig), most notably, an increased frequency of DC’s in the spleen at day 7 p. i. , but decreased frequency in the liver. To test whether Blimp-1-dependent IL-10 production by T cells was responsible for inefficient control of parasite growth, we infected Il-10fl/fl x Lck-Cre (Il-10ΔT) mice [59] that lacked IL-10 production by T cells with L. donovani. Similar to mice lacking Blimp-1 expression in T cells, control of parasite growth was dramatically improved, relative to Cre-negative (Il-10fl/fl) litter mate controls (Fig 6A). As indicated above, the improved parasite clearance in the absence of Blimp-1 in T cells was associated with more severe splenomegaly. IL-10 has previously been found to protect against tissue damage caused by parasite-mediated inflammation [23,24,60], and we show that in L. donovani-infected mice, increased spleen size was a consequence of a lack of IL-10 production by T cells (Fig 6B). Splenomegaly in mice lacking either Blimp-1 or IL-10 expression in T cells was also associated with a dramatic loss of splenic marginal zone macrophages (MZM) by day 14 p. i. (Fig 6C). Direct IL-10 signaling to myeloid cells contributed to the above phenotypes because mice lacking IL-10 receptor (IL-10R) expression specifically on these cells (IL-10Rfl/fl x LysM-Cre (Il-10rΔM) ) [61] and infected with L. donovani displayed dramatically improved control of parasite growth, which was again accompanied by splenomegaly, increased TNF and IFNγ production, and accelerated loss of MZM, relative to Cre-negative (Il-10rfl/fl) litter mate controls (Fig 6D–6G). Therefore, Blimp-1-dependent IL-10 produced by CD4+ T cells acts on myeloid cells, including MZM, to impair parasite killing, but also acts to limit splenomegaly. We previously showed that following the establishment of chronic L. donovani infection, TNF mediates the loss of MZM in the spleen, associated with severe disruption of lymphocyte trafficking [51]. Indeed, the accelerated development of splenomegaly and loss of MZM in mice lacking Blimp-1 expression in T cells was associated with disrupted cell trafficking into the spleen, which could be rescued by TNF blockade (Fig 7B–7E). Specifically, MZM were retained following TNF blockade, and this was associated with improved retention of injected, fluorescently-labelled lymphocytes in the T and B cell zones of the white pulp regions of the spleen (Fig 7E and 7F). Interestingly, despite the loss of MZM in mice lacking Blimp-1 expression in T cells, T and B cell zones were largely preserved at day 14 p. i. (Fig 7E and 7F). Importantly, improved parasite control in the absence of Blimp-1 in T cells was also dependent on TNF (Fig 7A), indicating that both beneficial and pathogenic effects of TNF were controlled by Blimp-1-regulated IL-10 production by T cells. While these data demonstrate that TNF blockade can prevent tissue damage during infection, they also highlight the importance of TNF for controlling parasite growth. However, in accordance with previous results [62], established infection could be controlled and anti-parasitic immunity maintained when TNF was blocked when mice were also treated with anti-parasitic drug (S7 Fig), thereby identifying a strategy for controlling TNF-mediated pathology while also controlling parasite growth. As outlined above, improved parasite control and exacerbated tissue pathology in mice lacking Blimp-1 or IL-10 expression in their T cells was strongly associated with increased numbers of Th1 cells and serum IFNγ levels (Fig 4). We therefore examined how IFNγ production influenced TNF-mediated consequences of L. donovani infection described above. Strikingly, despite uncontrolled hepatic parasite growth, as reported in IFNγ-deficient mice [63], mice lacking IFNγ receptor (IFNγR) showed no splenomegaly and complete preservation of MZM after 28 days of infection when MZM were lost in C57BL/6 controls (Fig 8A–8D). Critically, these mice produced minimal amounts of TNF, relative to wild type controls (Fig 8E). Thus, our results show that following L. donovani infection, IFNγ promoted TNF production, and this pathway was regulated by Blimp-1-mediated IL-10 production by T cells. Importantly, this regulatory pathway determined the balance between control of parasite growth and TNF-mediated pathology.
Here we show that Blimp-1 mRNA was elevated in CD4+ T cells from VL patients, along with IL-10 mRNA and elevated levels of plasma IL-10. Furthermore, in an experimental model of VL, Blimp-1-dependent IL-10 produced by Tr1 cells acted on myeloid cells to limit parasite killing, but was critical to prevent TNF-mediated tissue disruption. In the absence of IL-10 production by T cells, MZM were lost and this was associated with disrupted lymphocyte trafficking and splenomegaly. Thus, we have identified Tr1 cells as potent suppressors of anti-parasitic immunity, but critical regulators of IFNγ-dependent, TNF-mediated pathology. The transcriptional regulator Blimp-1 is important for IL-10 production by both Treg [64] and Tr1 [52,53] cells. However, Blimp-1 deficiency in Treg cells has minimal impact on immune responses and disease outcome in mice infected with L. donovani. Data from malaria [30,31] and VL [29] patients indicates that Tr1 cells are a major regulatory T cell population during protozoan diseases, and our results show a critical role for Blimp-1 in Tr1 cell function by promoting IL-10 production. Furthermore, results from mice that lack IL-10 in the T cell compartment indicate that IL-10 is a critical immune mediator being controlled by Blimp-1 following L. donovani infection. Another striking feature in mice lacking Blimp-1 expression in T cells was the increase in number and frequency of activated CD4+ T cells. Earlier work showed that Blimp-1-deficient T cells do not have an intrinsically better ability to proliferate [65,66], but they are more resistant to activation-induced cell death [65]. In addition, Blimp1 is an important repressor of T follicular helper (Tfh) cell development by suppressing Bcl6 [67]. Thus, one potential explanation for increased CD4+ T cell activation in our disease models is less cell death, as well as cell differentiation favoring Th1 cell development. The control of many intracellular infections requires Th1 cells which are generated in response to macrophage and dendritic cell derived IL-12 [9]. These Th1 cells produce IFNγ and TNF to activate phagocytes and kill intracellular pathogens [10]. However, if pathogens persist, there is a danger that these pro-inflammatory cytokines will damage tissue, and consequently, Tr1 cells develop to control this inflammation [12,33–35]. Our results suggest that Tr1 cells generated in our experimental setting derive from Th1 cells, a notion supported by recent results showing IL-12, along with IL-27, promoted Blimp-1-dependent IL-10 production by Tr1 cells [52]. Thus, our data suggest that IL-12 is not only required to generate Th1 cells following L. donovani infection, but also provides additional signals for the transition from Th1 to Tr1 cells in this model. Previous work by others has already identified an important role for IL-27 in Tr1 cell development in experimental malaria [23,68] and leishmaniasis [56,69]. TNF is involved in the pathogenesis of a range of diseases, including infectious and autoimmune diseases, and more recently, complications arising in immune-related adverse events as a consequence of immune check point inhibition [3,70]. Our results identify TNF as a major mediator of tissue pathology in the absence of T cell-derived IL-10. Hence, understanding how TNF is regulated in different disease settings is of major medical importance. Despite differences in the structure of rodent and human spleens [38], post-mortem studies on VL patients revealed extensive disruption to white pulp areas, associated with substantial changes in macrophage populations [47,48], also reported in experimental VL [50,51]. The MZ of the spleen is a specialized collection of cells separating the predominantly non-lymphoid red pulp regions and lymphoid dominated white pulp regions. It is also vascular and plays an important role in removing particulate antigen, as well as dead and dying cells, from the circulation [38,71]. Remarkably, the accelerated loss of MZM in mice lacking Blimp-1 or IL-10 expression in T cells resulted in significant disruption of lymphocyte trafficking into T and B cell zones of the white pulp that could be rescued by TNF blockade. These results support earlier studies showing important roles for MZM in directing lymphocyte traffic into the splenic white pulp [72,73]. Another striking feature of L. donovani-infected mice lacking Blimp-1 or IL-10 expression by T cells was dramatic splenomegaly. Angiogenesis is a dominant feature of splenomegaly during experimental VL [44], driven by inflammation-induced expression of neurotropic receptor on vascular endothelium and interactions with ligands produced by mononuclear phagocytes [43]. TNF produced by macrophages can promote angiogenesis [74,75], and chronic inflammation and vascular remodelling are intimately linked in autoimmune disease settings [76]. Critically, inhibition of vascularization with receptor tyrosine kinase inhibitors in mice with established L. donovani infection resulted in reduced mononuclear phagocyte number and reversed splenomegaly [44]. Therefore, our data supports a model whereby TNF-driven angiogenesis is regulated by Blimp-1-dependent IL-10 production by Tr1 cells. By identifying IL-10 produced by Tr1 cells as critical regulators of TNF, we can consider IL-10-related strategies for modulating TNF production. However, given the important role of TNF in controlling intracellular infections [62,77], these strategies should be designed to limit pathogenic functions of TNF while maintaining anti-microbial effects. TNF is produced by many different cell populations, and acts on a range of target cells [78]. We previously reported that TNF from CD4+ T cells was required for anti-parasitic functions in L. donovani-infected mice [77], but the cellular source of pathogenic TNF is not known. In ulcerative colitis patients, TNF from T cells appeared to be an important driver of disease [79], while in graft versus host disease, both T cell [80] and macrophage/monocyte-derived TNF cause gastrointestinal damage [81]. Therefore, different cellular sources of TNF are likely to be important for pathogen control and promoting disease in different immune environments. Further work is needed to better understand the cellular and molecular aspects of TNF regulation that will allow selective targeting of these two distinct functional outcomes of TNF biology. Treg cells are currently being generated and tested for a range of inflammatory conditions [82,83]. However, in situations where this approach fails, the use of Tr1 cells may be beneficial [84]. Our data indicate that Tr1 cells play a critical role in regulating inflammation in organs such as the spleen, in contrast to inflammation in the lung or gut, where Treg cells play critical protective roles [18,85,86]. Hence, under different clinical situations, either Treg or Tr1 cells may help to treat disease. Our findings identify Blimp-1-dependent IL-10 produced by Tr1 cells as a critical regulator of IFNγ-dependent, TNF-mediated tissue damage in the spleen in parasitic infections. Thus, Tr1 focused therapy may be an attractive modality in settings where TNF-mediated immunity is needed, but TNF-induced immunopathology instead dominates.
All patients presented with symptoms of VL at the Kala-azar Medical Research Center (Muzaffarpur, Bihar, India). VL diagnosis was confirmed either by the microscopic detection of amastigotes in splenic aspirate smears or by rk39 dipstick test. Patients were treated either with Amphotericin B or Ambisome. In total, 10 patients were enrolled in the study. The use of human subjects followed recommendations outlined in the Helsinki declaration. Written informed consent was obtained from all participants and/or their legal guardian when under 18 years of age. Ethical approval (Dean/2011-12/289) was obtained from the ethical review board of Banaras Hindu University (BHU), Varanasi, India. All animal procedures were approved by the QIMR Berghofer Medical Research Institute Animal Ethics Committee. This work was conducted under QIMR Berghofer animal ethics approval number A02-634M, in accordance with the “Australian Code of Practice for the Care and Use of Animals for Scientific Purposes” (Australian National Health and Medical Research Council). Female C57BL/6J mice, 8–12 weeks old were purchased from the Australian Resource Centre (Canning Vale, WA, Australia) and the Walter and Eliza Hall Institute (Melbourne, VIC, Australia). Prdm1fl/fl, Prdm1ΔT, Prdm1GFP/+ Il-10fl/fl, Il-10 ΔT, Il-10rfl/fl and Il-10r ΔM mice were bred in-house under specific-pathogen free conditions. Prdm1ΔTreg mice were bred at the Walter and Eliza Hall Institute. All Prdm1fl/fl [54] and Prdm1GFP/+ [55] mice were on a pure C57BL/6 background, while Il-10fl/fl [59] and Il-10rfl/fl [61] mice were backcrossed to C57BL/6 for at least 10 generations. Leishmania donovani (LV9) parasites were maintained by passage in B6. Rag1-/- mice and amastigotes were isolated from the spleens of chronically infected mice. Mice were infected with 2 x 107 LV9 amastigotes intravenously (i. v.) via the lateral tail vein. Spleen and liver impression smears were used to determine parasite burdens and were expressed as Leishman Donovan Units (LDU; number of amastigotes per 1000 host nuclei multiplied by the organ weight (in grams) ). Plasmodium chabaudi chabaudi AS (PcAS) and Plasmodium berghei ANKA (PbA) strains were used in all experiments after one in vivo passage in a C57BL/6 mouse. All mice received a dose of 105 pRBCs i. v. via the lateral tail vein. Thin blood smears from tail bleeds were stained with Clini Pure- stains (HD Scientific Supplies, Willawong, Australia). Parasitemia was used to monitor the course of infection and was determined by flow cytometry (see below). Briefly, 1–2 drops of blood from a tail bleed was diluted and mixed in 250 μl RPMI/PS containing 5 U/ml heparin sulphate. Diluted blood was stained simultaneously with Syto84 (5 μM; Life Technologies, Mulgrave, Australia) to detect RNA/DNA and Hoechst33342 (10 μg/ml; Sigma-Aldrich, Castle Hill, Australia) to detect DNA for 30 minutes at room temperature, protected from light. 2 ml RPMI/PS was added to stop the reaction, and samples were immediately placed on ice until acquisition on a BD FACSCanto II Analyzer (BD Biosciences, Franklin Lakes, NJ). Data was analysed using FlowJo software (Treestar), where pRBC were readily detected as being Hoechst33342+ Syto84+, with lymphocytes excluded on the basis of size, granularity, and higher levels of Hoechst33342/Syto84 staining compared with pRBCs. Heparinized blood was collected from patients before and 28 days after commencement of drug treatment, and PBMC were isolated by Ficoll-Hypaque (GE Healthcare, NJ) gradient centrifugation and used for the positive selection of CD4+ T cells using magnetic beads and columns (Miltenyi Biotech, Bergisch Gladbach, Germany). Cells were collected directly into RNAlater (Sigma), and stored at -70°C until mRNA isolation and analysis. Total RNA was isolated using RNeasy mini kits and QiaShredder homogenizers (Qiagen, Valencia, CA), according to the manufacturer’s protocol. The quality of RNA was assessed by denaturing agarose gel electrophoresis. cDNA synthesis was performed in 20 μL reactions on 0. 5–1. 0 μg RNA using High-Capacity cDNA Archive kit (Applied Biosystems, CA, USA). Real-time PCR was performed on an ABI Prism 7500 sequence detection system (Applied Biosystems) using cDNA-specific FAM–MGB labelled primer/probe for PRDM1. The relative quantification of products was determined by the number of cycles over 18S mRNA endogenous control required to detect PRDM1 gene expression. Spleens were processed through a 100μm cell strainer in order to obtain a single-cell suspension. Splenocyte cell suspensions were then counted and adjusted to a concentration of 2 x 106 cells/ml. LV9 amastigotes (fixed in 4% PFA) were thawed and washed in RPMI media containing Penincillin-Streptomycin and then counted and adjusted to a final concentration of 4 x 107/ml. Cells and parasites were plated into a 96-U well plate at a 1: 20 ratio, where each well contained 1 x 105 cells and 2 x 106 parasites. Cells were cultured in the presence of antigen for a period of 24 and 72 hours. Culture supernatants were harvested at 24 and 72 hours and intracellular cytokine staining was performed at both time points. A transgenic PbA line (231c11) expressing luciferase (PbA-luc) and GFP under the control of the ef1-α promoter [87] was used for all PbA experiments. PbA-infected mice were monitored and scored, as previously described [88]. The in vivo imaging system 100 (Xenogen, Alameda, CA) was used to detect the level of bioluminescence as a measure of whole body parasite burden in each mouse. At selected time-points, PbA-luc-infected mice were anaesthetised with isofluorane and injected with 150 mg/kg i. p. of D-luciferin (Xenogen) 5 minutes prior to imaging. Bioluminescence was measured in p/s/cm²/sr using Living Image (Xenogen), as previously described [88]. For IL-12 neutralisation experiments, mice were administered 500 μg of rat IgG (Sigma) or anti-IL-12 (clone: C17. 8; BioXcell; West Lebanon, NH) i. p. , on the day of infection and every 3 days post infection until day 14 p. i. For TNF blockade experiments, mice administered 200 μg of Human Normal Immunoglobulin (INTRAGAM P; CSL, Melbourne, Australia) or anti-TNF (Enbrel; Amgen, Thousand Oaks, CA) i. p. , on the day of infection and every 2 days post infection until day 14 p. i. . Mice were injected with 100 μg i. v. of FITC dextran (Life Technologies, Melbourne, Australia) one day prior to collection of organs. Spleen tissue was collected into 4% PFA, incubated at room temperature for 1-2hrs and then transferred to a 30% sucrose solution (in MilliQ water) (Sigma, Sydney, Australia) overnight at 4°C. Fixed spleen tissue was then preserved in Tissue-Tek O. C. T. compound (Sakura, Torrance, CA). Splenic architecture and distribution of marginal zone macrophages (MZMs) were analysed in 20 μm sections counter-stained with DAPI and visualized on the Aperio FL slide scanner. Image analysis was performed using Image Scope to determine area of the sections and Metamorph 7. 8 (Integrated Morphometry analysis tool; Molecular Devices, Sunnyvale, CA) to count the MZMs. In some experiments, mice were injected intravenously with 2 x 107 naïve splenocytes labelled with Cell Trace Far Red DDAO-SE (Life Technologies, Mulgrave, Australia), 2 hours prior to sacrifice. The same 20 μm sections were used to assess cell trafficking, where sections were stained with CD3 biotin (5 μg/ml) + SA AF594 (5 μg/ml), B220 PE (5 μg/ml) (Biolegend, San Diego, CA), counter-stained with DAPI (1: 25000, Sigma-Aldrich, Castle Hill, Australia) and mounted with Pro-Long Gold anti-fade (Life Technologies). Slides were visualized on a Carl Zeiss 780 NLO laser scanning confocal microscope under 10x magnification. Image analysis was performed using Metamorph 7. 8 (Counting App and Region Measurement tool). Briefly, since each image was identical in size (1416. 30μm x 1416. 30μm), the total number of cells was counted in each image. The drawing tool was used to delineate the T and B cell zones in the white pulp (WP) of the spleen and the area of these zones in each image was measured (in mm2). Number of cells in WP per mm2 was calculated as the cumulative number of cells in WP in each image divided by cumulative areas of WP in each image. Allophycocyanin-conjugated anti–IFNγ (XMG1. 2), anti-IL-10 (JES5-16E3), PE-conjugated TNFα (MP6-XT22) allophycocyanin-Cy7–conjugated anti-CD4 (GK1. 5), FITC-conjugated anti-CD11a (M17/4), Brilliant Violet 421–conjugated anti–IFNy (clone XMG1. 2), allophycocyanin-Cy7–conjugated anti-NK1. 1 (clone PK136), Alexa Fluor 700–conjugated anti-CD8a (clone 53–6. 7), biotinylated-CD49d (R1-2), PeCy7-conjugated Strepavidin and PerCP-Cy5. 5–conjugated anti–TCRb-chain (H57-597), Allophycocyanin-conjugated anti–anti-CD11c (N418), Pacific Blue-conjugated anti-MHCII (I-A/I-E) (M5/114-15. 3), PerCP-Cy5. 5–conjugated anti-CD11b (M1/70), FITC-conjugated anti-Ly6C (HK1. 4), PeCy7-conjugated anti-F4/80 (BM8) Brilliant Violet 605-conjugated anti-TCRβ (H57-597), allophycocyanin-Cy7–conjugated anti-B220 (RA3-6B2) and Alexa Fluor 700-conjugated Strepavidin were purchased from BioLegend (San Diego, CA) or BD Biosciences. Dead cells were excluded from the analysis using LIVE/DEAD Fixable Aqua Stain or LIVE/DEAD Fixable Near Infra-Red Stain (Invitrogen-Molecular Probes, Carlsbad, CA), according to the manufacturer’s instructions. The staining of cell surface antigens and intracellular cytokine staining were carried out as described previously [89,90]. FACS was performed on a FACSCanto II or LSRFortessa (BD Biosciences), and data was analyzed using FlowJo software (TreeStar). Gating strategies used for analysis are shown in Figs 1 and 2. Cytokine levels in the serum and culture supernatants were measured using a BD Cytometric Bead Array (CBA) Flex sets and the HTS system plate reader on the Fortessa 5 Flow cytometer (BD Biosciences) according to the manufacturer’s instructions. Comparisons between two groups were performed using non-parametric Mann-Whitney tests in mouse studies and Wilcoxon matched-pairs signed rank test in human studies. Comparisons between multiple groups were made using a Kruskal-Wallis test and corrected using Dunn’s multiple comparisons test. GraphPad Prism version 6 for Windows (GraphPad, San Diego, CA) was used for analysis; p<0. 05 was considered statistically significant. All data are presented as the mean ± SEM. | Many parasitic diseases are associated with the generation of potent inflammatory responses. These are often needed to control infection, but can also cause tissue damage if not appropriately regulated. IL-10 has emerged as an important immune regulator that protects tissues by dampening inflammation. Recently, some T cells that initially produce inflammatory cytokines have been found to start producing IL-10 as a mechanism of auto-regulation. We identified an important transcriptional regulator called B lymphocyte-induced maturation protein 1 (Blimp-1), which promotes IL-10 production by IFNγ-producing CD4+ T (Tr1) cells during malaria and visceral leishmaniasis, two important diseases caused by protozoan parasites. We found that Tr1 cell-derived IL-10 suppressed anti-parasitic immunity, but played a critical role in preventing tissue damage caused by the potent pro-inflammatory cytokine TNF. Specifically, IL-10 protected macrophages from TNF-mediated destruction, and this enabled lymphocytes to continue to migrate to regions in the spleen where T and B cell responses are generated. These findings allow us to better understand how parasites persist in a host, but also identify new opportunities to control inflammation to prevent disease. | Abstract
Introduction
Results
Discussion
Methods | 2016 | Blimp-1-Dependent IL-10 Production by Tr1 Cells Regulates TNF-Mediated Tissue Pathology | 8,941 | 282 |
|
Control of axial polarity during regeneration is a crucial open question. We developed a quantitative model of regenerating planaria, which elucidates self-assembly mechanisms of morphogen gradients required for robust body-plan control. The computational model has been developed to predict the fraction of heteromorphoses expected in a population of regenerating planaria fragments subjected to different treatments, and for fragments originating from different regions along the anterior-posterior and medio-lateral axis. This allows for a direct comparison between computational and experimental regeneration outcomes. Vector transport of morphogens was identified as a fundamental requirement to account for virtually scale-free self-assembly of the morphogen gradients observed in planarian homeostasis and regeneration. The model correctly describes altered body-plans following many known experimental manipulations, and accurately predicts outcomes of novel cutting scenarios, which we tested. We show that the vector transport field coincides with the alignment of nerve axons distributed throughout the planarian tissue, and demonstrate that the head-tail axis is controlled by the net polarity of neurons in a regenerating fragment. This model provides a comprehensive framework for mechanistically understanding fundamental aspects of body-plan regulation, and sheds new light on the role of the nervous system in directing growth and form.
Humanity’s inability to regenerate significant portions of anatomy (e. g. lost limbs, severed spinal cord, damaged organs) has prompted decades of intensive study into organisms that can. The planarian flatworm is one such model organism, as it can completely regrow from a piece as small as ~1/250th of the original worm [1,2]. Crucially, this regenerating organism exhibits complex behaviors and properties functioning well beyond the level of a single cell. It can maintain body-plan homeostasis under normal circumstances, including body-wide allometric remodeling during growth and shrinkage depending on availability of food, while also detecting injury to precisely reproduce missing features, and stopping regeneration once the correct body structure (the target morphology) has been restored [3,4]. Therefore, a deep and functional understanding of regeneration requires, not only an account of single cell activities such as gene expression [5], but also of the regenerating organism as an intricate system exhibiting complex dynamics and coordination over multiple levels of scale. How do patterning control systems scale to enable correct regeneration of fragments of vastly different sizes and shapes? What tissue-level properties underlie long-range coordination of anatomical outcomes? What is the role of the nervous system in regeneration? All of these open questions have major implications, not only for evolutionary and developmental biology, but also for implementing rational repair strategies in the regenerative medicine of complex organs. Body-plan consists of precisely-organized collectives of different cell types, where cells receive cues to proliferate and differentiate by responding to morphogens, including molecular-genetic [6–8], mechanical [9–11], and bioelectrical [12–14] signals. Therefore, spatial patterns of morphogens can serve as instructive pre-patterns to induce changes to single cell identity and ultimately specify aspects of the final body-plan [6,15–18] (Fig 1). Planaria maintain a population of pluripotent adult stem cells (neoblasts), which migrate to wounds to form a blastema, where subsequent proliferation and differentiation ultimately regenerates lost portions of head, trunk, and tail [19–22]. Neoblasts respond to morphogenetic gene products [19,20,22–24] and chemical messengers [25–27] to produce the needed cell types to transition the blastema at the former anterior portion of a wound into a new head, while creating tails at wound sites facing the original posterior. For whole planaria in morphological homeostasis, factors such as β-Catenin (β-Cat), Wingless/Integrated (Wnt), extracellular-regulated receptor kinase (ERK), Notum, as well as other position control genes (PCGs), are found as concentration gradients with characteristic polarities with respect to the anterior-posterior axis of the worm [18,24,28–35]. Remarkably, during early regeneration (from 15 to 72 hours after injury), these same morphogens spontaneously reform their original polarity in each fragment of a worm cut into multiple pieces [24,28–34]. Consistent with the concept of positional information regulating the body-plan, various heteromorphoses can be generated through experimentally-applied genetic and pharmacological interventions, which act by influencing levels of morphogens involved in anterior-posterior axis control [32,36,37] (Fig 2, Table 1). A major knowledge gap exists in determining how wound-induced repolarization of morphogens occurs in each fragment of a cut planaria. Reaction-diffusion mechanisms (e. g. diffusing proteinaceous transcription products or other signaling factors that regulate each other’s expression and activity levels), have long been proposed to account for self-assembly of molecular pre-patterns in a variety of biological systems [6,15,17,43,44], including the gradients observed in planaria homeostasis and regeneration [45–47]. However, the dependence of reaction-diffusion mechanisms on the size of a biological system—which is an intrinsic feature of most reaction-diffusion models—may limit their ability to account for chemical gradients in planaria fragments of diverse size [48,49]. Specifically, for a system of interacting molecules with fixed rates of diffusion and other physically-constrained factors, a gradient may spontaneously develop for a system at a certain size, but this pattern will drastically change (e. g. a categorical change in pattern type from smooth gradient to stripes or spots) when the system size increases [48,49]. This scale-dependence of reaction-diffusion schemes can be mitigated by incorporating directional transport, here termed vector transport, of substances into the reaction-diffusion scheme, which has been shown to robustly account for self-assembly of concentration gradients in a manner essentially independent of size scale [48,50,51]. The complex nature of emergent outcomes driven by such processes underscore the essential need to simulate models of regeneration in a spatialized framework to fully determine their large-scale patterning properties, extract testable predictions, and ultimately develop strategies for rational modulation of outcomes in biomedical and synthetic biology settings. Here we present, and quantitatively analyze, an inclusive model of the molecular regulation underlying planarian regeneration, synthesizing much published work on the results of physiological and molecular-genetic manipulations (Tables 1 and 2). Our model combines a molecular signaling network at the cellular level with vector transport of morphogens (e. g. the directional transport of morphogens by a vector field), and correctly predicts regenerative patterning outcomes in a variety of settings, including editing of regenerative pattern without use of molecular or genetic manipulations. Our model also explains several outstanding puzzles in planarian regeneration, such as how self-assembling, self-scaling morphogen gradients can form in pieces of different size and location across the worm; the mechanistic origin of various heteromorphoses (e. g. one-headed, two-headed, missing-head or missing-tail outcomes) under a range of pharmacological and RNAi interventions; and the instructive role of the nervous system in regeneration. We also tested a number of unique predictions of this model experimentally (Table 2). Hypothesizing that vector transport can take place on the aligned microtubules of nerves enabled the experimental discovery of the crucial role for nerve directionality in determining the AP polarity of fragments, including the re-setting of pre-existing polarity. We show how cooperation of biophysical and molecular-genetic processes can functionally link single cell physiology and long-range anatomical patterning. Importantly, while this work explores mechanisms enabling body-plan regeneration and target morphology in the context of planarian regeneration (Fig 1), the underlying concepts reveal a powerful, multi-scale patterning system that is applicable to a wide range of biological systems and contexts, such as embryogenesis, regeneration, cancer, and the bioengineering of organoids [52,53].
Our model is hierarchical and multifactorial in nature, connecting several major concepts. Details of our theory, model, and software are described in S1 Text. The planaria models presented herein were formally implemented and explored using the Planarian Interface for Modeling Body Organization (PLIMBO), a 1D and 2D finite volume method simulator written in Python3 with open-source tools utilized from Scipy, Numpy, Matplotlib, Scikit-learn [59,60], and BETSE [61,62]. PLIMBO was developed specifically to study all aspects of the planaria regeneration modeling reported in this manuscript. PLIMBO allows quantitative testing of the behavior of the regulatory network model reported on herein, in both 1D and 2D contexts, and under a range of experimental perturbations, with vector (convective) transport of morphogens on imported nerve polarity vector fields, as well as the extraction of novel testable predictions. PLIMBO also has the capacity to run parameter searches to automatically iterate parameters to assist model parameterization, to perform sensitivity analyses, and to perform scaling analysis of the model (where a body-shape is progressively scaled); these tools all assist in the development and exploration of complex biological models. PLIMBO is freely available from: https: //gitlab. com/betse/plimbo. Our planarian regeneration model requires the biophysical mechanism of vector transport to create virtually scale-free self-assembly of morphogen gradients instructive for planarian body-plan homeostasis and regeneration, a premise which extends conventional reaction-diffusion mechanisms by adding a convective transport term, the fundamental properties of which have been explored elsewhere [48,50,51]. Here we further propose transport of morphogens on polarized microtubules of nerve axons distributed throughout the planarian tissue (i. e. not just the ventral nerve cords, VNC) as the specific vector transport mechanism (Fig 1B, 1C and 1D). Vector transport of Hh on the VNC has been previously hypothesized and qualitatively reported, but not subjected to a formal analysis to determine feasibility and predictions [30,31,39]. Experimental investigations into neuronal polarity and axoplasmic transport networks of planaria [63] are reportedly very difficult undertakings, and have produced results that are not useful for the scope required by this present work. Therefore, we estimated maps of nerve axon polarities (Fig 1B), which were derived by manually tracing synapsin stains to determine approximate axon angle at various locations in the planaria tissue cross-section (Fig 1A). The manual traces were assumed to be vector fields supporting directional transport of morphogens (Fig 1B). A summary of all nerve map models is shown in S1 Fig. Planaria of the species Dugesia japonica were used in all experiments. Planaria were maintained as described in [77]. Animals were kept at 13 °C in Poland Spring water with weekly feedings of calf liver paste and twice-weekly cleanings. Animals were starved for at least 7 days before experiments were performed. Planaria were cut as described in Nogi et al [78], using a scalpel on moist and cooled layers of filter paper. 2H worms were generated via 3 day incubation of cut fragments in 127 μM octanol as described in [79,80]. A 2H worm was characterized by at least one eye on each end of the worm [79]. A variety of cutting/amputation scenarios were tested, each of which were designed to cut the proposed neural transport map in key geometries to influence outcomes according to the main hypothesis. In order to explore the role of proposed VNC cord polarity in directing regeneration outcomes, a nerve-deviation experiment inspired by the work of Kiortsis in Mediterranean Fanworm (Spirographis spallanzanii) [81], was designed. Trunk fragments of planaria were excised and immediately further cut using upwards or inverted L-shaped incisions. Specifically, fragments were cut perpendicular to the head-tail axis halfway from margin to midline of the worm and further cut along the midline towards either the head (upwards) or tail (inverted), causing tissue movement exposing the VNC in a forward- or reverse- polarity, with N = 43 and N = 46 replicates in upwards and inverted L-cut groups, respectively. The cuts were re-enforced every day for 7 days while worms regenerated at 20 °C, to prevent simple wound healing. All worms were monitored once a day for 14 days, with specimens reserved from each group for synapsin staining every other day for 11 days. Computational modeling using the same upwards and inverted L-shaped incisions on planaria models was performed to obtain predictions for these cutting scenarios. Control L-cuts were made similar to the L-cuts above, but not cutting so far from the margin to the midpoint as to including the VNC, thereby creating an exposed side area free of VNC but attached at the base to the main worm, with N = 30. To explore the role of net nerve alignment in fragments, rectangular fragments with the ~1 mm long-edge corresponding to the original anterior-posterior worm axis (thereby allowing for identification of axis-polarity in the regenerates) were cut to include the ventral nerve cord (VNC-containing) or from the side margin of the worm, excluding the VNC (VNC-free), with N = 94 and N = 103 replicates in the VNC-free and VNC-containing groups, respectively. The fragments were cut from the margin of 1H worms, with the worm placed ventral side up, which allowed the VNC to be seen and to thereby be avoided in the VNC-free fragments. The fragments were checked 1 day after cutting to ensure that all 4 sides were cut. Fragments that did not fulfill this condition were discarded. All fragments were monitored once a day for 14 days, with additional specimens from each group reserved for fixation and synapsin staining at days 1 through 7,10 and 14. The net orientation of the synapsin signal was determined using the Directionality plugin in ImageJ. Specifically confocal stacks of the whole fragments with a z-resolution of 2 μm were z-stacked using the average intensity projection in ImageJ and oriented so that the long edge of the fragment was vertical in the images. Fragments in which no long edge could be determined were not analysed. The Directionality plugin was run using the Fourier component method, 90 bins and a histogram start at 0. Resulting data was averaged for all fragments, binned into 10° sections and plotted using Excel. Brightfield images and respective synapsin stains taken over multiple days and showing further replicates, can be found in S1 and S2 Datasets. To explore regeneration outcomes for different cutting scenarios in two-headed worms, 2H worms were generated as described above, and had completed regeneration at least 1 month prior to subsequent amputation. The 2H worms were cut using either decapitation, or cuts through the central symmetry line of the worm, cuts along the diagonal of the worm [80], or a combination of transverse cuts producing fragments. The number of replicates varied in each cutting scenario, with N = 401 used for short fragments containing a head, N = 90 for medium fragments with head and second head amputation cut not crossing the central symmetry line, N = 140 for fragments with one head and second amputation crossing the central symmetry line, N = 90 for dual head amputation containing the central symmetry line, and N = 41 for fragments with dual head amputation and not containing the central symmetry line. Diagonal cuts of 2H worms were performed through the midpoint of the worm (as estimated by the pharyngial opening) and the base of the two heads (N = 97). Cut fragments were allowed to regenerate for 1 week at 20°C and 1 week at 13°C (to prevent fissioning) before scoring regenerative outcomes. Worms were fixed in Carnoy’s fixative as described in [78], were bleached in methanol, and were subsequently stained using the mouse-anti-synapsin primary antibody 3C11 at 1: 50 dilution. 3C11 (anti SYNORF1) was deposited to the DSHB by [82]. A standard immunohistochemistry protocol was used as outlined in [78], using a goat-anti-mouse-Alexa488 secondary antibody (Invitrogen) at 1: 250. Samples were mounted using VetraShield (Vector Laboratories) and imaged on a Nikon AZ100 Multizoom Macroscope or on a Leica SP8. The dsRNA synthesis was performed as described in [83]. The template used was Dugesia japonica mRNA for β-catenin-B, partial cds (GenBank: AB499795). Each dsRNA injection was performed in the pre-pharyngeal area of 1~1. 5 cm long worms, consisting of three pulses of 32. 2 nL each, for three consecutive days. In order to achieve only a partial block of β-Cat, which was required to test the model prediction, the amputation was performed on the third day of dsRNA injection, at least 6 hours after the injection. Regeneration was at 13°C. As a control, VenusGFP dsRNA was injected, and displayed no phenotype different from uninjected worms. N = 20 replicates were used for experimental and control groups, respectively. Dynein transport function was inhibited by pre-soaking intact animals in 3 μM Ciliobrevin D (Sigma, stock 25 mM in ethanol) for 3 days before cutting and placing trunk, pharynx or pre-tail fragments into fresh 3 μM Ciliobrevin D solution for 3 days before removing the inhibitor and replacing with Poland Spring water. Regeneration was at 20°C and outcomes were scored after 14 days. N = 47 replicates were performed as well as N = 50 controls with matching concentration of ethanol. For all other pharmacological treatments, excised pre-tail fragments were treated immediately after cutting for the first three days of regeneration, after which the regenerating fragments were washed in Poland Spring Water and allowed to regenerate for 10 days at 20 °C before scoring regenerative outcomes. All pharmacological stock solutions were aliquoted and stored at -20 °C. Bromocriptine (Tocris 0427) 50 mM stock solution in DMSO was used at a final concentration of 2 μM, MDL12330A (Tocris 1436) 50 mM DMSO stock solution was diluted to 20 μM, U0126 (Tocris 1144) was used at a concentration of 18 μM, and 3-isobutyl-1-methylxanthine (IBMX) (Sigma I5879) was dissolved in DMSO and used at a concentration of 200 μM. A referenced summary of projected downstream targets for these pharmacological agents is given in Table 3. Bromocriptine treatment was also used on N = 52 worms cut into 5 equal fragments each. To detect the pattern of cilia driven flow, an experimental procedure similar to [56] was used. Planaria were placed ventral side up in a small dish so that they attached to the surface of the water and a small amount of Carmine powder (Sigma) was sprinkled onto their ventral side. The movement of the powder particles was recorded on a Nikon AZ 100M microscope with an Andor DL-604M camera at 100 ms frame rate for 2 mins at a time. Slime and powder build-up was periodically removed. The functionality of the method was controlled by applying the same assay to worms treated with 3% ethanol to block cilia function [85] and worms cooled to 4°C, both of which showed a complete lack of cilia-driven powder movement. To analyze the flow pattern, recordings of 1. 5 to 2 s were selected in which the worm showed no muscle-driven movement and analyzed using the PIVlab application [86,87] in MATLAB (MathWorks). Particle image velocimetry (PIV) was calculated in an interrogation area of 64 x 48 x 32 pixels for subsequent passes for all frames, data was smoothed, and all frames were averaged to result in the shown flow pattern. Ciliary beat was used as an indicator of cellular polarity, as the establishment of long-range order by physiological mechanisms is a key implication of our model. As a polarized microtubule array within individual cells is a required prerequisite for establishment of planar cell polarity [88–92], and cilia beat direction is in turn controlled by planar cell polarity [93], the beat direction of planaria cilia are read-outs of a net microtubule array alignment in individual cells, and ciliary beat direction is widely used as a readout of planar polarization in tissues [40,94–99].
It is crucial to develop tissue simulations of proposed signaling networks to fully test quantitative models of regeneration, ensuring that they produce the correct outcomes for known functional data in the field. Additionally, an existing gap in understanding planarian regeneration is how to describe self-regenerating morphogen gradients independent of size-scale, while maintaining adherence to realistic time-scales and conditions. Using realistic values for biological parameters (see S1 Text) and known regulatory interactions (e. g. the canonical Wnt/β-Cat pathway), we developed a model to address these knowledge gaps, and tested its predictive abilities by simulating diverse experimental conditions and comparing the outcomes, either against published results, or new experimental data, in a quantitatively rigorous manner. The main predictions of the full model, and their state of validation by previously-reported or presently-reported experimental evidence, are summarized in Table 2. The core of the model is a regulatory network including major molecular signals known to be involved in regulation of anterior-posterior axis regeneration in planaria, namely Wnt, β-Cat, ERK, Notum, and Hh (see Table 1). Predicted morphogen signaling gradients were found to be consistent with experimentally-observed planaria body-plan outcomes for homeostasis (Fig 2A) and normal wild-type regeneration (Fig 2A‘o’). We first tested the model’s ability to correctly predict outcomes of experimental manipulations known to alter regenerative outcomes in planaria by simulating 10 different RNAi and pharmacological interventions as perturbations to key components of the regulatory network (Fig 2). Specifically, we modeled RNAi against Wnt, Hh, β-Cat, APC, and Notum, as well as chemical manipulations of ERK, cAMP, dopamine, and serotonin (Fig 2 and Tables 1 and 3). Of these 10 different interventions, cAMP inhibition by MDL12330A and cAMP enhancement by the phosphodiesterase inhibitor IBMX have not been previously reported. Overall, the model generated stable morphogen concentration patterns, which were predictive of body-plan outcomes observed in experiments (Fig 2) for interventions associated with both abnormal posterior regeneration (e. g. head on the posterior to regenerate 2H, Fig 2Ai to 2Aiii), abnormal anterior regeneration (e. g. loss of head or tail on the anterior, Fig 2Av to 2Ax), or loss of tail (Fig 2Aiv). These model predictions, where regenerated body-plan is interpreted in terms of stable morphogen gradients after cutting the model, matched corresponding experimental outcomes, which we compiled from published work from a number of different labs, as well as from new experiments (Table 1 and 3). Details of predicted morphogen gradients and morphological outcomes for all interventions considered in the model are summarized in S3 Fig. The model also explains puzzling observations that have been reported experimentally with regards to Notum [32,34]. In the model, Notum is transcribed at anterior-facing wounds (S4 Fig) and therefore appears as an anterior-polarized gradient under normal regeneration conditions, as is observed experimentally [28,29,32] (Fig 4A). Consistent with this anterior localization, experimental RNAi data shows that Notum is required for head induction [32], which our model reflects as a loss of anterior signals following simulated RNAi against Notum (Fig 2Aix). Paradoxically, the same experimental data [32] shows loss of Notum signal in 2H heteromorphoses, which is also recapitulated in our model (Fig 4B and 4C). Conversely, treatments leading to loss of anterior development (e. g. APC RNAi, S3L Fig) are predicted by the model to show very high Notum levels in spite of producing 0H regenerates—this is further consistent with the previously reported data [34]. The model provides insight into these otherwise paradoxical observations with respect to Notum, showing how they occur due to a confluence of four factors (Fig 1E): (1) expression of Notum is induced by β-Cat (indirectly via β-Cat’s induction of expression of a predicted Notum Regulating Factor (NRF), where NRF induces Notum expression); (2) NRF is subjected to vector transport by dynein (towards the anterior of fragments, and in the opposite direction to transport of Hh); (3) Notum inactivates Wnt at the anterior of regenerating fragments; and (4) β-Cat is ultimately what inhibits ERK signaling and therefore anterior regeneration. Taken together, while Notum is required to decrease levels of Wnt [32,100,101], and thereby induce anterior regeneration under normal circumstances, interventions that are downstream of Wnt signaling will show non-intuitive outcomes with respect to Notum. Manipulations leading to enhanced destruction of β-Cat (including decreases in Wnt and cAMP levels), decrease β-Cat, and therefore increase ERK and anterior regeneration signaling, while reducing levels of Notum. In contrast, as APC (and other elements of the β-Cat destruction complex) signal downstream of Wnt and Notum, RNAi to APC leads to significant increases in β-Cat levels and therefore to a strong inhibition of head formation with concurrent strong increases in Notum levels due to stimulated expression by increased β-Cat. Finally, the model correctly predicts that Notum works in conjunction with Patched (Ptc) to inhibit β-Cat signaling, such that inhibition of Notum or Ptc both lead to 0H or 2T heteromorphoses [39,40] (Fig 2Aviii and 2Aix). Please see S3 Fig, which presents modeling results illustrating these trends. Due to the spatialized nature of our model and the ability to flexibly simulate different physical perturbations, the model can predict regenerative outcomes for fragments of different sizes, shapes and locations within the original animal. This enabled us to extract predictions of differential regenerative outcomes for fragments along the anterior-posterior axis (Fig 4). In the standard case, the model predicts polarized gradients of ERK, β-Cat, and Notum consistent with normal regeneration of single-headed worms in each fragment (Fig 4A). For a partial RNAi block of β-Cat, such that β-Cat is no longer being synthesized but levels have not fully decayed in the system, the model predicts higher levels of morphogen gradient polarity disruption for more anterior fragments (Fig 4B and 4D), which translates to a prediction of higher numbers of 2H regenerates closer to the head. This prediction fundamentally depends on the premise that Hh is synthesized and transported in neural tissue. To test this prediction, we performed a novel investigation of the effects of partial RNAi against β-Cat in changing the incidence of regenerative outcomes of worms cut into 5 fragments along the head-tail axis immediately after 3 consecutive days of RNAi injections. Our experimental observations show the same trend as the prediction of the model, with 100% 2H organisms regenerating from head fragments, and progressively lower incidence of 2H heteromorphoses forming in fragments cut closer to the tail (N = 20), see Fig 4B and 4D. In contrast to the observation of high 2H regeneration frequencies in more anterior fragments for a partial β-Cat RNAi, the model predicts the reverse pattern for perturbations inhibiting cAMP (or any perturbation enhancing degradation of β-Cat in the canonical Wnt/β-Cat signaling pathway) where a higher level of morphogen polarity disruption, and therefore more 2H incidence in a population of regenerates, is predicted in fragments cut closer to the tail (Fig 4C and 4E). We tested this prediction by chemically inhibiting cAMP signaling through the application of Bromocriptine [25,42], and did indeed observe the highest number of 2H in the pre-tail fragment, with very low 2H numbers in the pre-pharynx and head fragments (Fig 4C and 4E), with the experimental results showing the same trend as model predictions. Note that these contrasting trends are not limited to a small sub-set of parameters, but appear to be a natural component of the planaria model, which can also easily be seen in the simpler 1D simulations (see S2C and S2D Fig). In summary, these results indicate that our proposed model is capable of recapturing and accurately reflecting past and novel experimental data, as well as offering explanations for previously puzzling observations. At the same time, the model can make new, non-obvious predictions for the outcomes of genetic and chemical manipulations, especially in combination with complex physical manipulations, some of which we validated here (see Table 2). In addition to the signaling network within the individual cell, our model relies on vector transport of certain signaling factors, specifically Hh and NRF, which were identified as necessary in order to create a model capable of describing outcomes of RNAi experiments and the variety of new experiments reported herein (see Table 2). While the vector transport aspect of the model is compatible with several mechanisms of directional transport involving an array of potential cell types, including transcytosis [102], cytonemes [103], and electrophoretic transfer of small, charged morphogens via gap junctions [48,104], experimental data presented here suggests a role for the nervous system anatomy of planaria in directing regeneration outcomes. We therefore proposed that the directional transport of morphogenic factors occurs with respect to aligned microtubules of the neuronal axons [105–107], where vesicles of Hh are hypothesized to be transported by the motor protein kinesin, while the here-proposed Notum Regulating Factor (NRF) is hypothesized to be moved by dynein. Neuronal transport or expression of these factors has previously been shown for Hh [30,31,39]. To test the importance of cellular motor protein activity for the patterning of the AP axis, we inhibited dynein-based transport through the application of low doses of Ciliobrevin D to regenerating fragments, which did not impact ciliary motion as assessed by gliding behavior. Blocking dynein-based transport led to inhibition of head formation during regeneration (N = 23/47 0H), consistent with the phenotype observed with Notum RNAi (see S3I, S3K and S5 Figs). Cell division was not disrupted by Ciliobrevin D treatment, as the tail blastema regenerated at a normal rate. Although we cannot rule out the possibility that other morphogen-regulating factors are also being transported by dynein, the consistency between the dynein inhibition and Notum RNAi supports the hypothesis that NRF is being transported via dynein towards the anterior blastema where we predict it is responsible for inducing Notum expression and subsequent head formation. As vector transport of morphogens is proposed to be determined by nerve axon polarity, and morphogen concentration is instructive for regenerated body-plan, we predicted that the average (net) polarity of nerves contained in a cut fragment would drive the patterning of the primary axis in head/tail regeneration. We sought to test this prediction by analyzing regeneration outcomes in a variety of different cutting scenarios involving the monopolar and bipolar nerve polarity maps of 1H and 2H heteromorphoses (Figs 5,6 and 7). The high nerve axon density and strong anterior-posterior alignment of axons in the VNC [57] are predicted to lead to VNC dominance of the transport direction in fragments containing the VNC (Fig 5). Therefore, when a 1H worm is cut perpendicular to its head-tail axis, transport along the VNC is predicted to lead to re-establishment of morphogen gradients with their original anterior-posterior polarity in each cut fragment (Fig 1G and 1F), a prediction that has been previously observed in experiments [29,34]. To further test the involvement of VNC, L-shaped cuts were made into the margins of 1H worm trunk fragments, such that the VNC was diverted towards the induced third blastema via cutting, which induced a side-outgrowth containing either a forward-facing or a backward-facing VNC (Fig 5). Dependent on this directionality, the side outgrowths formed either a head when the forward-facing VNC was diverted (N = 46/46), or a tail when the backward-facing VNC was diverted outwards (N = 43/43), see Fig 5. When the L-shaped cuts were performed so that only tissue outside of the VNC was diverted towards a side outgrowth, no head regeneration was observed (N = 25/25), see S6 Fig. Taken together this data indicates that the VNC is important in generating head or tail regenerative outcomes, depending on cut orientation, and that other tissues, such as muscle, which are known to play an important role in patterning in planaria [18,20], are not alone sufficient to determine new head/tail identity in tissues. These data further indicate that directional transport on the VNC has the ability to determine head/tail identity, as predicted by the steady-state morphogen gradients of ERK and β-Cat produced by the model for these same two cutting scenarios (Fig 5G, 5H, 5O and 5P). To be able to functionally test the hypothesis that transport along the vector field formed by the net distribution of nerve axons distributed throughout the planarian tissue is the driver of subsequent patterning events, we sought a test case involving a cut fragment where the average nerve direction was not aligned with the prior head-tail axis and the gradient of morphogens that existed before the injury. Specifically we tested small fragments cut from the margins of the worm between the VNC and the body edge, where synapsin stains indicated a net nerve polarity oriented 90° to the anterior-posterior axis and thereby perpendicular to the original morphogen gradient (Fig 6B and 6C). The regeneration of rectangular fragments cut from this margin area (with cuts on all 4 sides so as to not bias blastema formation, and the long edge of the rectangle corresponding to the original anterior-posterior axis), was tracked over time, as well as observed in detail via a time course of synapsin staining. This revealed that the newly forming head in these fragments developed perpendicular to the original body axis and morphogen gradient and in alignment with the nerve fiber direction (Fig 6C–6F), exactly as predicted by the model (N = 94/94 regenerated 1H). These side pieces very likely contain nerve bodies in addition to the axons oriented 90° to the anterior-posterior axis, as seen by PC2 staining [63] and persistent synapsin staining over multiple days. These nerve bodies are hypothesized, by our model, to be the site of production of Hh and NRF, which are transported along the axons to the long edge of the fragments, perpendicular to the previous axis. In contrast, control fragments of the same size and shape, but containing a piece of the VNC, regenerated following the orientation of the VNC and in accordance with the head-tail orientation of the original tissue (N = 103/103 regenerated 1H), see Fig 6G–6K. Further depictions and quantification of nerve directionality in regenerating fragments for the VNC-free and VNC-containing fragments are shown in S7 Fig, as well as the S1 Dataset. This confirms a highly non-intuitive prediction of our model, given that it is, to our knowledge, the first case where regenerative polarity of a fragment is fundamentally different from that of the original tissue without any applied treatment. This indicates that the morphogen gradient expressed in planaria tissue for various position control genes (PCGs) is not deterministic for regenerative outcomes but rather is set up by the underlying neuronal polarity which determines patterning. The 2H worms feature a non-trivial nerve polarity map (Fig 7G and 7H), which is proposed to exhibit bipolarity associated with each head and the location of the midpoint of the animal (Fig 7, where white arrow shows midpoint). This bipolarity was directly observable using cilia flow assay (Fig 7A–7F and S5 and S6 Videos). Cutting 2H worms with respect to the nerve polarity map lead to experimental regeneration outcomes consistent with modeled morphogen gradients resulting from vector transport on the field derived from nerve polarity for a variety of cutting scenarios (Fig 7I–7N). Experiments showed 0% 2H (0/401) for short fragments containing the head (Fig 7I), 8% 2H (11/140) for longer fragments with head but posterior cut not crossing the symmetry line (Fig 7J), 94% 2H (85/90) for cuts containing the head but posterior cut passing the symmetry line (Fig 7K), 99% 2H (89/90) for amputations of both heads in fragments containing the symmetry line (Fig 7L), and 7% 2H (3/41) for amputation of both heads in fragments that did not contain the symmetry line (Fig 7M). Cutting 2H worms diagonally at angles greater than 45° from the symmetry line regenerated 88% 2H (85/97) in fragments that showed regeneration (Fig 7N). Experimental regeneration outcomes showed statistically significant corroboration with model-predicted outcomes (based on Markov-model body-plan inference from predicted morphogen gradients) for the same fragments (Chi2 test, p < 0. 05). This modeling and experimental data in 2H worms further indicates that transport of morphogens occurs with respect to nervous system polarity, and demonstrates further correct predictions of a wide range of non-intuitive regeneration outcomes. Models were run with planarian body shapes ranging from 0. 3 cm to 2. 4 cm in length, and were found to produce morphogen gradients with similar pattern form and of similar magnitude in models of very different size (S8 Fig). Moreover, cutting these models into 3 fragments showed that all models, except for the smallest, were able to fully re-polarize morphogen gradients (S8 Fig). Small fragments (below approximately 1 mm in length) begin failing to regenerate normal gradient polarities, with 0T and 2H heteromorphoses predicted as fragments become smaller (S8 Fig), an outcome which is experimentally observed for small or thin fragments [108–110]. Taken together, the results of the above-described analyses and new experiments show that our model correctly predicts the outcomes of a range of diverse amputations that probe the relationship between positional information in the intact worm and mechanistic transport processes that set the axial polarity of the resulting fragments. We conclude that regenerated anterior-posterior axial polarity depends on the average polarity of nerve axons contained in a regenerating fragment, and that these are dominant drivers in cases where their direction diverges from the pre-existing directionality of the prior AP axis.
Planaria are an important model for a very fundamental phenomenon: pattern homeostasis, which allows regulative regeneration to restore complex anatomical structures. It is widely acknowledged that the discovery of molecular signaling relationships has far outpaced our understanding of the origin of order and the ability of complex systems to remodel to correct anatomical patterns from diverse starting conditions. Fully quantitative models are an essential step for the field of developmental biology, and are an important emerging tool for rational discovery of interventions in regenerative medicine. The framework presented here is a powerful environment for general use, within which future findings in planarian regeneration and other systems can be predicted and interpreted. The model and simulation environment facilitate the addition of new data in this field to explore self-organizing dynamics that cannot be inferred from arrow diagrams alone, and make novel predictions about the large-scale behavior of molecular pathways. Our model contains a regulatory network which recapitulates a majority of published genetic and pharmacological manipulations, and makes new predictions for regenerative outcomes for a wide variety of experimental manipulations, both genetic, chemical and physical, which we successfully validated here. Furthermore, the modeling of neuronally-mediated vector transport of morphogens was validated in new experiments indicating the importance of dynein-based transport in morphogen localization as well as showing the reorientation of head-tail polarity in fragments, according to their nerve polarity, without any treatment. Even more importantly, our model and its analysis are proposed as a proof-of-principle effort, applicable potentially to a wide range of patterning contexts, as a first step towards an integrated modeling system within which complex, emergent dynamics can be explored. | Understanding how large-scale anatomy emerges from the activity of cellular pathways is a key goal of evolutionary developmental biology. Elucidating the rules of body-wide morphogenesis is especially essential for transitioning molecular signaling data at the cellular level into advances in regenerative biomedicine. We constructed and analyzed a comprehensive, multiscale computational model to explain the determination of axial polarity during planarian regeneration. Uniquely, our model explains the various head-tail patterning outcomes of a wide range of molecular and physiological manipulations. Testing the novel predictions of this model revealed the nervous system as an instructive regulator of axial patterning. | Abstract
Introduction
Methods
Results
Discussion | invertebrates
rna interference
cell processes
animals
nerve regeneration
head regeneration
developmental biology
organism development
network analysis
molecular development
epigenetics
morphogenesis
tail regeneration
planarians
computer and information sciences
genetic interference
gene expression
flatworms
regulatory networks
axonal transport
morphogens
biochemistry
rna
eukaryota
cell biology
nucleic acids
regeneration
genetics
biology and life sciences
organisms | 2019 | Neural control of body-plan axis in regenerating planaria | 9,923 | 145 |
Graph theory has evolved into a useful tool for studying complex brain networks inferred from a variety of measures of neural activity, including fMRI, DTI, MEG and EEG. In the study of neurological disorders, recent work has discovered differences in the structure of graphs inferred from patient and control cohorts. However, most of these studies pursue a purely observational approach; identifying correlations between properties of graphs and the cohort which they describe, without consideration of the underlying mechanisms. To move beyond this necessitates the development of computational modeling approaches to appropriately interpret network interactions and the alterations in brain dynamics they permit, which in the field of complexity sciences is known as dynamics on networks. In this study we describe the development and application of this framework using modular networks of Kuramoto oscillators. We use this framework to understand functional networks inferred from resting state EEG recordings of a cohort of 35 adults with heterogeneous idiopathic generalized epilepsies and 40 healthy adult controls. Taking emergent synchrony across the global network as a proxy for seizures, our study finds that the critical strength of coupling required to synchronize the global network is significantly decreased for the epilepsy cohort for functional networks inferred from both theta (3–6 Hz) and low-alpha (6–9 Hz) bands. We further identify left frontal regions as a potential driver of seizure activity within these networks. We also explore the ability of our method to identify individuals with epilepsy, observing up to 80 predictive power through use of receiver operating characteristic analysis. Collectively these findings demonstrate that a computer model based analysis of routine clinical EEG provides significant additional information beyond standard clinical interpretation, which should ultimately enable a more appropriate mechanistic stratification of people with epilepsy leading to improved diagnostics and therapeutics.
The human brain is perhaps the best example of a multiscale complex network, with organizational hierarchies spanning many spatial and temporal scales. At the microscale, neurons communicate with other neurons through both chemical and electrical coupling, with estimates varying from 1000 to 10000 synapses per individual neuron [1]. At the mesoscale within the cerebral cortex, networks of between fifty and one hundred and fifty neurons form minicolumns, which in turn aggregate to form cortical columns, each containing around 5000 neurons [2]. At the macroscale, networks of tightly coupled cortical columns form distinct regions of the cerebral cortex that communicate with each other in a functionally specific manner. There is now increasing evidence for the concept of a core of such brain regions that form structural hubs that are essential for facilitating normal cognitive processing [3]–[5]. Whilst the precise mechanisms by which communication between large-scale brain regions occurs remains an open question, it is widely accepted that many critical brain functions such as cognition and motor coordination result from the emergent dynamics of large networks of neurons [6] and phase synchronization across regions is thought to play a critical role in facilitating communication between regions [7], [8]. From an experimental perspective, a window into the underlying macroscopic structural network may be given by functional networks that can be inferred from imaging modalities such as fMRI, MEG or EEG [9], and a substantial number of methods has been developed and applied to derive functional networks, ranging from cross-correlation [10] and phase coherence [11], to Granger causality [12] and transfer entropy [13], [14]. Whilst strong functional connectivity during the resting state has been shown to be a good indicator of underlying structural connectivity [15], [16], it is important to note that there is no one-to-one translation, thus a degree of fluctuation in functional connectivity is to be expected. From a theoretical perspective, effective connectivity can be considered a mathematical model driven representation of the functional connectivity inferred from data space [17], and to understand differences between cohorts of patients and controls, several recent studies have used methods from the mathematical field of graph theory [9] to explore either effective or functional networks across a variety of neurological conditions including schizophrenia, dementia and Parkinson' s [18]–[21]. A further debilitating neurological disorder that is associated with abnormal synchronization between brain regions is epilepsy; a disorder characterized by the tendency to have recurrent seizures. The International League Against Epilepsy (ILAE) define an epileptic seizure to be “a transient occurrence of signs and/or symptoms due to abnormal excessive or synchronous neuronal activity in the brain” [22]. The role of neural synchronization in seizures has caused some controversy in recent years, in part due to how synchrony is defined [23]. For example, if synchrony at the microscale is strictly defined as single-unit action potentials occurring at the same instance in time, then it can appear that synchrony is decreased during seizures [24]. However, if a broader definition of synchrony, such as phase coherence or generalized synchronization [25], [26], is applied to macroscopic recordings such as EEG then evidence for hypersynchronous activity is commonplace [27]–[30]. An open-question when pursuing a purely graph-theoretic approach is the relationship between the observed network structure and the emergent dynamics supported by that structure; particularly if alterations in function relate to symptoms of the neurological disease (Figure 1). To address this question, it is necessary to introduce a model of the dynamics of each node within the network, and to study the interplay between local dynamics and network structure on the emergent activity. Mathematically, a number of approaches has been used to study the mechanisms of seizure activity. At the physiological level, the use of neural mass and neural field models [31], [32] has become increasingly established to describe the evolution of both spike-wave discharges [33]–[35] and focal epilepsies [36], [37]. These frameworks have enabled important steps toward patient specific representations of these models to be taken using both genetic algorithms [38] and Kalman filtering [39]. Alternatively, at the opposing level of detail, phenomenological models are used to qualitatively describe the critical features associated with different brain states [40]–[43]. These models are typically computationally inexpensive (at least for small networks) making them potentially applicable in a clinical setting, however, they are often only suitable for considering a network at a single level of description and thus represent a coarse simplification of the underlying neurobiology. In this paper we pursue this phenomenological approach, but choose a model – the Kuramoto model [44] – that is more naturally suited to elucidate the mechanisms by which multiscale network structures can lead to hypersynchrony within or between large-scale brain regions. The Kuramoto model has become a standard model to study synchronization phenomena across physics, chemistry, biology and neuroscience (see [45]–[47] and references therein). Mathematically, the relationship between the Kuramoto model and the Wilson-Cowan model [48], which is a prototypical neural mass model, has been established by Schuster and Wagner [49], [50], and more recently by Daffertshofer and van Wijk [51]. This transition is made in the limit of weak coupling, and as a consequence the amplitude of the original model is neglected. For our purpose, we treat the Kuramoto model as a purely phenomenological model (we show that it mimics the critical features of both background activity and seizures), enabling us to analytically study synchronization phenomena in large-scale networks. As a result we are not limited to the case of weak coupling, since we are not attempting to relate back to a more detailed physiological model. Instead, the approach we pursue is to consider brain activity, for example as reflected in the macroscopic electrical activity measured using EEG, to be the result of networks of oscillators coupled at two distinct scales of activity: The macroscale electrical activity that is recorded by a scalp electrode is the sum over the dipoles generated by cortical columns (mesoscale) in the vicinity of the electrode. We assume these cortical columns to be strongly connected at close range and form a fully connected network (a so-called complete graph) (Figure 2). In turn, each of these connected networks forms a node (or module) within a larger network, the structure of which is defined by positions of the scalp electrodes. At this larger scale, separate brain regions may share connections, yet do not form complete graphs. Instead the network structure is typically sparse and directed. The remainder of our paper is arranged as follows. First we introduce the mathematical framework, the method we use to derive functional networks from EEG data, and details of the statistical analysis we perform. Next, we present our results in three parts. In the first part, we demonstrate the conditions required for the emergence of global synchrony across a network. In the second part, we use simple motifs to illustrate how subtle changes in the connectivity structures at different scales can have a dramatic influence on the degree of emergent synchronization, and further illustrate particular structures that can support the emergence of synchrony across either part of or the whole motif. In the final part of our results, we infer networks using clinically recorded EEG from a cohort of people with heterogeneous idiopathic generalized epilepsies, as well as a cohort of healthy controls. We then use our mathematical framework to explore the mechanisms by which seizures can emerge from these networks and find statistically significant differences in the networks of our epilepsy cohort, further demonstrating their potential predictive value at the individual level. We conclude with a discussion of the significance of our findings from both a theoretical and clinical perspective, some limitations of our approach, and suggest avenues for future research.
We build a modular network of nodes, using the Kuramoto model as a basis for each node: (1) The Kuramoto model is a mathematical representation of a network of oscillators, coupled together uniformly with strength through their phase. is the natural frequency of the oscillator, and we assume all these frequencies are drawn from a normal distribution function with mean and standard deviation 1. This model has previously been used to study neural oscillations, for example dynamic connectivity mimicking synaptic plasticity [46], [52], planar models with specific synaptic footprint [47], as well as the study of large-scale neural activity on realistic structural networks [53]. Here we consider each oscillator to represent the activity of a mass of neurons, such as a cortical column, for example. We couple together such nodes, each consisting of oscillators (in contrast to previous studies, which use one node per network, such as [51] or [54]), following the approach of Barreto et al. [55]. Here we introduce a coupling matrix with entries to describe the interaction between nodes and, weighted by a global coupling parameter: (2) The global connectivity matrix may be either a binary or weighted network, and for mathematical tractability we choose the natural frequencies from an identical frequency distribution with mean and standard deviation one, for every node. The degree of synchrony between oscillators within a single node and across the global network is controlled by the coupling parameters and. Focussing first on an individual node, Figure 3 demonstrates how the dynamic behavior of the Kuramoto model depends on the coupling constant. When this coupling constant is below a critical value, each oscillator behaves incoherently (i. e. they are uniformly spread around the unit circle) and the emergent signal is apparently stochastic and of low amplitude. However, when the coupling reaches a critical value, a phase transition occurs and oscillators become phase-locked (which in this context is synonymous to synchronized), resulting in emergent large amplitude oscillations; analogous to the transition between background and spike-wave activity seen in the onset of seizures. To measure the degree of synchrony within the oscillators of an individual node, we use the order parameter defined by: (3) which measures the level of phase coherence between all oscillators, where is the average phase. The order parameter is low () when oscillators are incoherent, and high () when they become coherent. Using equation (3), we can reformulate equation (2) to obtain: (4) Exploiting our assumption that the natural frequencies within each node come from the same distribution with mean, and that all connections in the network are either zero or positive, all ensemble averages will be in-phase (for all combinations) when, and consequently the following inequality holds: (5) Equivalently, we can find an expression for the phase difference between an individual oscillator and the ensemble average: (6) In the thermodynamic limit () we can describe the distribution of natural frequencies with a function. The order parameter is then the integral of the product of the density of natural frequencies and the cosine of the corresponding phase differences over all natural frequencies for which phase-locking occurs: (7) The upper and lower limit of the integral in (7), and, are determined from the inequality (5), resulting in: (8) By using the definition of Bessel functions we can evaluate the integral in (7), which yields an implicit equation for each node: (9) where the function is given by: (10) Having obtained an expression for the order parameter at the level of individual nodes, we now seek an expression for the order parameter across the network, which we term the global order parameter. The global order parameter is defined as (11) This expression can be reformulated using (3) to obtain: (12) Once more, all connections in the matrix are non-negative, thus beyond the onset of synchronization the phases of the ensemble averages at each node are equal and we obtain (13) Thus, the global order parameter is given by the average over, demonstrating that the degree of global synchrony can be inferred from the degree of local synchrony. EEG recordings used in this study were collected from 35 people diagnosed with idiopathic generalized epilepsy (21 female, mean age 34. 4 years), and from 40 controls (20 female, mean age 30. 7 years) as previously described [56]. Ethical approval to use this data was obtained from Kings College Hospital Research Ethics Committee (08/H0808/157). Written informed consent was obtained from all participants. From these recordings, one artefact-free 20 second segment of background activity (eyes-closed (EC) ) was extracted from each recording. The segments were bandpass filtered between 1–70 Hz, and notch-filtered between 48–52 Hz to exclude power line interference. The pre-processed data were then divided into frequency bands as given in Table 1. Whilst these frequency bands are different to those standard in a clinical setting (where the bands are defined according to prominent features visible to an expert observer), they are hypothesized to contain maximally-independent information representing different neurobiological generators [57]. Furthermore, given that brain network features in the alpha band may show evidence of heritability [58], [59], and that antiepileptic drug treatment my alter peak alpha frequency [60], this motivates the subdivision of alpha range following the work of [57]. To infer the functional network structure from EEG recordings, we use a method based upon time-lagged cross-correlation [10] to infer weighted networks from the voltage signals of each electrode (see [61] for an evaluation of linear and non-linear methods for inferring functional connectivity). Our specific choice is motivated by the predominantly linear nature of resting-state EEG [62]. To account for false connections due to common sources, we only consider those connections with non-zero time-lag since previous studies have demonstrated that volume conduction primarily manifests as an instantaneous correlation [63], [64]. Entries for the connectivity matrix are given by (14) with (15) A potential source of spurious cross-correlation are autocorrelation effects due to finite length time-series data. To account for this, we create surrogate datasets from our original EEG data via the iterative amplitude-adjusted Fourier transform (IAAFT) method (iterations) [65], which preserves autocorrelation whilst removing genuine pairwise cross-correlations within the time-series. Applying our method pairwise within each of these surrogate datasets creates a spectrum of cross-correlation values which could arise as a consequence of autocorrelation alone within the specific pair. Therefore we reject connections from the original EEG data if they do not exceed the level of significance. Next, we create a directional matrix by setting if, and if. If, we set in order to remove zero time-lag connections. Further, we remove spurious connections by setting if, at first order, there exists a node such that. At second order, we set if there exist two nodes, such that. In other words, direct connections between nodes are removed if there exist stronger, indirect connections via one or two other nodes. A graphical representation of this procedure is given in Figure 4. Finally, this functional connectivity matrix feeds into equation (2) in what may be thought of as an effective connectivity representation of the observed dynamics. In order to test for statistically significant differences in the model-based measures at the group level, we use the Wilcoxon rank sum test [66]. In comparison to parametric tests, such as the t-test, this method does not assume the existence of an underlying normal distribution. The test yields the -value (likelihood) that the medians of both samples are the same (null-hypothesis). As our analysis involves multiple hypotheses (frequency bands, nodes in the network) we correct the -value using a conservative Bonferroni correction (effectively multiplying the -value by the number of hypotheses considered). If this corrected -value is below, we consider the difference between the samples to be significant. For model-based measures found to be significant on this basis, we then explore the discriminative power of the measure through computation of the receiver operating characteristic (ROC). The resulting ROC curve plots the true positive rate (TPR) against the false positive rate (FPR), which is achieved through varying the threshold (that parametrizes the ROC curve), and counting all sample points below this threshold as positives. Next, we identify the point on the curve which is closest to the the point of perfect classification, , the upper left-hand corner. For this point, we compute the positive predictive value (PPV) defined as: A PPV of indicates the measure has no discriminative power, whilst a PPV of indicates perfect classification. The False Discovery Rate (FDR) is defined to be.
First, we address the conditions necessary for the global network of nodes, or a subset thereof, to synchronize. As for the case of the standard Kuramoto model, an individual node or a subset of nodes can synchronize if their intrinsic coupling strengths are greater than the critical coupling strength. However, suppose that all nodes are individually in the desynchronized state (i. e.). Does there exist a critical value of the global coupling parameter such that synchrony across some or all of the nodes in the global network emerges? To explore the existence of such a critical value, we first linearize (9) around giving: (16) where is a -dimensional diagonal matrix with elements, . Trivially, the zero-solution for all order parameters () exists for any choice of. However, non-zero solutions for exist if the following determinant condition holds: (17) Solving this determinant problem is computationally efficient in comparison to the corresponding full nonlinear problem (9). Alternatively, since is invertible (as), we can reformulate (16) as a standard eigenvalue problem with as eigenvalue: (18) As is diagonal, its inverse is diagonal as well, with elements. Finally, as all real-valued eigenvalues of represent the inverse of a coupling constant that permits non-trivial solutions around the zero-state, we identify the critical coupling constant with the inverse of the largest (real) eigenvalue of: (19) We refer to this scenario as “network-driven synchronization”. Alternatively, if is zero, then no critical value of the global coupling exists, since must be positive. This is the case when represents a network with hierarchical flow, and has upper triangular form. Here is nilpotent and all its eigenvalues are zero. In this scenario there may exist a node or nodes which, if synchronized (i. e.), can drive other nodes (with intrinsic coupling parameters less than) to synchronize due to the topology of the hierarchical network. We term this scenario “node-driven synchronization”. For the specific problem of epilepsy, we might consider these two scenarios equivalent to seizure onset as a distributed network property versus the existence of an epileptogenic zone, for example. To better understand these different conditions for emergent synchronization, we focus initially on motifs with a small number of nodes. First, we consider the case of two nodes that are either unidirectionally or bidirectionally coupled. In graph-theoretical terms, two bi-directionally coupled nodes are the simplest example of a feedback-loop or cycle, which, in turn, is the simplest form of a strongly connected component. A strongly connected component is a network configuration such that each node can be reached from all other nodes by following directed connections. Hence, a strongly connected component must contain at least one cycle. Conversely, two nodes that are uni-directionally coupled represent the simplest form of a network with a hierarchical flow, such that each node can be assigned its own level of hierarchy. First, we consider two coupled nodes in the most general scenario, where the connectivity matrix has entries and. Then from (17) we obtain the following expression for the critical value of the global coupling parameter, above which both nodes become phase-locked and their order parameters increase: (20) Again, is the critical value of for the onset of synchrony within an individual node. is only real-valued for, verifying that neither node is synchronized in isolation (a prerequisite for network-driven synchronization). In this scenario the value of depends on the distance of the intrinsic coupling parameters from. If both nodes are identical with, and we obtain the simple expression (21) U ni-directional coupling (i. e. either or) means (20) is undefined, making network-driven synchrony impossible. In this scenario, only node-driven synchrony can arise as a consequence of their intrinsic coupling exceeding the critical value. A comparison of both cases is shown in Figure 5, accompanied by numerical results for oscillators, which leads to an increasing divergence between analytical and numerical results for the uni-directional case with increasing global coupling. We now generalize these ideas to larger networks, for which we make the distinction as to whether there exist cycles (or feedback loops, as per the bi-directional two node case) or a hierarchical structure (as per the uni-directional case). Having studied the conditions necessary for the emergence of global synchrony to be network or node-driven, we now apply this understanding to functional networks inferred from EEG recordings collected from our cohorts of people with epilepsy and controls. For each individual and each frequency band we obtain a functional connectivity matrix, where each node within the network corresponds to a specific EEG channel. We study these matrices from two perspectives: First, we consider the critical value for the global coupling parameter, above which network-driven synchrony emerges and compare these values for functional networks inferred from people with epilepsy and those inferred from controls. Where the critical coupling strength required to enable global synchrony is smaller, this suggests that those networks are more seizure prone than others. Second, we study whether there exist specific nodes within these functional networks which may drive emergent synchrony across the wider network. This latter study is motivated by recent studies from human and rodent models that suggest generalized seizures in IGE appear to have a focal onset [67]. Here, we set each node to be self-synchronized, and analyze the effect this has on the rest of the network by computing the global order parameter, which indicates the global degree of synchronicity.
We have previously described a phenomenological approach to studying network abnormalities in epilepsy from a purely theoretical standpoint [42], illustrating that seizures could occur due to either abnormalities in the dynamics of brain regions or the connectivity structures between them. Here, we advance this understanding using a modular network of embedded Kuramoto oscillators, that has enabled us to explore the interplay between node dynamics and network structure in the emergence of hypersynchrony, analogous to the generation of seizures, both in theory and in real data. We have derived necessary conditions for the emergence of synchronization within large scale networks in terms of the pattern of directed edges in the network, the intrinsic coupling parameters within each node, and the macroscopic coupling parameters between nodes. Specifically, we demonstrate that strongly connected components (i. e. disparate regions that form complete cycles) are necessary for the emergence of global synchrony for a collection of nodes that individually are sub threshold, whereas directed networks with hierarchical flowcan only result in global synchronization if nodes at the top of the hierarchy become synchronized themselves. In binary networks, strongly connected components can be created by adding connections to an existing network, or they can be destroyed by removing specific connections. In general, an indicator of the degree of network-driven synchronization is the critical value of the global coupling parameter; a small indicates a strong disposition for nodes to synchronize due to the particular structure of the network. In larger, weighted networks we might reasonably expect to observe mixtures of strongly connected components and hierarchical subnetworks. Indeed, this is demonstrated in our results of functional networks inferred from EEG data. On the one hand, we find that the critical coupling strength, that is necessary to enable the emergence of synchrony in the global network, is significantly lower for networks inferred from the EEG recordings of our cohort of people with epilepsy in comparison to healthy controls. This indicates an increased presence of strongly connected components in the network and suggests a fundamental mechanism for the tendency to experience recurrent seizures in people with epilepsy. On the other hand, we observe that left frontal brain regions (represented by EEG channels Fp1 and F7) can drive increased levels of global synchronization when they are self-synchronized, which indicates an increased presence of hierarchical flows as well. These latter findings, that generalized seizures may be driven by activity in left frontal regions of the brain, complement previous findings using other imaging modalities, for example Pavone and Niedermeyer [71] who identified a cortical, mostly frontal lobe involvement in absence seizures and primary generalized seizures. Likewise, Holmes et al. [77] and Amor et al. [72] identified frontal areas as highly involved during absence seizures. This evidence is supported by the fact that working memory - a frontal lobe function - is suspended during typical absence seizures. A critical advantage of our approach is that these differences are identified from epochs of data from inter-ictal time-periods (i. e. away from seizures). MRI studies [73]–[76] in one particular IGE syndrome, juvenile myoclonic epilepsy, have identified a structural abnormality of medial frontal cortex, and abnormalities of structural connections (using DTI) and functional connections (using fMRI) of this area with motor cortex, frontopolar cortex, thalamus and contralateral medial frontal cortex, supporting the EEG/MEG data implicating frontal abnormalities. Further, Yan and Li [53] inferred human brain networks from diffusion-magnetic resonance imaging in healthy controls, and postulated that frontal hubs could drive seizure activity when placing these data inferred networks onto a computational model utilizing a delayed version of the Kuramoto model. Our study extends this research by comparing the networks of people with epilepsy directly with those of healthy controls, and demonstrating an increased propensity for seizure generation as a consequence of the functional network structure. This current study complements our earlier work [41], [42], where we used an alternative model formulation (a subcritical Hopf bifurcation to reflect the rapid transition from background activity to seizures) to explore the role of network structure in driving the onset of seizures. Our present study has focussed on analysis of routine clinical EEG in ‘sensor’ space, by which we mean functional networks were inferred through studying the interactions spanning EEG electrodes, rather than the interactions between the brain sources responsible for generating the activity. This is necessitated by the limited spatial sampling of the clinical data (19 channels) that does not readily permit the use of source reconstruction techniques [78]. Volume conduction can be a further concern when utilizing data from scalp electrode recordings. To compensate for this, we used a time-lagged cross correlation function (excluding the zero time lag) for inferring functional networks. It is therefore important that future work should extend our approach to larger networks inferred from either high density EEG, fMRI or DTI data. Given that we determine the point of onset of global synchronization using an analytic expression, our framework can be applied naturally to networks of any size. Further research should also seek to understand long-term disease progression through studying longitudinal clinical recordings collected at regular intervals post-diagnosis, or in response to changes in treatment. Here, alterations in network structure or dynamics may point towards remission or successful drug response. Of further importance is the non-trivial relationship between structural, functional and effective connectivity. For example, a large repertoire of functional networks can be supported by the same underlying structural network [79]. Further, effective connectivity is dependent both on the choice of generative model, as well as the observation data. Thus, effective connectivity should not be thought of as a unique representation of our data, and more likely there can be different (model, network) pairs that are consistent with the observed functional connectivity structure. As a simple example of this, consider the auxiliary approach for detecting generalized synchronization introduced by Abarbanel et al [80]. Here, a system drives a system, and the dynamics of these systems may become coherent depending on the nature of the coupling. If the dynamics of and are chaotic, then this coherent relationship is highly non-trivial. However, by considering a copy of system, Abarbanel and colleagues show that one can infer the existence of this synchronization between and, by observing that it is possible to infer a much simpler functional relationship between and its copy, even though there is in reality no direct connection between them. Now reverse-engineering this scenario, suppose that we can only observe and, with no knowledge of system. As an effective connectivity structure, we would identify a bidirectional link between and with a model representation of the simple functional form. However, in reality, there is an alternative effective connectivity structure that links to and to with the original more complex relationship. Whilst it has been shown that there can still exist a wide repertoire of functional networks [81], we might reasonably expect differences across cohorts to become apparent in resting-state functional networks at the group level. The inherent variability in functional expression may reflect the overlap between the patient cohort and the control cohort. In conclusion, our findings are significant for a number of reasons. First, they demonstrate the power of pursuing a computational modeling approach to elucidate the mechanisms underlying differences observed in graph theory measures of data inferred functional brain networks. In this regard, the approaches we describe may have potential for understanding inferred brain networks from other neurological conditions, for example dementia (for which there is a strong association with seizures [82]) and schizophrenia [83]. Second, our study was performed inferring functional networks using epochs of background activity (i. e. away from seizures), which suggests that network structure is an enduring and critical marker of the propensity for seizures that offers the potential for diagnosis of epilepsy without the need to induce seizures within the clinical environment. Indeed, in this regard, beyond the group level differences we identify, the ROC analysis we performed demonstrated up to 80 predictive power of this method for discriminating at an individual level, despite neither the model nor data being optimized for this purpose. This strongly motivates the potential of this approach. Third, our methods identified these networks using routine clinical EEG recordings, with low spatial and temporal sampling. Despite these apparent limitations, our study identified candidate regions that drive the onset of seizure activity, which are consistent with those obtained using more expensive MEG and fMRI modalities. Finally, deriving a mathematical equation for the global synchrony of the network makes it computationally tractable to analyze patient data in close to real time (through removing the need to numerically simulate large networks of oscillators). Taken collectively, these findings suggest that a computational modeling approach to analyze routine clinical data can be used in real time within the clinic as a diagnostic aid for clinicians treating epilepsy, as well as other neurological disorders, for which synchrony may potentially play a role. | In this paper we show that within modular networks (that is, networks with multiple scales of connections), two distinct mechanisms may drive the emergence of synchrony at the global level. We term the first of these mechanisms “network-driven synchrony”, which is characterized by the presence of cycles within the macroscopic network. The second mechanism we term “node-driven”, which is characterized by the ability of an individual node (or nodes) to drive synchrony across the rest of the network, due to the hierarchical structure of the macroscopic network. By applying this framework to routine clinically collected resting state data from people with idiopathic generalized epilepsy and from age matched healthy controls, we demonstrate that functional networks of people with epilepsy have a significantly enhanced capacity to synchronize than those of people without epilepsy. This finding suggests a critical role for the connectivity structure of large-scale networks in the tendency to have seizures. Further, by deriving a mathematical equation for the global synchrony of the network, we make it computationally tractable to analyze data in close to real time. This gives our method potential to be used within the clinic as a diagnostic aid for clinicians treating neurological disease. | Abstract
Introduction
Materials and Methods
Results
Discussion | epilepsy
computer and information sciences
medicine and health sciences
computer modeling
systems science
network analysis
neurology
epileptic seizures | 2014 | Dynamics on Networks: The Role of Local Dynamics and Global Networks on the Emergence of Hypersynchronous Neural Activity | 7,557 | 272 |
Computer science has become ubiquitous in many areas of biological research, yet most high school and even college students are unaware of this. As a result, many college biology majors graduate without adequate computational skills for contemporary fields of biology. The absence of a computational element in secondary school biology classrooms is of growing concern to the computational biology community and biology teachers who would like to acquaint their students with updated approaches in the discipline. We present a first attempt to correct this absence by introducing a computational biology element to teach genetic evolution into advanced biology classes in two local high schools. Our primary goal was to show students how computation is used in biology and why a basic understanding of computation is necessary for research in many fields of biology. This curriculum is intended to be taught by a computational biologist who has worked with a high school advanced biology teacher to adapt the unit for his/her classroom, but a motivated high school teacher comfortable with mathematics and computing may be able to teach this alone. In this paper, we present our curriculum, which takes into consideration the constraints of the required curriculum, and discuss our experiences teaching it. We describe the successes and challenges we encountered while bringing this unit to high school students, discuss how we addressed these challenges, and make suggestions for future versions of this curriculum. We believe that our curriculum can be a valuable seed for further development of computational activities aimed at high school biology students. Further, our experiences may be of value to others teaching computational biology at this level. Our curriculum can be obtained at http: //ecsite. cs. colorado. edu/? page_id=149#biology or by contacting the authors.
The absence of a computational element in secondary biology classrooms is of growing concern to the computational biology community. While computer science is increasingly important in modern biology research [1]–[4], it plays almost no role in high school and undergraduate biology classes. Few of these students are even aware that computation plays any role in biology [5]. We developed a computational biology unit aimed at advanced high school students and taught this unit in three different classes in two local high schools. The goal of this unit is to teach students the connection between computer science and biology and demonstrate how computational techniques can lead to biological discoveries. This project was undertaken as part of the Engaging Computer Science in Traditional Education (eCSite) Program, which is funded by the National Science Foundation (NSF) GK-12 program. The goals of the GK-12 program include training graduate students to communicate their research to the general public while providing K-12 students a glimpse of the work and life of scientists. eCSite is a 5-year program that began in 2009 to bring computer science into traditional K-12 classrooms. The goals of eCSite include improving computational literacy, which is important to all people living in the 21st century, and raising awareness of the possibilities for impacting our world with a career in computer science by showing K-12 students and teachers some of the ways computing affects modern life. Rather than create new computer science classes that only students already interested in computer science are likely to take, the goal is to bring computer science into classes such as geography, physics, government, art, and biology. In our first year, two eCSite fellows (graduate students in computer science) worked with Advanced Placement (AP) Biology teachers at two different high schools (as described below, the content was also brought to other advanced high school students taught by these teachers). The eCSite fellows and teachers developed and presented a unit to help students gain a deeper understanding of genetic evolution (a requirement in the AP biology curriculum) using the Basic Local Alignment Search Tool (BLAST) [6] and phylogenetic trees in three distinct lessons. In the first lesson, we gave a brief introduction to algorithms. In the second, we gave an introduction to BLAST, including demonstrating the sequence alignment problem and the challenges associated with it. In the third lesson, we showed them an application of BLAST, building evolutionary trees using the matched sequence. Our lessons were largely successful in showing students the relative scale of computation involved in genomic research and giving them an appreciation for the methodology that goes into any computational research. Students were able to complete the lessons and gain an understanding of how algorithms such as BLAST might be used in biological research and why some understanding of these algorithms is important to students of biology. While few students had the necessary background to construct complex algorithms, they were able to use and analyze existing algorithms. Teachers reported that student reaction to the curriculum was mixed. Some students were interested in the computational element, seeing it as a tool used by “real” biologists and eager to learn more about how it was used. These students, upon seeing the amount of genetic data available in the various databases, realized that advances in this area would be impossible without good computational methods to analyze the data. Other students, however, either failed to see the importance of computation to their current studies or indicated that they did not have time to learn about something that would not explicitly be on the AP or International Baccalaureate (IB) test. We discuss these challenges as well as possible ways of overcoming them below. Despite these challenges, we believe that our curriculum can serve as a starting point for further development of computational biology activities for high school students. We also believe that our experiences, including our difficulties and our methods of overcoming them, will be of value to other computational biologists interested in updating the content of high school biology classes.
Two eCSite fellows (computer science graduate students) with research experience in computational biology worked closely with two AP biology teachers in order to introduce computational methods into required AP biology lessons. Beginning in September, they met weekly to brainstorm possible units in an effort to minimize time taken away from required content. In addition, the two eCSite fellows met with each other on a weekly basis to coordinate the planning with both teachers, and monthly with an eCSite co-director whose research interests are in computational biology to help guide the content. There were periodic meetings of all involved (the eCSite fellows, AP Biology teachers, and co-director). Curriculum was developed during the fall semester and delivered to the classes during the spring semester. Our curriculum consisted of three basic lessons: an introduction to algorithms, a basic discussion of the BLAST algorithm [6], and algorithms to construct phylogenetic trees. The first lesson, an introduction to algorithms, was designed to introduce students to the basic vocabulary involved with algorithms and get students thinking algorithmically. We started by introducing students to algorithms for every day tasks, such as making coffee, then asking them to write their own algorithms to make peanut butter sandwiches. The next activity had students participating in a “living computer” algorithm: each student was given a numbered instruction that represented one step in the algorithm. Students then stood in line in front of the class in the order that the steps would be performed in the computer. Each student performed their step and handed the “output” to the next student in line until the final step had been reached. After each step was acted out, it was executed on a computer using the same inputs so that the students could see the same output coming from the computer as their “living computer”. The students executed the algorithm, and then were shown the computer code that did the same thing. This allowed students to see what sort of basic steps should be included in an algorithm and how those steps could be translated to computer code. Finally, we had students write algorithms to create Punnett squares, diagrams to predict the outcome of a cross-breed given the genotypes of the parents. Students were already familiar with the creation of Punnett squares, so this helped students think about biological problems in algorithmic terms. The second lesson focused on DNA sequence comparison using BLAST. This lesson had two parts. In the first part of the lesson, we explained BLAST in terms of a word search. We gave students three word search puzzles containing various vocabulary words from biology and genomics (see Figure 1). The first was a “perfect” word search, simulating the problem of finding an unmutated gene in another genome. In the second puzzle, students were given a “wrong key” word search created by someone who would occasionally hit the wrong key, so some letters might be wrong. This served as a simulation of the problem in a genome with single nucleotide polymorphism (SNP) mutations. The third puzzle was designed to simulate the problem in a genome that contains both SNP and insertion/deletion (indel) mutations. Students were given a “ditz” word search that contained not only wrong letters, but missing or added letters, as if the person writing the word search may have skipped parts of the puzzle or forgotten what he was doing and started typing in a shopping list or telephone message. The second part of the lesson introduced students to an implementation of the BLAST algorithm. We asked students to choose a disease with a genetic basis, then search for that gene in the National Center for Biotechnology Information (NCBI) database. Once they obtained the DNA sequence for that gene, they were asked to run it through the BLAST implementation on the NCBI website to find similar genes in humans and other species. Students were then asked to answer several questions about the BLAST results, including: give the scientific and common names of species that have similar genes, give the percent of base pairs that could be matched in a given alignment, and give an example of an alignment that contains an indel. The third lesson was building phylogenetic trees. Students were given sequence data on the COX15 gene from eight different species. These genes were identified using only first names (“Alex”, “Chris”, etc.) without any information about what species they came from. Students were then asked to construct a phylogenetic tree by running pairwise BLAST to compute a similarity score between every pair of genes, and to cluster based on these results. Students were taught a simple hierarchical agglomerative clustering algorithm [7], and different students were given different metrics for determining distance between clusters: some students were asked to cluster by considering the distance between clusters to be the distance between the closest two points in the clusters (single linkage), some students were asked to consider the distance between clusters to be the distance between the furthest two points in the clusters (complete linkage), and the remainder were asked to consider the distance to be the average distance between all points in the clusters (average linkage). Students then compared the results of their different clustering algorithms to see that seemingly minor differences in algorithm implementation can result in different biological conclusions, reinforcing the notion that biologists need to understand details of the computational tools they use, even if they aren' t themselves developing the tools. Finally, students used BLAST to identify species from their sample genomes. Once species were identified, students could compare their results to their predictions of species relatedness. This unit was taught in three different classes in two different high schools: Centaurus High School and Monarch High School. Centaurus is a high school located in Lafayette, Colorado, with 28% of students receiving free or reduced lunches. Monarch is a high school in Louisville, Colorado, with 4% of students receiving free or reduced lunches (data is from the Boulder Valley School District website at http: //www. bvsd. org/and is based on numbers collected in October 2009). In Centaurus High School, the unit was taught to a combined AP/IB Standard Level (SL) Biology class composed mostly of high school juniors with some seniors. At Monarch High School, it was taught to an AP Biology class and a dedicated Biotechnology science elective class. There was little overlap between students in the Biotechnology class and those taking any AP science classes. Due to scheduling constraints in the schools, the lessons in the unit were spread out over a period of approximately three months. Lessons were taught primarily by the eCSite fellows in collaboration with the classroom teachers. For various reasons, not all activities were done in all classrooms. Lessons were evaluated through discussion with the classroom teachers. In formal and informal discussions, the teachers reported to the fellows their observations about which part of the lessons the students had enjoyed, which parts they had struggled with, and whether or not students were able to connect these lessons to other lessons from biology class. Teachers also reported on the general types of comments, questions, and complaints they had received from students about the curriculum.
Our activities were generally successful and had the students interested and engaged. Most students enjoyed the first lesson. They liked seeing an algorithm on how to get a cup of coffee and making their own algorithms to make peanut butter sandwiches. They also enjoyed the living computer activity and seeing how an algorithm worked and translated into computer code. This suggests that there is interest in computational and algorithmic activities as long as these activities are presented at the correct level. In the later, more biology-oriented lessons, students enjoyed doing the word search puzzles. Likewise, they seemed interested in seeing how different hierarchical clustering algorithms created different phylogenetic trees (Figure 2). Also in the third lesson, they seemed to enjoy trying to figure out from which species their genome sequences came. Again, we found that as long as students were readily successful at the task at hand, they were willing to engage with the material. The unit was a success in that most students were able to complete the assigned activities. Though there were multiple difficulties and challenges (detailed below), the vast majority of students were able to run a genetic sequence through BLAST and find multiple species with similar genes. They were able to use BLAST to compare two genetic sequences and build a phylogenetic tree using the results. Over the course of the unit we encountered several challenges. We detail these here along with possible responses to them. The curriculum is designed to be taught by computational biologists interested in updating the content of high school biology classes. While the fellows who taught this particular curriculum were embedded in the classroom for most of the school year, this would not be necessary. The lessons are designed to be done in one or two class periods and could be taught either by a computational biologist coming into the classroom for a few days or as part of the high school outreach science camps that many colleges offer during weekends or the summer. While the lessons are designed to be introduced by a computational biologist, we hope that after team-teaching the lessons with a high school biology teacher, the teacher will be able to teach the lessons without assistance. The lessons do not require the teacher to know how to program; the computer portions of the lessons use either pre-existing, publicly available tools (such as the implementation of BLAST on the NCBI website), or programs that are already written and can be downloaded along with the curriculum. One of the two biology teachers involved in the first year of our program successfully taught the lesson independently in the second year. In addition, since high school biology teachers helped develop these lessons so that they integrate well with the required AP Biology curriculum and address required “essential knowledge, ” teachers who are motivated to incorporate modern methodology and are comfortable with computers may be able to teach them independently from the beginning. The curriculum can be used as is; however, we encourage those using the curriculum to modify it, either to suit their own needs, to correct potential flaws in the curriculum (see below), or otherwise improve it. The curriculum that we have developed is a first attempt to bring a computational element into the high school biology curriculum, not a final product. It is a functional lesson plan and could be used by those not interested in developing their own curriculum, but there is much room for improvement by those interested. Based on our experience and feedback from other computational biologists, we believe there are several areas where a review of the current lessons would be of value. Some lessons would benefit from changes in the activities. Other lessons need a more thorough explanation of how the lesson ties into the broader picture of biological research. Though simplifications are appropriate for the high school lesson, we would highlight these simplifications to show students how researchers would do things differently. In reviewing the first lesson, the general introduction to algorithms, our first question was, “Is this lesson necessary? Do the activities involving general algorithms enhance the understanding of the later, biological algorithms, or are they an unnecessary sideshow? ” Critics of the first lesson felt that it should be eliminated because it is largely unrelated to biology and makes students more inclined to see the entire unit as a distraction. Countering that view is the argument that BLAST and phylogeny algorithms represent breakthroughs in computer science as well as biology, and students need to understand something about algorithms in order to get the “big picture. ” Most people involved in the development of the curriculum, including one of the two high school biology teachers, felt that the algorithms lesson was important, though it could use some improvements to tie it more closely to the study of biology. We plan to eliminate the activity where students were asked to develop the algorithm for Punnett squares. Our experience indicated that this activity was too difficult for most of the students and generally produced frustration rather than understanding. In order to accomplish the goal of the activity, to get students to think about biological problems in algorithmic terms, we may instead show students an algorithm for Punnett squares using the “living computer” or some other activity. We also discussed changing the order of the lessons in order to move the general discussion on algorithms from the first lesson to a place where students might better appreciate the connection between algorithms and biological research. One possible placement would be after the word search puzzles given in the lesson on searching for similar DNA sequences. After students complete the first (or all three) word search puzzles, students could be asked to describe general strategies that could be used for word search. Strategies could include looking for each word separately, looking for the start of all words simultaneously in the first pass of the word search, and looking through the word search for sequences that seem “word-like” without first reading the list of hidden words and then scrutinizing locally around these areas more carefully. Our experience indicates that such an approach would need to be scaffolded with ideas from the high school teacher, because even after a complete lesson on algorithms, students had difficulty describing their word search strategies when asked. Students could be asked what they read first (the word search or the list of hidden words). After coming up with different general approaches, the lesson would progress to details of an algorithm, perhaps in living algorithm format. Alternatively, the discussion on algorithms could be placed after both the lesson on searching DNA sequences and building phylogenetic trees. This would allow students to learn about algorithms after they have already seen the algorithms for BLAST and building phylogenetic trees. Examples would then be drawn from both these problems rather than from general tasks, making the lesson seem more relevant. Our review of the second lesson, the lesson on BLAST, showed two places where we feel changes would be beneficial. The first was that the “open-ended” nature of the computer activity had students feeling confused and directionless, especially those students who had trouble navigating the NCBI website. We have modified the lesson so that rather than looking for any genetic disease, all students look at sickle cell anemia. We have also included the code for the hemoglobin beta gene in both rat and human so that students would not need to navigate the website in order to find the correct code. These changes are already reflected in our curriculum posted online. In order to reinforce the idea that biologists whose work does not require programming still need to understand how computational tools work, we would introduce an element into the second lesson to use different BLAST parameters. We would have students try different seed lengths, different match/mismatch scores, and different gap costs. We would like students to come away from the lesson with a sense not only of the problem BLAST is trying to solve and how the program can be useful, but also how knowing more about the BLAST algorithm can aid them as biologists. As we add to this lesson, however, we must be mindful of the additional time it would take and whether or not this time can be justified given the rigid constraints of the AP and IB course requirements. Our review of the third lesson, building phylogenetic trees, focused primarily on whether or not we were misleading students about the methods used by biologists. We ask students to build evolutionary trees using a clustering procedure based on sequence similarity. While this, in broad strokes, is what evolutionary biologists would do, the methods are not up to standard. BLAST does not give a good distance metric for phylogenetic trees, and our clustering methods are all based on the unweighted pair-group method using arithmetic averages (UPGMA), which is known to perform poorly in some cases. However, our initial methods have two relevant advantages. Using BLAST rather than another method of sequence similarity lets us build on the second lesson and gives students more practice using BLAST. Also, by using a tool with which the students were already familiar, we avoided having to teach another tool. While some tools for building phylogenetic trees from sequence data are relatively easy to learn, if we were to use a different tool for determining similarity, we would have to either teach students a general understanding of the underlying algorithms as well as how to use that tool, or give them a pre-computed matrix of similarity scores. The former option might take time that is not available within the course constraints, while the latter option would deny students the opportunity for further practice with a computational element. Introducing the concepts of parsimony, maximum likelihood, and Bayesian inference would likely take too much additional class time, given that these concepts are not tested on the AP or IB exams. On the other hand, the simple UPGMA clustering algorithms require little teaching time and are easy for the students to both do and understand, while more complicated methods might either require significant time or leave students simply plugging numbers into equations that they don' t understand. The advantages of using these simplifications must be weighed when deciding whether to use a more biologically accurate method. Again, a possible compromise might be to have students do the lesson as it is, but then explain to them the flaws in this methodology and have them compare their trees to ones built using more sophisticated algorithms [8]–[13]. As mentioned above, the curriculum that we have developed is a first attempt to bring a computational element into the high school biology curriculum, not a final product. In future years of the eCSite project, the curriculum will be modified to take into account teaching experience as well as feedback from students, teachers, and members of the computational biology community. In addition, we encourage those who use the curriculum to modify and improve it (please tell us if you do, so others can benefit from your changes). Our unit focuses on one of many areas of biology in which computers now play a significant role. Future lessons on computer modeling in areas such as epidemiology and ecology could be adapted from other eCSite units developed for middle school and high school classes [14]. New units could be developed on biological networks (metabolic networks, predator–prey relationships) and protein structure and function. On his blog, Professor Kevin Karplus of UC Santa Cruz has gone through the curriculum framework of the AP Biology course [15] and suggests several essential knowledge requirements that can be addressed in a computational manner. Dr. Karplus suggests that several requirements could be addressed using computational methods (this discussion can be found at http: //gasstationwithoutpumps. wordpress. com/2011/01/08/advanced-placement-bio-changes-announced/). While our initial curriculum addressed some of these requirements, his analysis will prove valuable for planning new units. In modifying lessons and developing new ones, we will have to keep in mind our goals. We want to introduce students to the connection between biology and computer science. We would like to show students computational techniques used in “real” biology while at the same time respecting the limits of their abilities: the students to whom this curriculum is directed have only an advanced high school knowledge of biology and little, if any, knowledge of computer science. Finally, we need to respect the limited amount of time available to devote to new content in a high school AP or IB Biology class. A good unit for high school age students should be geared towards understanding existing algorithms rather than trying to build complex algorithms. Lessons should be based around understanding the algorithms used to solve biological problems and using that understanding to make better use of the algorithms.
We incorporated a computational biology unit into advanced high school biology classes to teach students about the importance of computer science to biology. We presented a general introduction to algorithms, an overview of algorithms for searching for related DNA sequences (including BLAST), and algorithms for building phylogenetic trees. Our unit enjoyed a great deal of success. Students got an idea of the volume of data involved in modern biological research and an appreciation of the need for computational methods to handle this data. They gained experience with a computational tool used by working biologists. Students also developed an appreciation for why biologists should have some understanding of how these computational tools work, even if they are not interested in developing new computational tools. High school teachers were able to independently incorporate the lessons they helped develop. We encountered challenges associated with teaching high school students due to their lack of experience and the pressure they are under to pass high-stakes exams. A lesson geared towards students at this level needs to be compatible with their limited experience with mathematics and computer science and also be geared towards enhancing understanding of required material rather than replacing it. It should be made clear to students that they are learning the required curriculum using computational methods. The lessons must be closely tied both to the existing local curriculum at the school and the broader AP or IB curriculum for advanced students. There should be connections between the computational lessons and past lessons in the class, as well as connections between the lessons and the high-stakes exams. To do this well, the regular teacher must be heavily involved. The curriculum presented here is a work in progress. All activities are being evaluated to see how they further our goal of demonstrating the connection between computer science and biology. In future classes, some activities may be eliminated or modified to enhance our goals or to better reflect current biological research. This curriculum serves as a seed for further development, and our experiences teaching it can guide that development. In future years of this project, we hope to improve the curriculum in response to student and teacher feedback (locally and from readers of this journal), as well as expanding it to include other areas of computational biology. | We have designed and implemented a curriculum to teach basic computational biology to advanced high school students. The curriculum includes an introduction to the concept of algorithms, an overview of the Basic Local Alignment Search Tool (BLAST) algorithm used to compare DNA sequences, and methods for building phylogenetic trees. We taught this curriculum in advanced biology classes at two local high schools. As a result of this, we were able to give many students an appreciation of the role computers play in biology and an idea of why computational methods are needed in biological research. We found that while the high school students lacked the necessary background in math and computer science to be able to write their own algorithms, they were able to use existing algorithms, analyze them, and compare the results. We also encountered a number of challenges that could arise in other attempts to teach computational biology to students at this level, whether using our curriculum or another. We discuss each of these challenges and possible ways that they can be overcome. | Abstract
Introduction
Methods
Results/Discussion
Conclusions | education
biology
computational biology | 2011 | A First Attempt to Bring Computational Biology into Advanced High School Biology Classrooms | 5,804 | 208 |
Telomere length (TL) predicts health and survival across taxa. Variation in TL between individuals is thought to be largely of genetic origin, but telomere inheritance is unusual, because zygotes already express a TL phenotype, the TL of the parental gametes. Offspring TL changes with paternal age in many species including humans, presumably through age-related TL changes in sperm, suggesting an epigenetic inheritance mechanism. However, present evidence is based on cross-sectional analyses, and age at reproduction is confounded with between-father variation in TL. Furthermore, the quantitative importance of epigenetic TL inheritance is unknown. Using longitudinal data of free-living jackdaws Corvus monedula, we show that erythrocyte TL of subsequent offspring decreases with parental age within individual fathers, but not mothers. By cross-fostering eggs, we confirmed the paternal age effect to be independent of paternal age dependent care. Epigenetic inheritance accounted for a minimum of 34% of the variance in offspring TL that was explained by paternal TL. This is a minimum estimate, because it ignores the epigenetic component in paternal TL variation and sperm TL heterogeneity within ejaculates. Our results indicate an important epigenetic component in the heritability of TL with potential consequences for offspring fitness prospects.
Telomeres are evolutionarily conserved DNA sequence repeats, which form the ends of chromosomes together with associated proteins and contribute to genome stability [1]. Telomeres shorten due to incomplete replication during cell division, which can be accelerated by DNA and protein damaging factors and attenuated or counter-acted by maintenance processes, mainly based on telomerase activity, a telomere-elongating ribonucleoprotein [2]. On the organismal level, telomere length (TL) generally declines with age and short TL relates to ageing-associated disorders and reduced survival in humans [3,4] and other organisms [5,6]. Given this relationship of telomeres with health and lifespan it is of importance to understand how variation in TL among individuals arises, which is already present early in life [7–9]. TL has a genetic basis, but heritability estimates for TL are highly variable [10]. Compared with other traits, inheritance of TL is also unusual in that the TL phenotype is directly expressed in the zygote without any effect of its own genome. This is because the zygote’s set of chromosomes carries the telomeres of the two parental gametes. Subsequently, during development of the embryo, different telomere maintenance and restoration mechanisms, under the control of multiple genes, potentially regulate TL, but this process is poorly understood [11]. In the course of early development, such mechanisms can potentially compensate fully for gamete derived differences in TL (as suggested by e. g. [12,13]), in which case the effect of gamete TL is transient (Fig 1A). Alternatively, differences in gamete TL are carried over to later life (as suggested by e. g. [14]; Fig 1B). The latter case would imply the inheritance of parental TL, which is independent of DNA sequence variation (in vertebrates (TTAGGG) n [15]), but a change in telomere sequence length (n). We interpret this as a form of epigenetic inheritance component on TL [16,17]. Note that this epigenetic inheritance mechanism differs from better known epigenetic mechanisms such as DNA methylation in that it does not affect the phenotype (TL) by modulating gene expression, but instead through direct inheritance of the phenotype itself and therefore has also been referred to as “epigenetic-like” [17]. Strongest evidence for an epigenetic mechanism of TL inheritance comes from studies that show a relationship between parental (usually paternal) age and offspring TL [18–23] with a particularly interesting example showing a cumulative effect over generations in humans [24]. In humans, where offspring TL increases with paternal age, this trend parallels a qualitatively similar change in sperm TL with age, which is generally assumed to underlie the TL increase in offspring [25]. However, studies of parental age effects in other species show mixed results and trends differ in direction between and within taxa [20,26]. More importantly, some critical uncertainties remain unresolved in any species. Firstly, studies to date are all cross-sectional [18–23], thus, comparing offspring of different parents that reproduced at different ages. Such cross-sectional trends may differ from age related changes within individual parents if, for example, individuals with long TL are more likely to reproduce at older ages, which is not unlikely given the positive correlation between human TL and reproductive lifespan [27,28]. Secondly, parental age effects on offspring TL may arise from effects of parental age on pre- and postnatal conditions prior to sampling. Because telomere attrition is highest early in life e. g. [29,30], these effects can be substantial, as illustrated by parental age effects on TL dynamics during the nestling phase in European shags Phalacrocorax aristotelis [31] and Alpine swifts Apus melba [21]. Lastly, due to their cross-sectional character, studies to date could not test whether changes in TL within parents over their lifetimes are predictive of changes in TL of the offspring in relation to parental age at conception. These points need to be resolved to establish whether the correlations between parental age and offspring TL can be attributed to epigenetic inheritance of TL, and before we can begin to understand why parental age effects on offspring TL appear to differ between and within taxa [20,26]. To investigate whether offspring TL changes with parental age at conception over the lifetime of individual parents we used our long-term, individual-based dataset of free-living jackdaws Corvus monedula. Telomere length was measured in nucleated erythrocytes using terminal telomere restriction fragment analysis [32] from multiple chicks of the same parents that hatched up to 9 years apart. As telomere attrition is highest early in life, we took blood samples for telomere analysis shortly after hatching, when the oldest chick in a brood was 4 days old. To test if TL was influenced by age-dependent parental care prior to sampling, we cross-fostered clutches between nests immediately after laying and tested whether foster parent age affected offspring TL. To investigate if the rate of telomere attrition within parents predicts the change in TL of the offspring they produce over consecutive years, we measured TL of the parents repeatedly over their lifetimes. For the first time, we here show that offspring TL declines as individual fathers age and that the change in TL over time in fathers is reflected in the TL of their offspring, which explains a substantial part of the telomere resemblance between fathers and offspring and can be interpreted as an epigenetic component in the inheritance of TL. Mother offspring resemblance on the other hand was independent of maternal age and within mother variation in TL was not associated with variation in the TL of her offspring.
To be able to separately evaluate between- and within-individual patterns of parental age, we used within-subject centering [33]. Instead of using age in our models, we used the mean age per individual over multiple years as one variable, and delta age, the deviation from that mean as a second variable. Thus, the coefficient of mean age estimates the parental age effect compared between individual parents, while the coefficient of delta age estimates the age effect on offspring TL within parents. As fathers aged, they produced offspring with 56±20 bp shorter TL for each additional year (variable ‘delta age father’ in Table 2A, Fig 2A), showing that offspring TL declined with paternal age at conception within individual males. This effect was not apparent when comparing offspring of different fathers reproducing at different ages (cross-sectional component of the statistical model, variable ‘mean age father’ in Table 2A). In contrast, there was no effect of maternal age on offspring TL (Table 2B), neither when compared cross-sectionally, between offspring of different mothers over age (mean age mother, Table 2B), nor within mothers as they age (delta age mother). The negative, non-significant effect of maternal age on offspring telomere length we observed (Table 2B) we attribute to the age of their mates, because pair bonds in jackdaws are maintained over many years (pers. obs.) and hence maternal and paternal age are correlated. This interpretation is confirmed by the finding that the observed maternal age estimate (delta age mother) is close to what would be expected based on the estimate found in fathers and the observed correlation of r = 0. 75 (n = 298) between maternal and paternal age (i. e. 0. 75 * 56 bp = 42 bp, which is very close to the estimate ± s. e. for delta mother age, which was 38±23 bp; Table 2; see also [23]). Thus, we conclude a maternal age effect on offspring TL other than through the age of the females’ mates to be unlikely. The decline in offspring TL with fathers’ age was lower than the rate of TL attrition in the fathers themselves (-56±20 versus -87±15 bp/year, respectively). Individual variation in telomere attrition slopes was negligible both between individual fathers (additional variance explained by random slopes 1%) and in their offspring produced across the fathers’ lifetimes as well (variance explained by random slopes 0. 3%). The paternal age effect on offspring TL could potentially be caused by age-dependent paternal care (e. g. age-related feeding of the incubating partner, or the chicks prior to sampling), if this affects telomere dynamics between conception and the sampling age of 4 days. We tested this hypothesis by exchanging clutches between pairs shortly after clutch completion. Our analysis is based on telomere data of 61 chicks that hatched from 31 cross-fostered clutches. In a first test, we added the age of foster father or mother to the model in Table 2A, and neither parental age significantly affected offspring TL (age foster father: 5. 6±28. 5, p = 0. 85; age foster mother: -9. 7±17. 4, p = 0. 58). To avoid basing a conclusion solely on a negative statistical result, in a second analysis we compared the estimate of the age of the caring father (i. e. the genetic father if not cross-fostered) on offspring TL with the estimate of the age difference between genetic and foster father (which is 0 in case of no cross-fostering or matching ages between genetic and foster father) on offspring TL. Both estimates were negative and very similar (Table 3, Fig 2B). Because the age of the caring father and the age difference between the caring father and the genetic father add up to the age of the genetic father for the cross-fostered offspring, the similarity of the estimates implies that there was no effect of age-related care between conception and sampling on offspring TL. While the estimate of the age difference did not quite reach statistical significance in a two-tailed test (p<0. 09), we consider the similarity of the estimates (10% difference) the more salient result. Thus, the older the father, the shorter the TL of his offspring, independent of the age of the male that cares for the eggs and offspring up to sampling. These results show that the paternal age effect on offspring TL is explained by the age of the genetic father and that the influence of the age of the foster fathers on offspring TL at age 4 days is negligible. The paternal age effect on offspring TL raises the question whether changes in paternal TL with age predict the change in early life TL of the offspring produced over the fathers’ lifetimes. We tested this by replacing the two age terms in the model in Table 2 by TL at conception (i. e. mean and delta TL) of the father in the year the offspring hatched. Fathers’ mean TL as well as delta TL were strongly and positively correlated with offspring TL (Table 4A, Fig 3). The effect of father’s mean TL on offspring TL can be attributed to additive genetic inheritance, possibly augmented by effects of a shared environment [10]. The effect of fathers’ delta TL on offspring TL cannot be attributed to a genetic effect, because delta TL refers to variation of TL within fathers over their lifetime. We therefore consider an epigenetic effect the most likely explanation for the effect of fathers’ delta TL on offspring TL. The variance in offspring TL explained by mean and delta TL of the father was 1. 87 and 0. 96 respectively. This indicates that 34% (0. 96 / 2. 83) of the variance in offspring TL that was explained by paternal TL can be attributed to the paternal-age related epigenetic effect. In agreement with our finding that maternal age did not affect offspring TL, when we performed the same analysis for mother TL, we found that maternal TL shortening (delta TL mother) was not related to the TL of her subsequent offspring, with a slope of the variable delta maternal age that was more than 90% lower than the comparable slope in males (Table 4B). However, mean maternal TL, reflecting a similarity between maternal and offspring TL per se (independent of a maternal TL change over time), based on a combination of additive genetic and age-independent maternal effects, was highly significant (Table 4B).
Resemblance of TL between parents and offspring is potentially due to a dual inheritance mechanism, with on the one hand a ‘classic’ additive genetic effect and on the other hand an epigenetic effect of variation in TL in the gametes that at least in part carries through to later life (Fig 1). Suggestive evidence for an epigenetic contribution to the inheritance of TL comes from studies showing a paternal age effect on offspring TL, but available results are based on cross-sectional analyses [18–23]. Using a unique longitudinal dataset on free-living birds, and a high precision TL measurement technique (CV within individuals <3%), we show for the first time that offspring TL changes with age within individual fathers (i. e. longitudinally). We used a cross-foster experiment to test whether the paternal age effect may be due to paternal age-dependent parental care prior to offspring sampling. This showed that the paternal age effect is already present at laying. Mother age was not significantly associated with offspring TL, and the non-significant estimate of the maternal age effect matched almost exactly the expected estimate based on the observed paternal age effect in combination with the correlation between the ages of pair members. Thus, we conclude that offspring TL declined with parental age within individual fathers, but not mothers. The parental sex dependent age effect on offspring TL is in agreement with most other studies [18–23,25,34], and is usually attributed to the different replicative history of the gametes of the two sexes. Male gametes are newly formed throughout reproductive life, while a female’s complete stock of gametes is formed before birth [35,36]. Hence TL of female gametes is less prone to changes with female age compared to TL of male gametes [37–39, but 40]. This is not to say that there is no epigenetic inheritance of TL through the female line, but only that its contribution to offspring TL does not depend on mothers’ age. While we consider epigenetic inheritance of TL via a carry-over effect from paternal gamete TL the most parsimonious explanation of our findings, we acknowledge that we cannot yet fully exclude other mechanisms. There is some scope for females to modulate the contents of their eggs, which may affect TL dynamics [41]. Thus, it remains a possibility that females adjusted the content of their eggs in response to the age of their partner in a way that causes the paternal age effect on the TL of their offspring. However, if there were such an effect, one would perhaps also expect it to be expressed in egg volume (which varies considerably in jackdaws), but there was no evidence that females adjusted the volume of their eggs to the age of their partner (p = 0. 35, n = 683 clutches, model including female identity and year as random effects). Another mechanism we cannot rule out is paternal age dependent expression of genes that control telomere dynamics of offspring. However, genetic influences on telomere dynamics are modest compared to environmental influences or heritability of TL itself [42], making it unlikely that this hypothetical mechanism explains a substantial part of the paternal age effect. We tentatively estimated the relative contributions of additive genetic and epigenetic effects to the resemblance between males and their offspring using a statistical model in which we separated between- and within-individual variation in parental TL as predictors of offspring TL. In this model, the within-male component (‘delta TL father’, Table 4A) shows the strong epigenetic effect over the years within males on their offspring, while the between male component (‘mean TL father’) shows the putative additive genetic effect on offspring TL. When comparing the relative contributions of the two inheritance mechanisms, it appeared that 34% of the variance explained by paternal TL can be attributed to the epigenetic effect. Telomere loss within mothers (‘delta TL mother’) was unrelated to the TL of offspring produced over years (Table 4B). Estimates of the between-male effect (0. 26±0. 08, ‘mean TL father’, Table 4A) and the between-female effect (0. 46±0. 09, ‘mean TL mother’, Table 4B) together equate to a narrow sense heritability of jackdaw TL of 0. 72, which is similar to results observed in humans [43] and within the range observed in other vertebrates [10] and is in line with other studies on birds estimating higher similarity between mothers and offspring [44,45]. We stress however that we measured telomere length in parental blood and not in sperm and that the estimates for the additive genetic and the epigenetic effects are tentative. Firstly, with respect to the additive genetic effect, it is of importance that shared environment effects are not controlled for in the present analysis. We note however that a more extensive analysis using multigenerational pedigree information and controlling for shared environmental effects [46] yielded a very similar estimate of the narrow sense heritability of TL in our study population (Bauch et al. in prep). Secondly, the variance in TL between males is not only of genetic origin, given that in addition there appears to be an epigenetic contribution to the between-male variance. Thus, the effect of ‘mean TL father’ (Table 4A) will to an unknown extent contribute to the epigenetic effect, as well as heterogeneity of sperm TL in ejaculates. Hence the epigenetic contribution to the resemblance between father and offspring TL will be more than the 34% we estimated based on parent-offspring regression over a single generation. Narrow sense heritability of human TL has been estimated using monozygotic and dizygotic twins [e. g. 47], assuming that a weaker resemblance between dizygotic twins compared to monozygotic twins can be attributed to the difference in genetic relatedness. However, as monozygotic twins develop from a single zygote, and hence from a single sperm cell and oocyte, the difference in resemblance within a monozygotic versus a dizygotic twin pair may in part be due to an epigenetic effect of having developed from the same or different gametes [14]. This process would lead to an overestimation of the narrow sense heritability compared to techniques that do not depend on twins. The direction of the paternal age effect in jackdaws (decreasing) is opposite to the direction of the paternal age effect in humans and chimpanzees (increasing) [20]. Assuming that paternal age effects in humans and chimpanzees [20] on the one hand and several bird species (including our study species) [20–23] and lab mice [34] on the other hand all reflect paternal age effects on sperm TL, this raises the question why these age effects on sperm TL are in opposite directions. Seasonality of reproduction may well play a role, with species that produce sperm for a small part of the year having less need to maintain sperm TL than species with year-round sperm production [20]. The lengthening of TL in human sperm with age has been interpreted as the result of an overshoot in telomere maintenance [25] that can be viewed as a safety margin in the maintenance process. Such a safety margin can be expected to be larger when the rate of sperm production and hence telomere attrition is higher. This may explain why chimpanzees, with a higher sperm production rate than humans, due to their promiscuous mating system, show a steeper paternal age effect on offspring TL compared to humans [48]. Information on the sign of the association between paternal age and offspring TL in strongly seasonal mammal species and / or continuously reproducing bird species would allow a test of this hypothesis. The epigenetic inheritance of TL potentially has more general implications. Parental age at conception has previously been shown to have negative effects on offspring fitness prospects in diverse taxa, a phenomenon known as the Lansing effect [22,49–52]. The underlying mechanisms are likely to be diverse, but in taxa where the paternal age effect on offspring TL is negative, given that TL predicts survival in wild vertebrates [6], and TL early in life correlates strongly with TL in adulthood in jackdaws [7], offspring born to older fathers may have a shorter life expectancy due to their epigenetically inherited shorter TL. A further implication is that there may be cumulative changes in TL over multiple generations [24]. This could lead to population level changes in TL when the age structure of the population changes, as has for example been observed in birds in response to urbanisation [53]. A population level change in TL may in itself have further demographic consequences [54], providing a positive or negative feedback, depending on whether increasing paternal age has a positive or negative effect on offspring TL.
Data were collected under license of the animal experimentation committee of the University of Groningen (Dierexperimenten Commissie, DEC, license numbers: 4071,5871,6832A). License was awarded in accordance with the Dutch national law on animal experimentation (“Wet op de dierproeven”) and research was carried out following the guidelines of the Association for the Study of Animal Behaviour (ASAB) [55]. Life-history data and blood samples originate from an individual-based long-term project on free-living jackdaws Corvus monedula breeding in nest boxes south of Groningen, the Netherlands (53. 14° N, 6. 64° E). Jackdaws produce one brood per year with mostly 4 or 5 chicks. They are philopatric breeders and socially monogamous with close to zero extra-pair paternity as shown in different populations [56,57]. Females incubate the eggs, while males feed their female partners. Chick provisioning is shared by the sexes. Each year, during the breeding season around the hatching date nest boxes were checked daily for chicks. Freshly hatched chicks were marked by clipping the tips of the toenails in specific combinations and therefore the exact ages of offspring were known. Between 2005 and 2016,715 jackdaw chicks were blood sampled when the oldest chick (s) was (were) 4 days (note that chicks hatch asynchronously). These chicks originated from 298 nests, of 197 different fathers, whereof 66 were blood sampled repeatedly over years (max. difference of age between offspring 8 years) and 194 different mothers, whereof 62 were blood sampled repeatedly over years (max. difference of age between chicks 9 years; see Table 1 for more information). 61 chicks (that contributed telomere data) hatched from 31 cross-fostered nests, i. e. eggs were exchanged between nest boxes (selected for equal clutch sizes and laying dates (or up to one day difference), but otherwise randomly) soon after clutch completion. 54 (89%) of those chicks were fostered by a father of different age. Jackdaws in this project are marked with a unique colour ring combination and a metal ring. Parents were identified by (camera) observation during incubation and also later during chick rearing when caught for blood sampling (by puncturing the vena brachialis). Unringed adults were caught, ringed and assigned a minimum age of 2 years, as this is the modal recruitment age of breeders that fledged in our study colony. All jackdaws were of known sex (molecular sexing [58]). Blood samples were first stored in 2% EDTA buffer at 4–7°C and within 3 weeks snap frozen in a 40% glycerol buffer for permanent storage at -80°C. Terminally located telomere lengths were measured in DNA from erythrocytes performing telomere restriction fragment analysis under non-denaturing conditions [29]. In brief, we removed the glycerol buffer, washed the cells and isolated DNA from 5 μl of erythrocytes using CHEF Genomic DNA Plug kit (Bio-Rad, Hercules, CA, USA). Cells in the agarose plugs were digested overnight with Proteinase K at 50°C. Half of a plug per sample was restricted simultaneously with HindIII (60 U), HinfI (30 U) and MspI (60 U) for ~18 h in NEB2 buffer (New England Biolabs Inc. , Beverly, MA, USA). The restricted DNA was then separated by pulsed-field gel electrophoresis in a 0. 8% agarose gel (Pulsed Field Certified Agarose, Bio-Rad) at 14°C for 24h, 3V/cm, initial switch time 0. 5 s, final switch time 7. 0 s. For size calibration, we added 32P-labelled size ladders (1kb DNA ladder, New England Biolabs Inc. , Ipswich, MA, USA; DNA Molecular Weight Marker XV, Roche Diagnostics, Basel, Switzerland). Gels were dried (gel dryer, Bio-Rad, model 538) at room temperature and hybridized overnight at 37°C with a 32P-endlabelled oligonucleotide (5’-CCCTAA-3’) 4 that binds to the single-strand overhang of telomeres of non-denatured DNA. Subsequently, unbound oligonucleotides were removed by washing the gel for 30 min at 37°C with 0. 25x saline-sodium citrate buffer. The radioactive signal of the sample specific TL distribution was detected by a phosphor screen (MS, Perkin-Elmer Inc. , Waltham, MA, USA), exposed overnight, and visualized using a phosphor imager (Cyclone Storage Phosphor System, Perkin-Elmer Inc.). We calculated average TL using ImageJ (v. 1. 38x) as described by Salomons et al. [29]. In short, for each sample the limit at the side of the short telomeres of the distribution was lane-specifically set at the point of the lowest signal (i. e. background intensity). The limit on the side of the long telomeres of the distribution was set lane-specifically where the signal dropped below Y, where Y is the sum of the background intensity plus 10% of the difference between peak intensity and background intensity. We used the individual mean of the TL distribution for further analyses. Samples were run on 92 gels. Repeated samples of adults were run on the same gel, chicks were spread over different gels. The coefficient of variation of one control sample of a 30-day old jackdaw chick run on 26 gels was 6% and of one control sample of a goose, with a TL distribution within a similar range, run on 31 other gels was 7%. The within-individual coefficient of variation for samples run on the same gel was <3% [7] and the within-individual repeatability of TL was estimated to be 97% [59]. The relationships between parental age or parental TL and early-life TL of offspring were investigated in a linear mixed effects model framework using a restricted maximum-likelihood method (testing specific predictions). To be able to separately evaluate between- and within-individual patterns of parental age or parental TL, we used within-subject centering [33]. Thus, instead of father age, mother age or father TL, mother TL per se we introduced the mean value per individual over (if available) multiple years and delta age or delta TL, the deviation from that mean, respectively. To account for (genetic and potential other) similarities in TL between offspring of the same father or mother, we included father ID or mother ID as random effect in the model. As the dataset contains also siblings raised in the same nest, we additionally added a random effect of nest ID as a nested term in father ID or mother ID to the models investigating paternal or maternal age effects on TL, respectively. The age of chicks at sampling differed slightly (2–4 days) and as TL shortens with age [7], we included their age (in days) as a covariate. Offspring sex was never significant and was therefore excluded from the final models. We added gel ID as random effect. Analyses were performed separately for fathers and mothers as their ages are correlated. The cross-foster experiment was designed to test for potential effects of parental age on early-life telomere attrition between egg laying and sampling (age 2–4 days). First, we modified the linear mixed effect model with offspring TL as dependent variable testing for paternal age effects (see above) by adding the age of the foster father or mother as covariate. Second, in a linear mixed model with offspring TL as dependent variable, we included both the age of the father caring for the clutch after cross-fostering and the age difference between the genetic father and foster father as covariates (age genetic father-age foster father). When the paternal age effect is independent of age-dependent effects between conception and sampling, we predict the coefficients of the caring father’s age and the age difference between genetic father and foster father to be indistinguishable. This is so because the age of the male caring for the clutch, and the age difference between the genetic and the caring father add up to the age of the genetic father. In contrast, when the paternal age effect is entirely due to age-dependent paternal effects after laying, the coefficient will be the same, but opposite in sign. In case of a mixture of the two effects, the coefficient will be intermediate. In this analysis we used all offspring, i. e. also those that were not cross-fostered, and further included genetic father ID, nest ID, gel ID and year of telomere analysis as random effects, and offspring age at sampling as covariate. Statistics were performed using packages lme4 [60], lmerTest [61], MuMIn [62] in R (version 3. 3. 3) [63]. In the results mean ± standard error is given unless stated otherwise. | Telomeres are DNA-protein structures at chromosome ends and a short telomere length predicts reduced survival in humans, birds and other organisms. Variation in telomere length between individuals is thought to be largely of genetic origin, but telomere inheritance may be unusual because not only genes regulating telomere length are inherited, but a fertilised cell already has a telomere length (from the parental gametes). Using long-term individual-based data of jackdaw families (a small corvid species), we found that as fathers aged, they produced chicks with shorter telomeres. This suggests that telomere length inheritance has an epigenetic component. To investigate to what extent telomere length in the fertilised cell affects telomere length after birth, we compared telomere length over years within fathers with the telomere length of their consecutive offspring. This epigenetic component explained a substantial part (≥ one third) of the telomere length inheritance; whereas there was no such effect of maternal telomere length. The sex difference fits the idea that lifelong sperm formation leads to change in telomere length of the sperm cells, whereas female gametes are all formed before birth and their telomere length does not change over time. | Abstract
Introduction
Results
Discussion
Materials and methods | medicine and health sciences
animal genetics
chromosome structure and function
population genetics
reproductive physiology
germ cells
telomeres
clutches
epigenetics
population biology
bird genetics
sperm
telomere length
genetic polymorphism
chromosome biology
animal cells
cell biology
heredity
physiology
genetics
oviposition
biology and life sciences
cellular types
evolutionary biology
chromosomes | 2019 | Epigenetic inheritance of telomere length in wild birds | 7,542 | 308 |
Entomological surveys of Simulium vectors are an important component in the criteria used to determine if Onchocerca volvulus transmission has been interrupted and if focal elimination of the parasite has been achieved. However, because infection in the vector population is quite rare in areas where control has succeeded, large numbers of flies need to be examined to certify transmission interruption. Currently, this is accomplished through PCR pool screening of large numbers of flies. The efficiency of this process is limited by the size of the pools that may be screened, which is in turn determined by the constraints imposed by the biochemistry of the assay. The current method of DNA purification from pools of vector black flies relies upon silica adsorption. This method can be applied to screen pools containing a maximum of 50 individuals (from the Latin American vectors) or 100 individuals (from the African vectors). We have evaluated an alternative method of DNA purification for pool screening of black flies which relies upon oligonucleotide capture of Onchocerca volvulus genomic DNA from homogenates prepared from pools of Latin American and African vectors. The oligonucleotide capture assay was shown to reliably detect one O. volvulus infective larva in pools containing 200 African or Latin American flies, representing a two-four fold improvement over the conventional assay. The capture assay requires an equivalent amount of technical time to conduct as the conventional assay, resulting in a two-four fold reduction in labor costs per insect assayed and reduces reagent costs to $3. 81 per pool of 200 flies, or less than $0. 02 per insect assayed. The oligonucleotide capture assay represents a substantial improvement in the procedure used to detect parasite prevalence in the vector population, a major metric employed in the process of certifying the elimination of onchocerciasis.
Onchocerciasis, or river blindness has historically represented one of the most important neglected tropical diseases in the developing world as measured as a cause of socio-economic disruption [1]. It is also considered a candidate for elimination by the international community [2], [3]. As a result of these factors, the international community is currently supporting several programs whose goals are either to eliminate the disease as a public health problem, or to locally eliminate the causative agent of the disease, Onchocerca volvulus. These include the Onchocerciasis Elimination of the Americas (OEPA), the African Programme for Onchocerciasis Control (APOC), and the Ugandan Onchocerciasis Elimination Program (UOEP). Entomological criteria play an important role in the elimination criteria recommended by the World Health Organization (WHO) [4] and those currently utilized by OEPA [5]. Entomological data play an especially important role in the certification of elimination following the cessation of treatment, as the prevalence of infective stages of the parasite in the fly population is the timeliest measure of transmission in a given area. However, demonstrating that transmission is interrupted requires that large numbers of flies be tested. For example, current OEPA guidelines require that the prevalence of flies carrying infective larvae (L3) be less than 1/2000 in every sentinel community for transmission to be interrupted [5]. In order to be able to state with a 95% confidence that the prevalence of infective flies is less that 1/2000 requires examining approximately 6000 flies from each sentinel community. Examining such large numbers of insects using conventional methods (dissection) is impractical. For this reason, the current guidelines recommend the use of pool screening PCR based methods to conduct the entomological studies necessary to document transmission interruption [4]. Currently, the accepted method for pool screening vector black flies to detect O. volvulus relies upon screening DNA prepared from fly pools with a PCR assay targeting a repeated sequence family (the O-150 repeat [6]) specific for parasites of the genus Onchocerca. Algorithms have been developed that permit one to derive a point estimate of the prevalence of infection in the fly population (and an associated confidence interval) from the number of PCR positive pools and the number of flies contained in each pool [7]. Furthermore, because the infective stage of the parasite is the only form found in the black fly head capsule, separated pools of heads and bodies may be screened to obtain estimates of the prevalence of infective flies (flies with infective stages in their head) and the prevalence of infected flies (flies with immature larval stages in their bodies). This approach has been used to monitor transmission of O. volvulus in many foci of Latin America and Africa [8]–[11], as well as to certify the interruption of transmission in foci on both continents [5], [12]–[14]. Previous modeling studies have shown that increasing pool sizes has relatively little effect on the accuracy of the estimate of prevalence of infection obtained, so long as the proportion of positive pools remains less than the majority of pools screened [15]. Thus, in situations where pool screening is used to certify transmission interruption (where infective flies are extremely rare or non-existent) the pool size is only limited by the biochemical constraints of the assay. The current method of DNA extraction for the O-150 PCR assay is based upon adsorption to a silica matrix [16]. This preparation results in DNA samples that still contain inhibitors of the PCR, limiting the number of flies that may be included in each pool. Currently, pool sizes are limited to 50 individual heads or bodies (in the case of flies from Latin America) [9] or 100 individual heads or bodies (in the case of flies from Africa) [7]. Developing alternative methods to prepare DNA that would permit an increase in the maximum number of heads or bodies in each pool would decrease the cost and effort necessary to screen the requisite large numbers of flies necessary to certify transmission interruption. Magnetic bead based purification protocols have been developed for many different pathogens. Most of these involve direct capture of the pathogen using beads coated with pathogen-specific antibodies. This method, known as immunomagnetic separation (IMS), has been successfully used to purify and concentrate viruses [17], bacteria [18], [19] and fungal [20] pathogens. Similarly, methods have been developed which use oligonucleotides to magnetically purify pathogen genomic DNA [21]. Here, we describe a magnetic bead capture method to isolate O. volvulus DNA from homogenates prepared from pools of Latin American and African Simulium vector black flies. This method is shown to be an improvement upon the current DNA purification method utilizing organic extraction and silica adsorption.
Simulium ochraceum s. l. females were collected in public areas of the community of José María Morelos y Pavón, Chiapas, México between the hours of 0700 and 1000. Previous studies have demonstrated that the majority of flies captured during this period were nulliparous, and the risk of infection was therefore minimal. Simulium damnosum s. l. were obtained from breeding sites on public land located near the communities of Bodajugu and Sakora. These communities are located in the Region des Cascades in Southwestern Burkina Faso. This region is located within the area of the former Onchocerciasis Control Programme in West Africa, where onchocerciasis has been eliminated as a public health problem. O. volvulus L3 were obtained from experimentally infected Simulium damnosum s. l. flies 7 days after infection with skin microfilariae. The flies were kept at 25°C and 80% relative humidity to allow the microfilariae to develop into L3. Larvae were isolated from the flies by dissection into dissecting medium (IMDM+10% FCS+2x Penicillin-Streptomycin-Fungizone) using a dissecting microscope. The cleaned larvae were frozen in 9% DMSO, 4 mM PVP, 10% FCS in Grace medium using Bio-Coll (freezing to −40°C at 1°C/minute followed by 30 minutes at −40°C) and then transferred to liquid nitrogen for long-term storage. The parasite material was prepared in the Tropical Medicine Research Station, Kumba, Cameroon, and is being stored at the New York Blood Center. Pools containing varying numbers of black flies were prepared and the heads and bodies separated by freezing and agitation, as previously described [7]. A single O. volvulus L3 was added to each pool. Head and body pools were placed in a 1. 5 ml microcentrifuge tube and purified using magnetic silica coated beads (Machery-Nagel GmbH & Co, Bethlehem, PA, USA) following the instructions provided by the manufacturer. In brief, the pools were homogenized in 200 µl of T1 buffer, 25 µl of proteinase K solution provided in the kit was added and the homogenates were incubated at 56°C for 30 minutes. The homogenates were subjected to centrifugation at 13,400×g for 5 minutes at room temperature. A total of 225 µl of the supernatant was transferred to a fresh tube containing 24 µl B-beads (Machery-Nagel) and 360 µl MB2 buffer (Machery-Nagel), and the tube shaken for 5 minutes at room temperature. The magnetic beads were isolated by placing the tubes in a six tube magnetic separator (Dynal MPC-S; Invitrogen). The supernatants were removed and discarded, and 600 µl of MB3 wash buffer was added to each sample. The bead/DNA complexes were washed by shaking for 5 min at room temperature. The beads were collected in the magnetic separator as before, and the washing procedure repeated with successive washes with 600 µl of MB4 and MB5 wash buffers (Machery-Nagel). Following the wash in the MP5 buffer, the beads were exposed to air for one minute to permit the traces of ethanol to evaporate. DNA was eluted from the beads by the addition of 100 µl of elution buffer. The beads were shaken for 5 minutes at room temperature to elute the DNA, and the beads removed by placing the tubes in the magnetic separator. The supernatant containing the purified DNA was then transferred to a fresh tube. Pools of spiked heads and bodies were prepared as described above. The head and body pools were homogenized in 500 µl of 10 mM Tris-HCl (pH 8. 0) 1 mM EDTA, and proteinase K added to a final concentration of 2 mg/ml. The homogenates were incubated at 56°C for 2 hours, and dithiothreitol added to a final concentration of 20 mM. The samples were heated to 100°C for 30 minutes and subjected to three freeze-thaw cycles. The homogenates were subjected to centrifugation at 13,400×g for 5 minutes and the supernatant placed into a new tube. The solutions were brought to a final concentration of 100 mM Tris-HCl (pH 7. 5) 100 mM NaCl. A total of 5 µl of a 0. 5 µM solution of OVS2-biotin primer (5′B-AATCTCAAAAAACGGGTACATA-3′, where B = biotin) was added to each sample. The samples were then heated to 95°C for three minutes and allowed to cool slowly to room temperature. While the probe was annealing to the DNA in the solution, 10 µl of Dynal M-280 strepavidin coated beads (Invitrogen) were placed in a single well of a 96 well tissue culture plate. The plate was placed on a magnetic capture unit (Dynal MPC-96, Invitrogen) and the beads collected for 2 minutes. The beads were then washed five times with 200 µl binding buffer (100 mM Tris-HCl (pH 7. 5) 100 mM NaCl) per wash, resuspended in 10 ul and added to the sample. The samples were incubated at 4°C overnight on a roller to permit the oligonucleotide-DNA hybrids to bind to the beads. The samples were placed in the magnetic separator for two minutes to capture the beads and the supernatant discarded. The beads were resuspended in 150 µl of binding buffer by pipetting, and the beads captured by placing the tubes in the magnetic separator for two minutes. The wash step was repeated five times. The beads were then resuspended in 20 µl of sterile water, heated to 80°C for 2 minutes and cooled rapidly on ice for two minutes. The beads were removed by placing the tubes in the magnetic capture apparatus, and the supernatant containing the purified DNA transferred to a new tube. A total of 2. 5 µl of the purified genomic DNA was used as a template for the PCR amplifications carried out in a total volume of 50 µL containing 0. 5 µmol/L of O-150 primer (5′-GATTYTTCCGRCGAANARCGC-3′) and 0. 5 µmol/L of biotinylated O-150 primer (5′-B-GCNRTRTAAATNTGNAAATTC- 3′, where B = biotin; N = A, G, C, or T; Y = C or T; and R = A or G). Reaction mixtures also contained 60 mM Tris-HCl, (pH 9. 0), 15 mM (NH4) 2SO4,2 mM MgCl2,0. 2 mM each of dATP, dCTP, dGTP and dTTP, and 2. 5 units of Taq polymerase (Invitrogen). Cycling conditions consisted of five cycles of one minute at 94°C, two minutes at 37°C, and 30 seconds at 72°C, followed by 35 cycles of 30 seconds each at 94°C, 37°C, and 72°C. The reaction was completed by incubating at 72°C for six minutes. Amplification products were detected by PCR enzyme-linked immunosorbent assay (ELISA), essentially as previously described [10]. Briefly, 5 µl of each PCR reaction was bound to a streptavidin-coated ELISA plate, and the DNA strands denatured by treatment with alkali. The bound PCR fragments were then hybridized to a fluorescein-labeled O. volvulus-specific oligonucleotide probe (OVS2: 5′-AATCTCAAAAAACGGGTACATA-FL-3′), and the bound probe detected with an alkaline phosphatase-labeled anti-fluorescein antibody (fragment FA; Roche Diagnostics). Bound antibody was detected using the ELISA amplification reagent (BluePhos) kit from KPL (Gaithersburg, USA) following the manufacturer' s instructions. Color development was stopped by the addition of 100 µl AP stop solution, and the plates read in an ELISA plate reader set at 630 nm. Samples were scored positive if their optical density exceeded either the mean plus three standard deviations of ten negative control wells run in parallel or 0. 1, whichever was greater.
An initial series of experiments were carried out with pools of heads and bodies of S. ochraceum s. l. (a major Latin American vector of onchocerciasis). Pools containing varying numbers of heads of bodies were spiked with a single O. volvulus L3, and DNA prepared from the pools using either the conventional method of organic extractions followed by adsorption to a silica matrix, or by oligonucleotide capture of O. volvulus genomic DNA followed by magnetic purification of the captured oligonucleotide-DNA complexes. The conventional method consistently produced a positive signal in pools containing up to 50 heads (Table 1). Pools containing greater than 50 heads were not positive in the assay. All samples containing 50,100 or 150 S. ochraceum s. l. bodies were positive, while none of the pools containing 200 bodies were positive (Table 1). In contrast, positive signals were obtained in all pools containing up to 200 heads or bodies in the assays performed on the oligonucleotide capture purified DNA samples (Table 1). The preliminary experiments suggested that the oligonucleotide capture method was capable of detecting one L3 in pools of up to 200 heads or bodies. To further explore the sensitivity of the assay, the experiment was repeated using 10 separate pools containing 200 heads or bodies spiked with a single L3. All pools were found to be positive, suggesting that the oligonucleotide capture assay was capable of consistently detecting a single L3 in pools of up to 200 heads or bodies of S. ochraceum s. l. (Table 2). Previous studies had demonstrated that the conventional silica adsorption method was capable of detecting a single infected S. damnosum s. l. (the major African vector of onchocerciasis) fly in a pool containing up to 99 uninfected flies [7]. To determine if the performance of the oligonucleotide capture assay was similar when applied to S. damnosum s. l. , the spiking experiments were repeated employing pools containing 200 S. damnosum s. l. heads or bodies. All spiked pools were found to be positive (Table 2), suggesting that the capture assay preformed equally well on both African and Latin American vectors of onchocerciasis. For the capture assay to be cost effective, it should be competitive with the cost of the conventional silica adsorption assay. The two assays require roughly the same amount of technical time, so labor costs may be assumed to be equivalent per sample for the two assays. However, because it will be possible to increase the size of the pools 2–4 fold when using the capture assay, a reduction in labor costs of between 50% and 75% would be realized when costs are considered on a per-fly-tested basis. Similarly, the per-sample cost of carrying out the conventional assay is roughly $2. 22 per pool, while the cost of the magnetic bead assay is $3. 81 per pool. However, because the magnetic bead permits more flies to be tested per pool, cost savings in reagents are realized when the costs are amortized on a per fly basis (Table 3).
Previous studies have demonstrated that the algorithms used to predict the prevalence of infection in a population from data derived from screening pools of samples are relatively insensitive to the size of the pool, so long as the proportion of positive pools does not represent a substantial majority of the samples screened [15]. Thus, the size of pools used in a pool screening protocol is more likely to be limited by the biochemistry of the detection assay than by the underlying statistical uncertainties associated with screening pools of samples. This is particularly true when positive samples are extremely rare, as in the case when monitoring for transmission interruption. The data presented above suggest that the oligonucleotide capture method of purifying O. volvulus DNA is superior to the conventional silica adsorption method. Mixing experiments have demonstrated that PCR inhibitors carried through the silica adsorption process limits the size of the pools that may be screened to 50 individual flies for S. ochraceum and 100 for S. damnosum s. l (data not shown). The oligonucleotide capture method appears to result in DNA preparations that are freer of PCR inhibitors than are those prepared using silica adsorption. The practical result of this improvement is that it permits a 2 to 4-fold increase in the number of black fly heads or bodies that can be included in a single pool. This increase in pool size results in a dramatic cost savings in the per-unit cost of the O-150 pool screen assay. Using the oligonucleotide capture assay, labor costs are reduced by 50–75%, while the overall cost of reagents needed is $3. 81 per pool of 200 flies or less than $0. 02 per individual fly. The decrease in cost and corresponding increase in the efficiency of the assay will make it more practical to screen the large numbers of flies necessary to demonstrate transmission interruption and to certify elimination of onchocerciasis. Because the oligonucleotide capture assay is a modification of the conventional silica adsorption assay, the equipment required to carry out both assays is quite similar. The only additional equipment necessary when replacing the conventional assay with the oligonucleotide assay is the magnetic capture apparatus. This unit is relatively inexpensive, costing less than $579 (Invitrogen' s list price). This would be recovered in reagent costs alone after screening just 114 pools of flies. The O-150 PCR is based upon the amplification of a genus specific tandemly repeated DNA sequence present in the genome of Onchocerca parasites [22]. Thus, the standard O-150 PCR will amplify sequences present in all Onchocerca species, including Onchocerca ochengi, a cattle parasite that is sympatric with O. volvulus in sub-Saharan Africa. Currently, the O-150 PCR is made species or strain specific by modifying the amplification conditions to limit amplification to species specific members of the O-150 repeat family [23] or by the use of species or strain specific probes to detect the resulting amplicons [8]. However the oligonucleotide used in the capture assay (OVS2) has previously been shown to hybridize specifically to O. volvulus under appropriate conditions [22]. It is therefore possible that modification of the current capture conditions could result in a purification process that would specifically capture O. volvulus but not O. ochengi genomic DNA. This would provide an extra layer of specificity to the assay when it is employed in areas where O. volvulus and O. ochengi are sympatric. | The absence of infective larvae of Onchocerca volvulus in the black fly vector of this parasite is a major criterion used to certify that transmission has been eliminated in a focus. This process requires screening large numbers of flies. Currently, this is accomplished by screening pools of flies using a PCR-based assay. The number of flies that may be included in each pool is currently limited by the DNA purification process to 50 flies for Latin American vectors and 100 flies for African vectors. Here, we describe a new method for DNA purification that relies upon a specific oligonucleotide to capture and immobilize the parasite DNA on a magnetic bead. This method permits the reliable detection of a single infective larva of O. volvulus in pools containing up to 200 individual flies. The method described here will dramatically improve the efficiency of pool screening of vector black flies, making the process of elimination certification easier and less expensive to implement. | Abstract
Introduction
Materials and Methods
Results
Discussion | medicine
infectious diseases
neglected tropical diseases
onchocerciasis | 2012 | Oligonucleotide Based Magnetic Bead Capture of Onchocerca volvulus DNA for PCR Pool Screening of Vector Black Flies | 5,211 | 233 |
The near exclusive use of praziquantel (PZQ) for treatment of human schistosomiasis has raised concerns about the possible emergence of drug-resistant schistosomes. We measured susceptibility to PZQ of isolates of Schistosoma mansoni obtained from patients from Kisumu, Kenya continuously exposed to infection as a consequence of their occupations as car washers or sand harvesters. We used a) an in vitro assay with miracidia, b) an in vivo assay targeting adult worms in mice and c) an in vitro assay targeting adult schistosomes perfused from mice. In the miracidia assay, in which miracidia from human patients were exposed to PZQ in vitro, reduced susceptibility was associated with previous treatment of the patient with PZQ. One isolate (“KCW”) that was less susceptible to PZQ and had been derived from a patient who had never fully cured despite multiple treatments was studied further. In an in vivo assay of adult worms, the KCW isolate was significantly less susceptible to PZQ than two other isolates from natural infections in Kenya and two lab-reared strains of S. mansoni. The in vitro adult assay, based on measuring length changes of adults following exposure to and recovery from PZQ, confirmed that the KCW isolate was less susceptible to PZQ than the other isolates tested. A sub-isolate of KCW maintained separately and tested after three years was susceptible to PZQ, indicative that the trait of reduced sensitivity could be lost if selection was not maintained. Isolates of S. mansoni from some patients in Kisumu have lower susceptibility to PZQ, including one from a patient who was never fully cured after repeated rounds of treatment administered over several years. As use of PZQ continues, continued selection for worms with diminished susceptibility is possible, and the probability of emergence of resistance will increase as large reservoirs of untreated worms diminish. The potential for rapid emergence of resistance should be an important consideration of treatment programs.
Schistosomiasis is one of the most common human parasitic diseases in the world. An estimated total of 207 million persons are infected worldwide, 97% of which are on the African continent [1]. The chronic and debilitating nature of the disease results in high costs in public health and economic productivity in developing countries, and has prompted the initiation of large scale control programs [2]. Schistosoma mansoni is one of the most common etiological agents for human schistosomiasis, and is estimated to infect more than 83 million humans in 54 countries [3]. Praziquantel (PZQ) is the least expensive, easiest to use and most readily available of all currently available schistosomicides [4]. It is highly effective against all schistosome species that are known to infect humans and is well-tolerated, making it suitable for mass treatment campaigns. These campaigns are particularly targeted at school-age children who represent the most heavily infected segment of the population [5], [6]. Although such programs have immediate and significant salutary effects, three general concerns are that they 1) inevitably leave some individuals untreated; 2) some individuals are treated but left uncured; and 3) they do not interrupt transmission, making re-infection a reality [7]. Another major concern of all anthelminthic and antibiotic drugs is the potential for resistance to develop and spread throughout the population, making the drug useless for treatment and control. Even though PZQ efficacy is generally high, reported cure rates are variable ranging from 60 to 95% [8]–[10]. Potential explanations for incomplete cures are use of doses that are actually sub-curative in people, or the presence of drug resistance traits present in natural populations of worms. The extensive use of PZQ for over 20 years in some African nations has raised concern regarding the selection of drug resistant worms [5], [11]–[13]. Artificial selection in the laboratory has produced resistant strains of S. mansoni in only 2 generations of repeated exposure to sub-lethal doses of the drug in mice [11], thus demonstrating that resistance is more than a hypothetical possibility. Low cure rates in response to PZQ in the field appeared 10–15 years after the beginning of its use on a mass scale in Egypt and after a recent introduction of the worms in Senegal [12]–[16]. In both of these cases, worms from the uncured patients were also less susceptible to PZQ when tested in a mouse model [17]. Therefore, traits of the worms themselves led to PZQ failure, although other factors were also suspected to contribute including host factors, heavy worm burdens, and pre-patent infections [13], [17]. Reports of difficulties in obtaining cures among travelers with schistosomiasis [18] further underscore the need to remain vigilant. Most published discussions of this topic conclude that convincing evidence for the clinically relevant emergence of PZQ resistance in the field is still lacking [4], [5], [16], [19], [20]. Once drug resistance reaches clinical relevance it becomes a difficult problem to solve. Therefore, vigilant monitoring aimed at the prevention of clinical drug resistance is critical to treatment and control of infectious diseases. Measuring the impact of PZQ on adult worms harbored by human subjects is at best an indirect process in that it relies on the relatively insensitive method of measuring reduction in schistosome egg excretion in treated individuals [21]. Also, several factors influence PZQ' s effects such as variable pharmacokinetics in different individuals, differences in immune responses to the worms, maturity of worms, and genetic variability of the worms themselves, both among individual hosts and geographic regions [13], [16], [20], [22], [23]. Obtaining an isolate from human patients for further in vivo tests in laboratory hosts such as mice can address some of these concerns [13], [17] and can be complemented with in vitro testing that removes host-induced effects. In vitro tests for PZQ sensitivity have been developed for use on schistosomes [24]–[26]. Within a given isolate, the susceptibility of miracidia correlates with the susceptibility of adult worms [26]. Using miracidia in these assays potentially allows the testing of schistosomes derived from a large number of human infections. The aim of this study was to evaluate the level of susceptibility to PZQ among a natural population of S. mansoni that were derived from Kenyan car washers and sand harvesters occupationally exposed to schistosome infection in Lake Victoria. These people have been enrolled in a longitudinal study designed to investigate immunologically-based host resistance to S. mansoni re-infection, and many of the participants have been treated with PZQ on multiple occasions throughout the study [27]–[30]. We first used an in vitro assay with miracidia to evaluate PZQ susceptibility of S. mansoni from several patients. Then, an isolate of S. mansoni was established in laboratory hosts from one of the patients, who had never fully cured after initial or several subsequent PZQ treatments. The susceptibility of adult worms of this and other Kenyan isolates and of two laboratory stocks of S. mansoni were compared using both conventional in vivo trials in mice, and with an in vitro assay with adult worms.
The patient-derived S. mansoni isolates were recovered from eggs in the fecal samples of adult males working as car washers or sand harvesters in or near Kisumu, western Kenya. The car washers use the shallow water along the Lake Victoria shoreline to wash cars and trucks and so are repeatedly in contact with water in an area where snails infected with S. mansoni snails have been repeatedly found [31]. The sand harvesters also experience extensive contact with lake water as they shovel sand from the submerged lake bottom into boats, often exposed to water from the chest down for hours each day. Since June 1995, the car washers have been continuously enrolled in several longitudinal studies that evaluate resistance to S. mansoni infection [27]–[30]. The sand harvesters were enrolled in similar studies starting in March 2005. For each patient, information is available regarding intensity of infection and history of exposure to treatment with PZQ since the time of their enrollment in these studies. As part of the ongoing work associated with these studies, fecal samples derived from each patient are regularly tested for S. mansoni using the modified Kato-Katz technique. For the purposes of the present study, miracidia hatched from S. mansoni eggs were used either directly in the PZQ susceptibility assay for miracidia described below or, in two cases, were used to establish laboratory isolates. Miracidia obtained from eggs in positive fecal samples from individual patients were used to establish the following isolates in laboratory raised B. sudanica and six to eight week old outbred mice: KCW: The KCW isolate was established from a car washer who had received 18 PZQ treatments since the beginning of the study. Following initial and all subsequent treatments, egg counts in this patient had never fallen to zero [28], suggesting the responsiveness of worms from this patient to PZQ was worthy of further study. At the second generation, this isolate was divided and maintained as separate “sub-isolates” in New Mexico and in Kenya. Worms of the New Mexico sub-isolate were used in the experiments below unless otherwise noted. KSH: The KSH isolate was established from a sand harvester who had never been treated with PZQ prior to this study. KAS: This isolate was established from a human fecal sample collected near a stream (Asao) about 38 Km south east from Kisumu in 2006 (−0. 33256°S, 34. 99914°E). KNY: The KNY isolate was established from cercariae obtained from B. sudanica collected in Nyabera marsh (−0. 10971°S, 34. 77461°E) on the outskirts of Kisumu on the road to Ahero. Both KAS and KNY isolates were from areas not previously included in PZQ treatment campaigns of local school children. In addition to the isolates obtained from Kenya patients, the following isolates were used for comparison: PR1: This laboratory stock is originally of Puerto Rican origin. It has been maintained in the laboratory, including at the University of New Mexico, for more than 20 years. NMRI: This laboratory stock also originated from Puerto Rico and has been maintained at the Biomedical Research Institute, Rockville, Maryland (www. afbr-bri. com). To test the idea that miracidia derived directly from different patients may differ in their susceptibility to PZQ, particularly given that some patients (see KCW above) had never fully cured following treatment with PZQ, we employed a modified version of the in vitro technique developed by Liang et al. [26]. Freshly hatched miracidia from the stools of sand harvesters (n = 14) and car washers (n = 9) were placed in each well (3–6 miracidia in each well) in a 96-well plate in 40 µl of aged tap water. Each row of the plate (one group of miracidia) received a different concentration of PZQ: 0 M (control), 10−5 M, or 10−6 M, in a 40 µl volume. PZQ was prepared as a stock solution of 10−4 M in 1% DMSO, and the final concentration of DMSO was 0. 1% in all wells including the control wells. The mean number of groups of miracidia used per patient per concentration of PZQ was 10. 2 (range: 4–12) (dependent on the number of miracidia obtained from a fecal sample). This design was repeated with miracidia from three different fecal samples from each patient. Miracidia were observed with the aid of a dissecting microscope prior to, 10 min and 20 min after addition of PZQ, and the observer had no knowledge of the PZQ concentration of each well. The number of miracidia that were alive and dead in each well was recorded. Miracidia were considered dead if they remained immobile. We present only results using 10−5 M at the 20 min observation time to simplify the presentation and because they are representative of the combinations of observation times (10 or 20 minutes) and PZQ concentrations (10−5 M, or 10−6 M) used. A Generalized Estimating Equations (GEE) approach was used to fit a logistic regression model to the data using the Proc GENMOD procedure in SAS 9. 1. Separate models were fit for each concentration of PZQ (0 M, 10−6 M and 10−5 M). The number of previous PZQ treatments the subjects received was included in the model using an indicator variable that was set to ‘1’ if a patient had received no previous treatments, and‘0’ for patients who had received one or more previous treatments, and the time of exposure to the drug in vitro (0,10 or 20 min). For each isolate of S. mansoni examined, 10 to 30 outbred mice were infected with 200 cercariae per mouse. Seven and a half weeks after exposure, mice were randomly divided into two groups, one of which was given 1000 mg/kg PZQ dissolved in 2% Cremaphor EL over 4 consecutive days (250 mg/kg per day). This dose was chosen because it will achieve a parasite reduction of at least 95% in mice [32]. The other group was given only the vehicle (2% Cremaphor EL) as a control. Two weeks after treatment, mice from both groups were euthanized and perfused with RPMI medium [33]. The body cavity, liver, and mesenteric veins were examined for worms after perfusion to ensure all worms were found. This assay was repeated two more times with subsequent generations of the KCW isolate, three more times for the PR1 isolate, and once more for both the NMRI and KNY isolates. We were unable to perpetuate the KAS isolate due to difficulties in hatching the eggs obtained from infected mice and to establish snail infections, making further experiments with it impossible. To eliminate the possibility that low PZQ susceptibility was due to a longer development time of the KCW worms, an additional in vivo assay was performed in which treatment was given at two time points. Mice were exposed to S. mansoni, randomly divided into 3 groups, and treatment was administered at either 7. 5 or 10. 5 weeks after exposure. The control group received the sham treatment at 7. 5 weeks post exposure. Immature worms are not susceptible to PZQ, and studies of the NMRI isolate indicated that susceptibility corresponds to the onset of reproductive development at 6 to 7 weeks post infection [34]. The sub-isolate of KCW that was kept at the Kenya Medical Research Institute (KEMRI) was tested for PZQ sensitivity in vivo after 3 years of passages in the absence of drug pressure or testing. Mice infected with this subset of KCW were treated at 10 weeks post-exposure, following the above protocol. The efficacy of treatment (ET) with PZQ as applied to the different isolates was measured as the percent of reduction of worm burdens based on the numbers recovered using the formula [35]: The number of worms recovered from the treated and control groups were compared across S. mansoni isolates with a randomized Generalized Linear Model 2-way ANOVA after log transformation of worm counts with SAS 9. 1. The main factors considered were S. mansoni isolate, and treatment; 2-way interactions among the main factors were analyzed. To complement the in vivo mouse treatment results an in vitro assay was devised to assess susceptibility of adult worms of S. mansoni to PZQ. Adult worms exposed to PZQ immediately contract. Although the mechanism by which PZQ causes this contraction is not fully understood, it is accompanied by a rapid influx of calcium ions, a slower influx of sodium ions and a decreased influx of potassium ions [36]. The rationale of our assay was to use the degree of contraction and subsequent recovery from contraction as a measure of PZQ susceptibility. We compared the length of the worms before exposure, during exposure (when the worms are contracted), and after a 24 hour recovery period to allow surviving worms to regain their normal lengths. Survival rates were also determined. Adult male worms (KCW, KNY and PR1, and the sub-isolate of KCW maintained in Kenya) were recovered by perfusion of mice and incubated overnight, at 37°C in an atmosphere of 5% CO2, in RPMI medium supplemented with 20% Fetal Bovine Serum, 100–500 IU Penicillin, and 100 µg/ml Streptomycin. Overnight incubation allowed for recovery from the stress caused by the perfusion. A single worm was placed in each well of a 96-well plate in 180 µl of supplemented RPMI medium. Before exposure to PZQ, each worm was photographed with a Nikon Coolpix 4500 camera mounted on a dissecting microscope phototube with a Thales Optem digital camera adapter. Then, 20 µl of either the control medium or a PZQ solution, of five different concentrations, was added to each well, resulting in final concentrations of 0,0. 8,8. 0,80. 0,400. 0, and 800. 0 µg/ml. The control solution contained 0. 1% DMSO, the same concentration as that in the experimental wells. The worms were incubated in the PZQ solution at 37°C for 3 hours and then photographed. The medium in each well was removed; the worms were washed three times, and were then transferred to a new 96-well plate in fresh RPMI medium, and held at 37°C. Each worm was re-photographed at the end of the 24 hour recovery period. The images of the worms before treatment and after recovery were processed with Metamorph v. 4. 65 software using the “fiber length” option which measures the object' s length. The length of the worms was compared across S. mansoni isolates and PZQ concentrations and analyzed statistically with SAS 9. 1 with a randomized ANOVA design; this was followed by Tukey' s multiple comparisons to find significant pair-wise differences. This research project has been reviewed and approved by the University of New Mexico' s Institutional Animal Care and Use Committee (IACUC), and the institutional review boards of the University of Georgia and the Centers for Disease Control and Prevention, the Scientific Steering Committee of the Kenya Medical Research Institute, and the KEMRI/National Ethics Review Board of Kenya. All investigators/assistants in this study have attained animal use certification regarding the ethical treatment of animals.
As expected, the in vitro effect of PZQ on the survival odds of miracidia was significantly influenced by drug concentration and time of exposure. To simplify data presentation, we show the results of exposure of miracidia derived from 23 different patients to 10−5 M PZQ after 20 minutes (Figure 1). For all time points examined, all miracidia held in water without PZQ continued to swim normally. There was considerable variation among the miracidia derived from different patients with respect to susceptibility to PZQ. The logistic regression model indicated that the odds of survival of miracidia in the in vitro assay upon exposure to 10−5 M PZQ are increased among those from patients who had received previous treatment compared to those from patients who had never been treated. Specifically, we obtained an estimated odds ratio of 2. 42 [with a corresponding 95% confidence interval of (1. 94,3. 01) ] for comparing survival of miracidia in a previously-treated versus untreated group. In other words, the odds of surviving PZQ exposure were 2. 42 times higher for miracidia collected from individuals who were previously treated than for those collected from patients who had never before been treated. Miracidia obtained from the car washer who had never fully cured following treatment and from whom the KCW isolate was derived were among the least susceptible to PZQ (# 22, Figure 1). We then passaged an isolate from a patient who had never been treated (KSH) and the KCW isolate through snails and mice in the lab and tested miracidia from the mice to determine their PZQ susceptibility. KCW miracidia were significantly less susceptible (p<0. 05) to PZQ (11% miracidia killed, from 6 trials) than KSH miracidia (75% miracidia killed, from 2 trials), suggesting that the reduced PZQ susceptibility of KCW was a heritable trait. With subsequent mouse infections we compared the PZQ susceptibility of adults of the KCW isolate with adults from other Kenyan isolates (KAS and KNY) or from long-maintained laboratory stocks (PR1 and NMRI). The results (Figure 2) are summarized based on multiple independent trials performed on successive generations of the worms, except for KAS, for which we had data from only one trial. The worm yields obtained from mice varied among isolates, which is not surprising given that the Kenyan isolates were recently derived from humans whereas the PR1 and NMRI isolates have been maintained in mice for decades. However, by reference to worm recoveries from corresponding untreated controls, we concluded that the efficacy of treatment was significantly lower in mice infected with KCW (31. 3%) than in mice infected with S. mansoni from the other 4 sources (over 98% for PR1, NMRI, and KNY and 67. 8% for KAS; p<0. 05), using a randomized ANOVA design. It should be noted that while the sex ratio for the worms recovered from mice infected with the KCW strain was male biased, the infections were not predominantly single sexed. Due to low infectivity for snails and mice and the inevitable attendant loss of genetic diversity, KAS worms recovered from mice were mostly or only males. The relatively low efficacy of treatment in mice infected with the KAS isolate (67. 8%) might be explained by this phenomenon, since worms from single-sex infections are less susceptible to PZQ than those from mixed-sex infections [24]. The results of the in vivo trial to examine the effects of delayed maturation showed that the efficacy of treatment was 28. 1% at 7. 5 weeks post-exposure and 26. 3% at 10. 5 weeks post-exposure. The mean number of worms recovered from control and treated mice did not differ significantly at either 7. 5 weeks post-exposure (control 32. 0±7. 9, treated 23. 0±6. 0, p = 0. 39) or 10. 5 weeks post-exposure (control 38. 7±4. 3, treated 28. 2±8. 1, p = 0. 29). Worms from untreated control mice at both times were sexually mature, and the livers from the mice harboring these infections contained many granulomas. These results indicate that the reduced susceptibility to PZQ observed in vivo with the KCW isolate was not a result of delayed maturation rate of KCW worms. In the in vitro assay to test adult worm PZQ susceptibility (Figure 3) we found that for all isolates, control worms not exposed to PZQ remained elongated at all observation times and there were no significant differences in worm length found among the three isolates (p = 0. 678). Following exposure to PZQ, regardless of the dose administered, worms from all three isolates contracted. However, for worms that were treated with 80 µg/ml of PZQ, there were differences among isolates in the extent that they remained contracted after the 24 hour recovery period. KNY and PR1 remained fully contracted and did not move. These worms were considered dead as per the definition of Pica-Mattoccia and Cioli [24]. In contrast, KCW worms partially elongated and were moving and active, and had increased their body lengths significantly more than KNY or PR1 worms (p = 0. 038). Worms of all three isolates recovered their normal length if they had been exposed to concentrations of PZQ lower than 80 µg/ml. Conversely, doses higher than 80 µg/ml of PZQ were lethal for all isolates. After the initial in vitro assay of miracidia that indicated reduced susceptibility of the KCW isolate, additional miracidia from this patient were divided into two sub-isolates in 2005, one was maintained in New Mexico and one in Kenya. After 8 generations of passaging the Kenyan sub-isolate through mice and snails, we performed the three susceptibility assays and found that the trait for reduced susceptibility had been lost. Before splitting the KCW isolate, in the in vitro assay of miracidia the mean percentage of dead miracidia recorded was 11%, which was significantly lower than after 8 generations (98%) and for all other isolates tested: 87%, 79%, 100%, and 78% for the PR1, NMRI, KAS and KSH isolates, respectively (p<0. 05). As shown in Figure 4, the subset of KCW maintained for eight passages over three years in Kenya (labeled KCW (2008) ) showed no diminished susceptibility to PZQ in the in vivo trials as compared to PR1 or KSH, and was significantly more susceptible to PZQ than the New Mexico subset of KCW (labeled KCW (2006) to indicate the year it was tested). The in vitro test of adult worms showed similar results. The difference in mean body length after the 24 hour recovery period between the New Mexico KCW sub-isolate and PR1, KNY and Kenyan KCW (2008) was significant (p = 0. 038). Kenyan KCW (2008), PR1 and KNY did not differ in this regard (p = 0. 211). In combination, these results indicate that the KCW sub-isolate maintained in Kenya lost the trait of diminished sensitivity to PZQ that we saw in miracidia when KCW was first isolated, and in comparison to the New Mexico sub-isolate.
In this study, we investigated PZQ susceptibility of S. mansoni from a high-transmission endemic area [28], [31] using in vitro assays involving both miracidia and adult worms, and an in vivo assay measuring drug susceptibility of adult schistosomes. In this population, we found worms with reduced susceptibility to PZQ. Miracidia collected from multiple patients varied significantly in their susceptibility to PZQ, and this susceptibility was associated with whether the patient had a history of exposure to the drug. Miracidia from car washers were more tolerant to PZQ than those from sand harvesters, which is not surprising because the car washer focus had been studied since 1995, while the sand harvesters were just being recruited in the study and had no prior history of exposure to PZQ. These results may suggest population level differences; however, we note that the sites are relatively close (within 5 km of each other) and our ongoing population genetics studies suggest no population subdivision between these groups or any other schistosomes collected throughout the Kenyan portion of Lake Victoria [22]. We hypothesize that treatment of individual patients with PZQ led to the accumulation of resistant worms within a single host rather than PZQ exposure inducing resistance within a patient, although induction of hycanthone resistance in schistosomes has been reported previously [37]. More in-depth studies were undertaken on an isolate (KCW) established from a patient who yielded miracidia with relatively low susceptibility to PZQ. Both in vivo and in vitro assays of adult worms confirmed this observation and showed that they had significantly lower susceptibility to PZQ than adults from other recent Kenyan isolates or from lab stocks. Drug resistance has evolved in many different microbial pathogens and parasites [38], [39], and is a looming concern for control of any pathogen, including schistosomes. Vigilant monitoring efforts are critical for preventing the spread and controlling the problem of drug resistance. The in vitro assays applied here were developed to monitor PZQ susceptibility of schistosomes, and have been validated by demonstrating for a variety of isolates that in vitro susceptibility of eggs, miracidia and cercariae to PZQ correlate with in vivo susceptibility of adult worms [26]. Such an in vitro assay is helpful given the inherent problems with measuring PZQ efficacy in humans, which is determined by measuring egg counts in fecal samples before and after treatment. Such an indirect measurement of efficacy can lead to erroneous conclusions about susceptibility of the schistosomes harbored by the treated person: they may harbor immature worms not susceptible to PZQ [16], [34], re-infections are likely in endemic areas and may confound interpretation of cure rates, and even though egg counts have diminished they may still harbor mature adult worms with temporarily inhibited egg production [40]. For such reasons, in vitro assays to determine drug susceptibility can be very useful, but they pose difficulties of their own. Most notably in our study was the difficulty of hatching eggs from field-collected fecal samples which made it hard to perform in vitro assays on populations of miracidia from some patients. Persistent difficulty in getting eggs to hatch ultimately led to the loss of the KCW sub-isolate maintained in New Mexico. Egg hatchability is an understudied phenomenon that could play an important role in the epidemiology of S. mansoni particularly if low hatching rates are an associated cost of drug resistance. Inclusion in our study of other Kenyan isolates that proved to be susceptible to PZQ provides evidence of variable susceptibility in natural populations and strengthens our findings of reduced susceptibility because the laboratory stocks potentially could have abnormal responses to praziquantel. Interestingly, the intensity of infection achieved with the Kenyan isolates was lower than that resulting from laboratory stocks. This difference likely is due to host adaptation as the laboratory stocks have been maintained in mice for several generations. It is unlikely that intensity influenced the outcome of the in vivo trials as efficacy of treatment was comparable across replicates regardless of intensity. In clinical reports, high intensity sometimes correlates with reduced efficacy of PZQ, but this could be due to high recruitment rates and the presence of immature worms within the host, or merely be a consequence of the fact that complete cures, as measured by cessation of egg production, are harder to obtain in people with high worm burdens even though the efficacy of treatment is the same. Even if reduced efficacy with high intensity infections was a phenomenon relevant to our in vivo trials, it would strengthen our conclusions because our more susceptible worms were present in higher intensities. Worm maturation is a potential alternative explanation for our results if the KCW worms develop more slowly in mice such that they were still immature and thus less susceptible to PZQ at standard treatment times. Although we saw no consistent evidence of a slower development rate, and KCW worms recovered upon perfusion at 7. 5 weeks post-infection were mature, we nonetheless addressed this possibility in an experiment that allowed three weeks of additional development before treatment was administered. The low PZQ susceptibility of these older, fully mature worms renders any argument invoking slower development times an unlikely explanation for the low PZQ susceptibility of KCW worms. In our study, diminished susceptibility to PZQ was a heritable trait that persisted across multiple generations: the KCW sub-isolate maintained in New Mexico retained reduced PZQ susceptibility over 6 generations, even in the absence of drug pressure. However, for the KCW sub-isolate maintained in Kenya, this trait was lost sometime during 8 generations of laboratory rearing. Loss of diminished susceptibility traits has also been reported for Egyptian field isolates brought into the laboratory [41]. The relative ease with which the trait of reduced PZQ susceptibility was lost may be associated with a fitness cost such as diminished asexual reproductive capacity in snails [41]. Fitness costs have been noted in other drug resistant organisms and play a large role in their evolution [42]. It is also interesting that with our divided isolates of KCW, the drug tolerant sub-isolate perished, but the susceptible sub-isolate persisted. The data from this study suggest that at least some members of the S. mansoni population in the Lake Victoria region have lower susceptibility to PZQ. As with most anthelmithic resistance, it is assumed that this trait has occurred naturally within the population even prior to drug treatment in the region. In fact, the patient from whom the KCW isolate was derived was never fully cured, even after the very first treatment. Furthermore, a longitudinal epidemiological study of these patients over the course of 12 years reported an average cure rate of 66%, but no evidence of increased treatment failure or clinically relevant level of resistance over the course of the study [28]. These findings suggest that worms with reduced susceptibility to praziquantel occur in the population, and are more common in patients that have been repeatedly treated, but this trait remains in relatively low frequency throughout the entire worm population. We suggest two hypotheses for this observation. First, although drug pressure is strong on the select target group of patients in Kisumu, the overall population of S. mansoni in snails, humans and other mammals in the Lake Victoria basin is vast [31], leaving a large refugium of untreated worms that are not subject to selection. Therefore, after treatment, the frequency of resistance alleles would remain low in the remaining worm population and likely in subsequent generations. The size of the untreated refugium has been shown to play an important role controlling the spread of resistance to other anthelminthics [43]. A second factor is that there may be a fitness cost associated with resistance, which would also work to prevent accumulation of the trait in the population unless the benefits of resistance outweighed the costs [41]. This reality, one that likely applies in many schistosomiasis endemic areas in Africa where treatment programs are underway, lessens current concerns about the spread of PZQ-resistance. However, as treatment expands to an ever larger proportion of infected individuals, then the refugia for unexposed parasites will diminish in size, increasing the chances for resistance to emerge. One of the most concerted programs of PZQ treatment, directed against S. haematobium in coastal Kenya, found variability in responsiveness to the drug, but no evidence of progressive emergence of PZQ resistance over an 8 year interval [19]. However, this program also involved a large refugium of untreated schistosomes because treatment was targeted to school children, and as the authors note, the limited coverage of the program may have reduced the tendency for resistance to spread. Based on computer modeling, it was predicted that emergence of PZQ resistance in this region should be anticipated within 10 to 20 years of its continued massive use [19]. Three aspects pertaining to the potential emergence of PZQ resistance are particularly deserving of additional study. The first is to determine if repeated cycles of infection and treatment of a single patient can favor the accumulation of a subset of S. mansoni adults more tolerant to PZQ and through recombination of alleles due to sexual reproduction, can lead to enhanced resistance in offspring. The pattern of decreased responsiveness to PZQ of miracidia derived from patients that had received multiple treatments, including KCW, is compatible with such a possibility. Trials to assess the safety and efficacy of higher doses for patients that fail to cure following exposure to standard PZQ dosages, or approval of more widespread availability of alternative treatments, such as oxamniquine, may both be prudent considerations when treatment failures occur. Second, because we lack a fundamental understanding of PZQ' s mode of action, we are also ignorant of the natural variability in PZQ' s targets in schistosome populations in endemic areas. It will be important to determine if variants inherently less susceptible to PZQ' s effects are particularly likely to occur in the large, genetically diverse populations of species like S. mansoni [23] that still thrive throughout much of sub-Saharan Africa. More explicit knowledge of PZQ' s targets will help us to devise much-needed improved assays for monitoring the emergence of resistance. Third, methods for genotyping individual eggs or miracidia need to be used to determine if the worm populations harbored by people before and after treatment show evidence of strong similarity, suggestive of the retention of fully adult worms that are not killed but merely temporarily silenced by PZQ treatment. Some Kisumu patients that have been treated remain negative for egg passage for long periods indicating they have been successfully cured and may even have acquired effective resistance to reinfection [30], but others resume egg production after treatment, indicative of acquisition of new worms, or possibly reactivation of existing worms. We conclude by noting that, ironically, the maintenance of PZQ susceptibility may come at the expense of poor coverage and continued high transmission. Continued monitoring of PZQ susceptibility using assays such as the ones employed here, and new and improved assays, is warranted as increased use of PZQ in control programs becomes a reality. | The emergence of drug resistant pathogens is a great challenge to the control of infectious diseases. Schistosomiasis is one of the world' s greatest neglected tropical diseases, and it is primarily controlled with the drug praziquantel. This drug is often used by repeatedly treating patients to maintain reduced worm burdens, an ideal situation to encourage the evolution of resistant worms. Although drug based control programs are increasing, monitoring efforts for drug resistance remain rare. We measured drug susceptibility of schistosomes from a cohort of patients in Kenya who are enrolled in a longitudinal study in which they are repeatedly treated with praziquantel. We found that schistosomes from previously treated patients were significantly less susceptible than those that were not. Also, schistosomes derived from a single patient who had been treated with praziquantel 18 times showed marked resistance. Although the findings of this study indicated that reduced drug susceptibility occurs in this population of schistosomes, this trait does not seem to be spreading widely or creating clinical levels of resistance. We hypothesize that the trait remains at low frequency because of the large population of schistosomes that are not exposed to the drug and/or potential fitness costs associated with reduced susceptibility. | Abstract
Introduction
Materials and Methods
Results
Discussion | evolutionary biology
infectious diseases/helminth infections
infectious diseases/antimicrobials and drug resistance | 2009 | Reduced Susceptibility to Praziquantel among Naturally Occurring Kenyan Isolates of Schistosoma mansoni | 8,704 | 288 |
Break-induced replication (BIR) has been implicated in restoring eroded telomeres and collapsed replication forks via single-ended invasion and extensive DNA synthesis on the recipient chromosome. Unlike other recombination subtypes, DNA synthesis in BIR likely relies heavily on mechanisms enabling efficient fork progression such as chromatin modification. Herein we report that deletion of HST3 and HST4, two redundant de-acetylases of histone H3 Lysine 56 (H3K56), inhibits BIR, sensitizes checkpoint deficient cells to deoxyribonucleotide triphosphate pool depletion, and elevates translocation-type gross chromosomal rearrangements (GCR). The basis for deficiency in BIR and gene conversion with long gap synthesis in hst3Δ hst4Δ cells can be traced to a defect in extensive DNA synthesis. Distinct from other cellular defects associated with deletion of HST3 and HST4 including thermo-sensitivity and elevated spontaneous mutagenesis, the BIR defect in hst3Δ hst4Δ cannot be offset by the deletion of RAD17 or MMS22, but rather by the loss of RTT109 or ASF1, or in combination with the H3K56R mutation, which also restores tolerance to replication stress in mrc1 mutants. Our studies suggest that acetylation of H3K56 limits extensive repair synthesis and interferes with efficient fork progression in BIR.
DNA damage drives mutagenesis and chromosomal rearrangements. Homologous recombination (HR) removes DNA lesions primarily at S/G2 phase of the cell cycle by pairing broken DNA ends with intact homologous template and copying across DNA breaks [1]. Break-induced replication (BIR) is the subtype of homologous recombination (HR) that eliminates one-ended DNA breaks or two-ended double strand breaks (DSBs) in the event when only one end of the DNA break is homologous to a template, such as a collapsed replication fork or eroded telomere. By searching for and copying from homologous sequences, often synthesizing hundreds of kilobases of DNA in the process, BIR has been implicated in catalyzing alternative lengthening of telomeres (ALT) and restoring replication forks [2,3]. Extensive DNA synthesis is unique to BIR or to only a subset of gene conversion events, both of which depend on Pol32, a non-essential subunit of DNA polymerase delta [4,5]. The large-scale DNA synthesis during BIR is also highly mutagenic in nature [6]. The initiation step of BIR is frequently associated with template switching, suggesting that the replication fork remains unstable and is prone to dissociation [7]. The precise mechanism behind the high propensity for mutations in BIR is not yet clear but might stem from its conservative mode of DNA synthesis and unique reliance on the helicase Pif1 for bubble migration [8,9, 10]. Every DNA transaction including DNA replication and repair occurs within chromatin. Therefore, histone modification and the reconfiguring of chromatin structure are not only a prerequisite, but also dictate both the efficiency and outcome of these events. Temporally disrupted chromatin in DNA replication and repair is then restored by the deposition of new nucleosomes and the re-establishment of unperturbed chromatin architecture globally and locally [11]. Histone H3K56 is an evolutionarily conserved residue that is subjected to reversible acetylation [12,13,14,15,16,17]. Unlike many other modifications at histone tails, H3K56 is located within the globular core domain near the entry and exit sites of nucleosomes, and does not appear to affect DNA-histone interactions as well as chromatin configuration [12,13,18,19]. In Saccharomyces cerevisiae, H3K56 is acetylated by the combined action of the acetyltransferase Rtt109/Vps75 and the histone chaperone Asf1, and deacetylated by two redundant class III histone deacetylases (HDACs) Hst3 and Hst4 [16,17,20,21,22]. Functionally, acetylated H3K56 facilitates replication-coupled nucleosome assembly and replication-independent histone exchange [23,24]. H3K56 acetylation is also required for histone eviction and transcription activation at certain promoters [25]. In mammals wherein acetylation of H3K56 is less abundant than in yeast, H3K56 is also methylated by the histone lysine methyltransferase G9a and promotes PCNA docking on chromatin in G1 [26]. In addition to its role in nucleosome assembly during replication and transcription, H3K56 also contributes to DNA repair and signaling [12,27]. Dys-regulation of histone H3K56 acetylation leads to a severe sensitivity to DNA damaging agents and elevated genome instability. Both excess and reduced levels of H3K56 acetylation disrupt cell cycle progression and induce spontaneous DNA damage and hypersensitivity to DNA damaging agent treatment [16,17,20,27,28]. Both hyper- and hypo-acetylation of H3K56 result in defects in sister chromatid cohesion and recombination [29,30] and ribosomal DNA (rDNA) amplification [31], as well as increased mutation frequency and GCR rate [32]. Furthermore, H3K56 is tightly regulated by the cell cycle and DNA damage-induced checkpoint, as the amount of Hst3 fluctuates at both the transcriptional and posttranscriptional level [16,29]. Deletion of HST3 and HST4 also leads to thermo-sensitivity, defective telomere silencing and increased chromosome loss [33,34]. Considering the role of H3K56 acetylation in nucleosome assembly, some of these repair defects could be attributed to low nucleosome density and the inability to recover from cell cycle arrest upon DNA damage, particularly in the absence of other histone chaperones [35,36,37]. However, this alone cannot explain why excess acetylation of H3K56 also leads to DNA damage sensitivity and genome instability. It is likely that hyper-acetylation of H3K56 interferes with some other aspect of DNA repair and signaling beyond nucleosome assembly. In this study, we investigated how chromatin modifiers and remodelers control BIR-mediated repair of an HO endonuclease-induced break on a disomic chromosome. Given the functional overlap of players between BIR and general DNA synthesis where chromatin modifications are tightly coupled to ongoing replication fork progression, we surmise that BIR should also depend on chromatin modification for efficient fork progression. We found that hyper-acetylation of H3K56 specifically inhibits efficient repair synthesis, therefore leading to BIR and gene conversion defects associated with long gap synthesis, but not other HR pathways. Deletion of HST3 and HST4 also impairs recovery from a collapsed replication fork in S-phase in DNA damage checkpoint deficient mrc1Δ cells and elevates translocation-type GCRs. We propose that acetylation of histone H3 impinges on the post-invasion step in recombination and thereby chromosomal integrity after DNA damage.
Extensive DNA synthesis in BIR likely relies on chromatin modification to achieve efficient fork progression. To address this question, we examined the effect of chromatin modifier gene deletions on BIR using an assay measuring BIR between disomic chromosomes [38]. Initial analysis was focused on chromatin remodeling complex factors (Swr1, INO80, RSC) and histone modifiers (Gcn5, Rtt109, Hst3, Hst4, and Dot1) implicated in DNA DSB repair. We induced a DSB at the MATa locus on a truncated copy of chromosome III, thereby triggering centromere proximal DNA ends to invade and copy from the intact MATα-inc allele on a full-length chromosome III as a DSB repair template (Fig. 1A). Since the truncated chromosome III possesses limited homology (46 bp) to the MATα-inc allele on the telomere proximal side of the DSB, BIR is the primary repair option, although a lower frequency of other repair events such as gene conversion (GC), chromosome loss or half crossover still occurs. The small percentage (>10%) of sectored Ade+/− colony formation likely represents the cases where only one of the two sister chromatids completed repair. Different types of repair events were deduced by the presence of marker genes located on the left arm of chromosome III (ADE1 and ADE3) and the LEU2 gene of the truncated chromosome III in surviving colonies. The survival frequency is nearly 100% because cells still retain a functional chromosome III copy even if repair is aborted and the truncated chromosome is lost. Inactivation of most chromatin modification factors altered the repair profile and BIR frequency but to a moderate extent (Fig. 1B). Increased BIR frequency in several of these mutants (htz1, arp8, and swr1) could be explained by their reduced resection function because extensive resection is known to suppress BIR [39,40]. However, the effect of SWR1 or ARP8 deletion on resection is variable among studies and modest at best [41,42,43]. We also found that deletion of DOT1 led to a dramatic decrease in BIR but increase in BIR/chromosome loss events. Since Dot1 recruits the checkpoint adaptor protein Rad9 at DNA break sites and plays an important role in DNA damage checkpoint activation [44], BIR deficiency in dot1Δ cells can be attributed to the lack of a functional cell cycle checkpoint. Indeed, dot1Δ and rad9Δ mutants showed very similar repair profiles with elevated BIR/chromosome loss (sectoring) events. Alternatively, BIR defects in dot1Δ or rad9Δ might be due to the increased resection in these mutants reported earlier [45]. Interestingly, deletion of two class III HDACs, HST3 and HST4 reduced BIR frequency but increased chromosome loss and half crossover events. The result somewhat mirrored that of pol32Δ mutants, albeit with less severity. Deletion of HST3 or HST4 alone or other sirtuin genes including SIR2 did not reduce BIR frequency. We also monitored BIR product formation by pulsed-field gel electrophoresis (PFGE) of genomic DNA isolated at various time points post-HO expression and Southern blot hybridization with a probe that anneals to the ADE1 gene on chromosome III. We found that deletion of HST3 and HST4 led to a 2-fold reduction in BIR product formation by Southern blot assay (S1 Fig.). To further confirm whether hst3Δ hst4Δ mutant cells are defective in BIR, we performed another ectopic BIR assay, wherein a galactose-inducible HO endonuclease creates a DSB on the left end of chromosome V and BIR restores an intact CAN1 gene that renders the cells canavanine sensitive, but leads to the loss of the HPH gene distal to the DSB (Fig. 1C). Using this ectopic assay we found that deletion of HST3 and HST4 led to a 5-fold reduction in BIR (Fig. 1D). These results indicate that hst3Δ hst4Δ mutants are significantly deficient in BIR. The strong resemblance in repair profiles between hst3Δ hst4Δ and pol32Δ mutants prompted us to examine if hst3Δ hst4Δ cells exhibit reduced expression of Pol32. We assessed the amount of triple HA-tagged Pol32 in hst3Δ hst4Δ cells via immunoblot assay with anti-HA antibody. We found that deletion of HST3 and HST4 did not reduce the expression level of Pol32 (S2 Fig.). Histone H3K56 is the best-known target of Hst3/Hst4 and represents the marker of newly deposited nucleosomes during replication or repair [12,17]. We therefore tested whether H3K56R or H3K56Q mutations that either block or partially mimic H3K56 acetylation will alter BIR frequency. We found that the H3K56R mutation alone did not exert any effect on BIR efficiency, whereas H3K56Q led to a slight but reproducible BIR defect (Fig. 2A). The modest BIR deficiency in H3K56Q mutants compared to hst3Δ hst4Δ cells supports the notion that H3K56Q does not fully mimic acetylated H3K56 [12]. We then analyzed the effect of the H3K56R mutation on BIR frequency in hst3Δ hst4Δ. If the observed BIR deficiency in hst3Δ hst4Δ is due to hyper-acetylation of H3K56, then the K56R substitution should offset this BIR defect. Indeed, H3K56R almost fully rescued the BIR defect in hst3Δ hst4Δ cells (Fig. 2A). These results suggest that hyper-acetylated H3K56 inhibits BIR. Rtt109 is an acetyltransferase responsible for acetylation of H3K56 [20,21]. Asf1 is a histone chaperone specific for H3K56 acetylation [22]. We thus examined the effect of RTT109 or ASF1 deletion on BIR frequency in wild-type and hst3Δ hst4Δ cells using the disomic BIR assay. As predicted, rtt109Δ or asf1Δ alone did not influence BIR frequency (Fig. 2A). However, deletion of ASF1 or RTT109 in hst3Δ hst4Δ mutants improved BIR frequency to a level almost identical to that in wild-type cells. The level of BIR product formation observed in this genetic assay was also confirmed by Southern blot analysis (S1 Fig.). These results further demonstrate that hyper-acetylation of H3K56 inhibits BIR. BIR deficiency in hst3Δ hst4Δ cells prompted us to test whether deletion of HST3 and HST4 generally affects other HR processes. We thus tested the effect of HST3 and HST4 deletion on mating-type switch recombination, ectopic gene conversion, and single strand annealing using genetic assays wherein the repair of an HO break occurs by different recombination subtypes (S3 Fig.). The recombination frequency was determined by survival frequency after HO expression and physical monitoring of the repair products by Southern blot-based assay. We found that none of the recombination subtypes tested were impaired in hst3Δ hst4Δ cells (S3 Fig.). We also determined the frequency of non-homologous end joining (NHEJ) in donorless hst3Δ hst4Δ cells by inducing an HO break at the MAT locus. Deletion of HST3 and HST4 did not lead to any changes in survival frequency following HO expression (S3 Fig.). This suggests that hst3Δ hst4Δ mutants are deficient in BIR but not other recombination or NHEJ processes. Since H3K56 acetylation marks newly assembled nucleosomes, half of the nucleosomes in each cell should bear this mark after DNA synthesis. Based on this premise, we set out to test if the physiologically relevant level of H3K56 acetylation is sufficient to inhibit BIR. To this end, we engineered a yeast strain expressing a degron-fused and the MYC epitope-tagged Hst3 as the sole H3K56 deacetylase, and then examined BIR efficiency in cells arrested at G2 after one round of DNA synthesis upon depletion of Hst3 using the ectopic BIR assay (Fig. 2B). Depletion of Hst3 at non-permissive temperature (37°C) was confirmed by thermosensitivity and immunoblot assay with anti-MYC antibody (S4A and S4C Fig.). Cell cycle progression was monitored by fluorescence activated cell sorting (FACS) after propidium iodide (PI) staining of cells (S4B Fig.). We found that depletion of Hst3 decreased BIR frequency 3-fold in cells arrested at G2 as compared to those expressing Hst3 (Fig. 2C, 2D). The relatively small reduction (3-fold) in BIR in Hst3-depleted cells compared to the five-fold reduction observed in hst3 hst4 deleted cells suggests that the level of BIR is proportional to the amount of intracellular H3K56 acetylation (S4D Fig.). Inactivation of Hst3 and Hst4 by the pan–sirtuin inhibitor nicotinamide (NAM) also reduced BIR frequency 2-fold (S5 Fig.). As shown previously, deletion of SIR2, HST1 and HST2 did not alter BIR frequency (Fig. 2A). The results suggest that physiological levels of H3K56 acetylation are sufficient to suppress BIR in cells. Replication forks stall when deoxyribonucleotide triphosphate (dNTP) pools are depleted and even collapse without a functional S-phase checkpoint [46]. Mrc1 is an S-phase checkpoint protein that stabilizes replication forks when hydroxyurea (HU) is used to induce dNTP pool depletion [47,48]. In the absence of Mrc1, stalled replication forks harbor extensive single strand DNA (ssDNA) and become collapsed [49]. Thus cell survival relies on pathways that repair collapsed replication forks [50]. We examined HU sensitivity of mrc1Δ cells deleted for HST3, HST4 or POL32 genes upon acute HU treatment. Unlike typical plate based assays where cells are exposed chronic DNA damage and dys-regulation of H3K56 acetylation leads to HU sensitivity [28], the acute HU exposure (1 to 4 h) employed here likely addresses the repair of collapsed fork [51]. We found that hst3Δ hst4Δ, pol32Δ or mrc1Δ cells are only mildly sensitive to HU treatment, but hst3Δ hst4Δ mrc1Δ or pol32Δ mrc1Δ cells showed far greater sensitivity than each single gene deletion mutant (Figs. 3A and S6A Fig.). Deletion of POL32 also sensitizes tof1- or csm3-deleted cells to HU treatment (S6B–S6C Fig.). Furthermore, the HU sensitivity of hst3Δ hst4Δ mrc1Δ mutants is fully rescued by an asf1Δ mutation (Fig. 3B). Deletion of ASF1 or RTT109 did not sensitize mrc1Δ to HU treatment (Figs. 3B and S6D). These results suggest that the observed HU-induced stall or collapse of replication forks in mrc1Δ cells relies upon Pol32- and H3K56 deacetylation dependent process for recovery. Rad53 is another S-phase checkpoint protein that sustains replication fork integrity in response to HU treatment [46]. Surprisingly, in contrast to mrc1Δ cells, deletion of HST3 and HST4 or POL32 did not further sensitize rad53Δ mutants to HU treatment Fig. 3D). These results suggest that not all collapsed forks depend on Hst3/Hst4 and Pol32-dependent process for repair. Mrc1 contributes to both damage-induced checkpoint induction and replication fork stabilization [47]. These two functions can be uncoupled by the mrc1AQ allele that abolishes Mrc1’s checkpoint function but retains its role in replication fork stabilization [47]. To determine which function of Mrc1 contributes to HU resistance in hst3Δ hst4Δ or pol32Δ mutants, we examined HU sensitivity in a strain bearing the mrc1AQ mutation and hst3 hst4 or pol32 gene deletions. We found that the mrc1AQ mutation did not confer synergistic HU sensitivity to hst3Δ hst4Δ or pol32Δ cells, suggesting that Mrc1’s checkpoint function is dispensable for HU sensitivity in hst3Δ hst4Δ or pol32Δ mutants (Figs. 3C and S6E). Additionally, deletion of RAD9, another DNA damage checkpoint adaptor but not a fork stabilizer, did not sensitize hst3Δ hst4Δ cells to HU treatment (S6F Fig.). The results suggest that stalled/collapsed replication forks but not checkpoint dysfunction lead to HU sensitivity in hst3Δ hst4Δ or pol32Δ mutants. The lethality upon acute HU treatment in hst3Δ hst4Δ mrc1∆ cells could be due to an inability to repair a collapsed fork or alternatively to a more extensive fork collapse. It has been shown that HU treatment induces Rad52-green fluorescent protein (GFP) focus formation in mrc1Δ but not in wild-type nuclei, likely marking the site of stressed and collapsed replication forks and their repair [51]. We thus monitored the kinetics of Rad52-GFP foci as a surrogate for the repair of collapsed replication forks in order to dissect HU-induced fork collapse in mrc1∆ cells. We found that Rad52-GFP foci appeared in mrc1Δ cells upon HU treatment, reached a maximum level after releasing cells into fresh YEPD medium, and gradually disappeared most likely due to repair (Fig. 4). Importantly, in hst3Δ hst4Δ mrc1Δ cells and pol32Δ mrc1Δ cells, the Rad52-GFP foci persisted for up to 8 h after HU treatment, suggesting that hst3Δ hst4Δ or pol32Δ cells are deficient in repairing collapsed replication forks (Fig. 4). To test if the repair deficiency in hst3Δ hst4Δ mrc1Δ mutants could be attributed to hyper-acetylation of H3K56, we deleted ASF1 and monitored Rad52-GFP kinetics following HU treatment. Deletion of ASF1 led to a high level of Rad52-GFP foci at 6–8 h after release from HU, consistent with the role of Asf1 in fork stability [52], yet significantly offset the level of Rad52-GFP in hst3Δ hst4Δ mrc1Δ cells (Fig. 4A and B). Deletion of HST3 and HST4 also led to an increase in spontaneous Rad52-GFP focus formation and sensitivity to acute HU treatment in the BY4741 strain, but not in another genetic background (JKM139, see Fig. 4A and C, and S6 Fig.). We surmise that Hst3 and Hst4 contribute to the integrity of replication forks in certain genetic backgrounds. The results support the role of Hst3, Hst4 and Pol32 in the repair of collapsed replication forks. Improper repair of replication stress is a likely source of gross chromosomal rearrangements (GCRs) [53]. Furthermore, deletion of ASF1 or RTT109 increases GCR rate [20,54]. We thus analyzed GCR frequency in hst3Δ hst4Δ or pol32Δ cells using the assay that measures the loss of GCR reporter genes on the left arm of chromosome V (Table 1). We found that deletion of HST3 and HST4 increased GCRs 457-fold (Table 1). Deletion of ASF1 partially reduced GCR rate to the level indistinguishable from that observed in asf1Δ cells, suggesting that hyper-acetylation of H3K56 is the likely cause of elevated GCRs in hst3Δ hst4Δ mutants. Interestingly, deletion of POL32 leads to only a mild increase in GCR. Analysis of the breakpoints of GCRs showed that most of the GCR events in hst3Δ hst4Δ cells are DNA translocations with 3–12 bp of microhomology at the junction, whereas all GCRs in pol32Δ cells are the result of de novo telomere formation (Tables 1 and S1). Cells bearing the hst3Δ hst4Δ mutations do not grow at 37°C but the deletion of genes involved in DNA damage checkpoint induction (RAD17) or sister chromatid establishment (CTF4) partially offset their thermosensitivity and DNA damage sensitivity phenotypes [28]. To test if the observed thermosensitivity and BIR defect share a common genetic cause, we deleted RAD17 or CTF4 and examined whether the gene deletions improve BIR in hst3Δ hst4Δ cells. We found that deletion of RAD17 or CTF4 partially offsets the growth deficiency at 37°C and resistance to MMS treatment as predicted, but does not affect defects in BIR in hst3Δ hst4Δ mutants (Fig. 5A and B). The results suggest that thermosensitivity, DNA damage sensitivity, and BIR deficiency are not genetically linked to each other in hst3Δ hst4Δ mutants and likely represent somewhat distinct molecular defects. Most recently, a ubiquitin ligase complex containing Rtt101, Mms1, and Mms22 has been implicated in H3K56 acetylation-dependent mutation avoidance and damage tolerance pathways [32]. Deletion of MMS22 also offsets the thermosensitivity of hst3Δ hst4Δ cells (Fig. 5A and [55]). We thus tested the effect of MMS22 gene deletion on BIR using the disomic BIR assay. We found that deletion of MMS22 did not alter the repair profile in wild-type or hst3Δ hst4Δ cells (Fig. 5B). The results further support the model that the BIR defect in hst3Δ hst4Δ mutants is genetically distinct from underlying molecular defects causing thermosensitivity, elevated mutagenesis and sensitivity to DNA damaging agents. Evidence suggests that acetylated H3K56 promotes SWR-C dependent exchange of H2AZ to H2A at promoter proximal regions and thereby regulates transcription [56]. If the observed BIR defect in hst3Δ hst4Δ mutants is due to excessive exchange of H2AZ to H2A, deletion of HTZ1 should lead to BIR defects as in hst3Δ hst4Δ cells. However, we found that deletion of HTZ1 or SWR1 does not result in BIR deficiency in the disomic BIR assay (Fig. 1B). Furthermore, deletion of HTZ1 or SWR1 did not rescue the BIR defect in hst3Δ hst4Δ cells (S7 Fig.). The results suggest that BIR deficiency in hst3Δ hst4Δ mutants is not due to excessive removal of H2AZ or dynamic histone exchange at promoters. We then asked how hyper-acetylation of H3K56 inhibits BIR, which involves a few common and unique recombination steps [2]. Following DSB induction, the broken DNA ends are processed by 5’ to 3’ resection to yield single stranded DNA (ssDNA) that serves as the binding site for Rad51 recombinase. The ssDNA-Rad51 complex (a. k. a. pre-synaptic filament) then invades into a homologous template and initiates strand pairing. The 3’ non-homologous tail should be removed from the annealed DNA so that repair synthesis will begin copying from the template until the end of the chromosome. We analyzed the formation of ssDNA, binding of Rad51 at donor and recipient DNA molecules, 3’ flap removal, and initiation of DNA synthesis in hst3Δ hst4Δ mutant cells using PCR-based assays and chromatin immunoprecipitation (ChIP) using anti-Rad51 antibody (S8, S9, S10 Figs.). We found that deletion of HST3 HST4 did not impair resection or the kinetics of Rad51 enrichment at either the break or homologous donor sites (see S8, S9 Figs. for experimental details). We discovered a slight reduction in flap removal efficiency in both hst3Δ hst4Δ and pol32Δ cells, but such a mild deficiency does not likely account for the BIR defects observed in these mutants (S10 Fig.). Furthermore, hst3Δ hst4Δ and pol32Δ mutants are not defective in single strand annealing wherein 3’ flap removal becomes the rate-limiting step, dictating repair frequency (S3 Fig.). We next examined the kinetics of initial repair synthesis by PCR amplification of repair products using a pair of primers in which one anneals to the donor site and the other anneals to the recipient DNA molecule. We discovered that both hst3Δ hst4Δ and pol32Δ mutants were severely defective in repair synthesis (Fig. 6). A single round of DNA replication after depletion of Hst3 in hst4 Δ cells also inhibits repair synthesis during BIR (S11 Fig.). In contrast, deletion of HST3 and HST4, or POL32 did not significantly impair or alter the kinetics of repair synthesis in mating type switching consistent with the finding that mating type switching is proficient in these mutants (S3 Fig.). The results suggest that the BIR defect in hst3Δ hst4Δ cells might be due to inefficient DNA synthesis. The results also reinforce that two-end break repair with no or limited gap is fundamentally different from two-end break repair with large gap or one-end break repair [57]. To further test if hst3Δ hst4Δ mutants are deficient in long repair DNA synthesis, we determined the frequency of gene conversion that requires 3. 4-kb gap synthesis in wild-type, hst3Δ hst4Δ, and pol32Δ cells (Fig. 7A). As evident in Fig. 7B, deletion of HST3 and HST4 significantly reduces gene conversion requiring large gap synthesis (tGI354-Gap), although pol32Δ cells exhibited a more severe defect (Fig. 7B). This is consistent with the relatively greater deficiency of pol32Δ mutants in BIR and repair synthesis (see Fig. 1B). Neither hst3Δ hst4Δ nor pol32Δ exhibited reduced gene conversion frequency without a gap (tGI354). Cumulatively, our data support the model that hyper-acetylation of H3K56 inhibits extensive repair synthesis.
BIR entails extensive DNA synthesis even to the end of the chromosome. Since chromatin may impede processive DNA synthesis [11], we surmise that BIR likely depends on chromatin modification, which then enables extensive DNA synthesis to the chromosome end. By surveying several candidate genes that have already been implicated in DNA synthesis and recombination, we discovered multiple chromatin modifiers that impinge on BIR events, supporting our original premise that chromatin modification represents an integral part of BIR. Interestingly, the effect of deleting several chromatin modifier genes on BIR can be explained by their role in resection, highlighting the impact of resection regulation on the frequency of BIR events [39,40,58]. The results also indicate that chromatin regulation comprises a major component in early repair steps, often contributing to the commitment stage of DNA repair. Consistent with this, chromatin configuration has been implicated in repair pathway channeling in several model systems [59,60]. Our results also led to the discovery that deletion of other histone modifiers, specifically HST3 and HST4, renders cells severely deficient in BIR using two different BIR assays, with a repair profile closely resembling that of pol32Δ mutants. Several observations indicated that hyper-acetylation of H3K56 causes BIR deficiency in hst3Δ hst4Δ cells. First, the H3K56Q mutation that mimics persistent acetylation reduced the frequency of BIR, albeit more mildly than hst3Δ hst4Δ. Second, deletion of ASF1 or RTT109, or replacement of lysine 56 with arginine in a histone H3 mutation that blocks H3K56 acetylation rescued the BIR defect in hst3Δ hst4Δ mutants. Third, hst3Δ hst4Δ cells are sensitive to HU in the presence of a collapsed replication fork, but become resistant to HU when ASF1 is deleted. We also showed that the intracellular level of H3K56 is proportional to BIR frequency and that a physiologically relevant level of H3K56 acetylation is sufficient to repress BIR. Lastly, deletion of HST3 HST4 elevated the frequency of translocation-type GCRs, but this increase was partially suppressed by the concomitant deletion of ASF1. Together these results suggest that H3K56 status, but no other HST3 and HST4 targets, is responsible for BIR deficiency and the associated chromosomal instability. To define how H3K56 acetylation inhibits BIR, we assessed the integrity of the four major recombination steps in BIR: resection, Rad51 binding at donor and recipient DNA molecules, 3’ flap removal and repair synthesis. Our findings suggest that deletion of HST3 and HST4 specifically impairs repair synthesis. This helps explain why only BIR and gene conversion via long gap repair synthesis, but no other recombination pathways, are deficient in hst3Δ hst4Δ cells. The results also support our premise that epigenetic regulation represents a key component of BIR. How does hyper-acetylation of H3K56 impede repair synthesis? It should be possible that acetylation of H3K56 might compromise the establishment of a stable replication fork at a BIR template. Interestingly, results from chromatin immunoprecipitation (ChIP) assays demonstrated that enrichment of Pol delta is reduced in hst3Δ hst4Δ cells, whereas Pol alpha and Pif1 still associate at the replication fork with almost equal efficiency (S12 Fig.). However, reduced association of Pol delta at a BIR template could simply be the effect but not the cause for BIR deficiency in hst3Δ hst4Δ mutants. Alternatively, acetylation might block replication fork movement by conferring nucleosomes inherently resistant to sliding or evicting from the template strand. Indeed, deletion of HST3 and HST4 in one genetic background (BY4741) led to an increase in spontaneous Rad52-GFP focus formation and elevated sensitivity to HU treatment. However, structural and biochemical analysis of histone octamers carrying H3K56Q did not reveal any major change in nucleosome architecture [18]. Furthermore, hyper-acetylation of H3K56 might impair access to or the activities of topoisomerase I or helicases, each of which relieves helical torsions or catalyzes strand unwinding ahead of a replication fork. In either case, we surmise that frequent fork stalling may ensue in hst3Δ hst4Δ mutants due to non-processive DNA synthesis, which may underlie the increased spontaneous DNA lesions and mutagenesis exhibited by these cells. The observations that many phenotypes of hst3Δ hst4Δ cells can be suppressed at least partially by improving replication processivity support this model [28]. Further studies are needed to determine precisely how H3K56 acetylation inhibits repair synthesis. In mitotic cells, limiting extensive repair synthesis in recombination should be beneficial as it could reduce mutagenesis inherent to repair synthesis [61]. The size of gene conversion track length also positively correlates with the frequency of crossovers and the rate of loss of heterozygosity [62]. Limiting the amount of repair synthesis should also helps avoid unequal sister chromatid exchange event and sustain repeat stability as seen in rDNA or triplet repeats. Interestingly, DNA damage induces proteolytic degradation of Hst3 [16,29], preserving intracellular level of H3K56 acetylation, which could then help limit excessive repair synthesis during recombinational repair of DNA damage. Since collapsed replication forks produce one-ended DSB, BIR has been implicated in the restart of replication upon fork collapse [2,3]. Nevertheless, multiple means are still available in cells to complete full genome duplication without re-establishing forks by BIR: a) use of a replication fork approaching from a neighboring origin; b) template switching or re-priming. Accordingly, the precise role of BIR in recovery from a collapsed fork has not yet been fully defined. Interestingly, we discovered that deletion of POL32 or HST3 and HST4 caused a severe sensitivity to HU when MRC1 is also deleted. Synergism between MRC1 and HST3/HST4 in resistance to HU treatment can be attributed to deficient maintenance of fork integrity in mrc1 mutants but not to an S-phase checkpoint defect, since the checkpoint-deficient mrc1AQ allele fully rescued HU sensitivity. Puzzlingly however, in rad53Δ, where a majority of replication forks collapse, deletion of POL32 or HST3 and HST4 did not sensitize cells to HU treatment. Evidence suggests that the structure of a collapsed fork in mrc1 mutants upon HU treatment associates with extensive ssDNA formation (2–3 kb long) due to uncoupling of the replicative helicase from stalled fork [49]. Extensive ssDNA at the fork then leads to DNA breakage and the repair requires extensive repair synthesis through newly deposited, H3K56 acetylated nucleosomes on the donor DNA molecule (S13 Fig.). The unique architecture associated with stalled forks in mrc1 cells might also be ill-suited for re-priming or bypass-mediated recovery, and thereby relies on the role of BIR like process as a bona fide replication restart mechanism. Alternatively, hyper-HU sensitivity in hst3Δ hst4Δ mrc1Δ mutant could be attributed to the excessive fork collapse due to a deficiency in stalled fork protection under HU treatment. However, we did not detect massive fork collapses in hst3Δ hst4Δ mrc1Δ or mrc1Δ cells as compared to rad53Δ using the assays that monitor the frequency of Rad52-GFP foci or DNA fragmentation by PFGE (Figs. 4 and S14). The typical collapsed replication forks in cells do not likely feature long ssDNA and such forks in principle contain acetylated H3K56 only at or behind the fork. This raises a possibility that the effect of persistent H3K56 acetylation shown in this study is not applicable to general collapsed fork recovery, but only to those with forks collapsed with long ssDNA as seen in mrc1- or tof1-deleted cells. Furthermore, H3K56 acetylation is increased in multiple types of cancers [14] and the results shown here could be more relevant to those pathological conditions or at regions of chromatin with more persistent H3K56 acetylation. Previous studies suggest that both hypo- and hyper-acetylation of histone H3K56 elevates GCR events [20,32,54]. Consistent with this, we also detected dramatically increased GCRs in hst3Δ hst4Δ mutants. However, unlike in an asf1 deletion [54], most GCR events in hst3Δ hst4Δ cells are chromosomal translocations. The results indicate that the molecular defects and pathways leading to GCR events in these strains are distinctly different. Accordingly, deletion of ASF1 partially decreased the GCR rate in hst3Δ hst4Δ cells to a level comparable to that in asf1Δ cells. We have evidence to suggest that it is unlikely that increased GCR observed in hst3Δ hst4Δ mutants is derived from a BIR defect. If BIR deficiency alone is sufficient for elevated GCR, deletion of POL32 that results in a more severe deficiency in BIR should increase GCR events at a greater extent than in hst3Δ hst4Δ cells. Instead, we found that deletion of POL32 only slightly increases GCRs. Analysis of the types of GCR and the breakpoint junctions in these two mutants further demonstrated that the BIR defect in hst3Δ hst4Δ cells cannot account for its elevated GCR because pol32Δ produces de novo telomere addition but not chromosomal translocations as seen in hst3Δ hst4Δ cells. Why then is GCR increased in hst3Δ hst4Δ mutants? One possibility is that DNA replication forks are unstable in hst3Δ hst4Δ cells and lead to frequent stall and collapse during S-phase. Spontaneous Rad52-GFP, DDC2-GFP, and Mre11-GFP focus formation, as well as Rad53 phosphorylation supports the increase in replication-associated DNA damage in hst3Δ hst4Δ cells, which then probably fuels the formation of GCRs. In fact, many cellular phenotypes of hst3Δ hst4Δ mutants can be explained by fork instability that might lead to such pleiotropic phenotypes. The idea is further supported by the observation that multiple phenotypes associated with hst3Δ hst4Δ mutants including thermosensitivity, drug sensitivity and spontaneous mutagenesis can be partially offset by RAD17, ELG1, or MMS22 deletion that catalyzes the S-phase checkpoint, PCNA unloading or replication fork stability, respectively. However, these mutations did not offset BIR deficiency, indicating that the basis of the observed BIR defect is distinctly different from that of other phenotypes. Analysis of GCR breakpoint junctions shows that most GCR events in hst3Δ hst4Δ cells are chromosomal translocations with microhomology (MH) at the breakpoint junctions. The presence of MH suggests that GCR events in hst3Δ hst4Δ mutants proceed by microhomology-mediated (MM) -BIR or microhomology-mediated end joining (MMEJ) mechanisms. Notably, deletion of POL32 not only impairs BIR but also the MMEJ pathway [5,63]. Therefore, lack of GCRs in pol32-deleted cells may be due to their inability to carry out MMEJ. Indeed, all the GCR events in pol32 are de novo telomere formation. We also found that hst3Δ hst4Δ mutants are not deficient in MMEJ. Alternatively, it is possible that the GCR events derived from residual BIR mechanisms in hst3Δ hst4Δ cells and the severe BIR deficiency in pol32Δ cells disable the formation of translocation-type GCRs. In all, our study has revealed the role of H3K56 acetylation in inhibiting extensive repair synthesis and restarting of stressed replication forks in checkpoint-deficient cells. It will be interesting to test if acetylation of H3K56 also inhibits collapsed fork recovery and BIR in human cells. It is noteworthy that HDAC or sirtuin inhibitors have been widely regarded as promising cancer therapeutic agents. Our study has implications for how these inhibitors target checkpoint-defective tumors cells and highlight their utility to treat cancer cells in combination with replication stress agents.
Disomic BIR strains are derivatives of AM1003 with genotype MATa-LEU2-tel/MATα-inc ade1 met13 ura3 leu2-3,112/leu2 thr4 lys5 hmlΔ: : ADE1/hmlΔ: : ADE3 hmrΔ: : HYG ade3: : Gal-HO FS2Δ: : NAT/FS2 [38]. Ectopic BIR strains are derivatives of JRL346 [40]. HU sensitivity was tested on derivatives of JKM139 with genotype hoΔ MATa hmlΔ: : ADE1 hmrΔ: : ADE1 ade1-100 leu2-3,112 lys5 trp1: : hisG’ ura3-52 ade3: : GAL: : HO [64]. JKM161, YMV80, tGI354 were used for mating-type switch, SSA, and ectopic recombination assays, respectively [65,66]. Gene conversion with a 3. 4-kb gap strain was constructed by insertion of a TRP1 containing DNA fragment originating from the pFA6a-TRP1 plasmid into the MATa-inc template of tGI354. The yeast strain expressing MYC epitope tags and degron fusion was constructed by PCR amplification of degron cassette as described in [17,67]. The strain list is shown in S2 Table. The assay was performed as previously described with minor changes in cell culture[38]. BIR derivative strains were cultured in SC-Ade medium overnight and then transferred to YEP-glycerol medium and cultured for more than 6 h before plating on YEP-galactose plates. Cells cultured in this way are selected against spontaneous chromosome loss. An aliquot of cells was also plated on YEPD plates. After colonies grew up on the plates for 3–5 days, they were replica-plated onto SC-Leu and SC-Ade plates to determine their genotypes. At least 500 individual colonies were scored and at least 2 different isolates with the same genotype were analyzed. Colonies replica-plated from YEPD plates were also examined to determine the background chromosome loss. BIR frequency of ectopic BIR assay was determined as described in [5]. To physically monitor BIR products, 3×107 cells were harvested to make DNA plugs for each time point. Pulsed-field gel electrophoresis (PFGE) was run in a 1. 2% agarose gel with switch time 10–35s, voltage 6v/cm, angle 120°, 14°C for 32–40 h (disomic BIR assay) or with switch time 50–70 s, voltage 6v/cm, angle 120°, 14°C for 21–24 h (ectopic BIR assay). PCR products from ADE1 gene or MCH2 fragment were radiolabeled and used to detect BIR bands by Southern blot assay. To test BIR in Hst3-depleted cells or those treated with NAM, HST4-deleted cells were arrested in G1 by treating them with 10 µg/ml alpha-factor, subjected to incubation at 37°C (non-permissive conditions) for an additional 1. 5 h, releasing cells into nocodazole containing (15 µg/ml) media for 2. 5 h and inducing HO endonuclease by addition of 2% (w/v) galactose. Aliquots of cells were harvested at each indicated time point and subjected to Western blot to investigate the level of Hst3, H3K56 acetylation using anti-MYC (Sigma) and H3K56 acetylation antibodies (Upstate) and FACS analysis for cell cycle progression. Cells were harvested to prepare DNA plugs for PFGE to analyze BIR products and PCR analysis of repair synthesis. PCR products from ADE1 gene or MCH2 fragment were radiolabeled and used to detect BIR bands or initial repair products by Southern blot. Logarithmically growing cells were alpha-factor arrested for 3 h and released into YEPD medium containing 150 mM HU. Cells were harvested at each time point, washed twice with water and plated onto YEPD plates. The survival rate was calculated by the number of colonies treated with HU divided by the number of colonies before HU treatment. For the mrc1AQ experiment, Uracil dropout medium instead of YEPD medium was used for cell culture. Images were taken with a DeltaVision microscopy system (Applied Precision). Each original image contained 12–15 Z-stacks. After deconvolution, images were subjected to quick-projection. The projected pictures were then counted for Rad52-GFP foci using SoftwoRx. 3’ flap cleavage efficiency during gene conversion was determined by measuring the retention frequency of yeast centromeric plasmids pFP120 and pFP122 after HO induced DSB as described in [68]. The percentage of plasmid retention is calculated as the fraction of colonies retaining the repaired plasmid on SC–Ura divided by the total number of colonies on YEPD. Chromatin immunoprecipitation (ChIP) assays were performed as described previously [69]. Briefly, JKM161 cell cultures grown to a density between 1 × 107 and 2 × 107 cells/ml in pre-induction medium (YEP-glycerol) were induced for HO endonuclease by adding 2% galactose. For immunoprecipitation, the sonicated extracts were incubated with Rad51 antibody (kindly provided by Dr. Patrick Sung P). Quantitative PCR was performed using MATa (break) specific primers and HML (donor) specific primers. JKM139 derivative cells grown to a density between 1 × 107 and 2 × 107 cells/ml in pre-induction medium (YEP-glycerol) were induced for HO endonuclease by adding 2% galactose. Five ml of cells at different time points were harvested and processed for genomic DNA isolation using Masterpure yeast genomic DNA purification kit (Epicentre). DNA equal to 0. 4 ml of cells was digested by 20 units of BsaJ1 enzyme or mock digested without BsaJ1 for 2 h in a 50 μl reaction volume. Samples were then diluted 10 times and subject to quantitative PCR to determine the percentage of resection. The principle of this assay is shown in S8 Fig. The rate of single stranded DNA formation (r) is calculated by r = (RSBsaJ1 digested/ RSMock digested) / (InputBsaJ1 digested/ InputMock digested), where RS indicates PCR value at the BsaJ1 restriction site flanking the HO break and Input indicates PCR value at a site from a different chromosome and does not contain the BsaJ1 restriction site. Since r = X/ (X+2 (1-X) ) where X is the rate of resection, the percentage of resection was calculated by: 100X% = 200r/ (r+1). Primer sequences for determining resection at 3. 3 kb, 12. 6 kb or input are 5’TCCGAACCAATGTCGTCATCCATAGTATC and 5’GGCGCCTTTAATTCATGGTCTTTACCTTT; 5’TTCTACGGCGACTGGTGGAATTGC and 5’ATGTAGCTTGGCTCTTGCTCAAATGC; 5’ACTTAGTTCTGGAGATTATCGCCTTATCG and 5’CTGTCTTTCGCTTAGTTCACCTCTACC, respectively. Semi-quantitative PCR was performed with Iproof polymerase (Biorad) with a 2 minute extension time and 20 cycles (allelic BIR assay) or 28 cycles (ectopic BIR assay) using Forward primer F (5’ TACTGGAGTTAGTTGAAGCATTAGG3’) upstream of the HO break inside the URA3 gene and Reverse primer R1 (5’ TGACTTCCAGACGCTATCCTGTGAA 3’) inside the MAT locus at an alpha-specific sequence or R2 (5’GATATTAAGTCCTCCGTCCAATCTG3’) downstream of MATalpha-inc, inside the TAF2 gene (allelic BIR assay) or Forward primer F2 (5’AATATTGTGTGTATGGGCACAAACCCTTG3’) inside the NPR2 gene and Reverse primer R3 (5’ CCTCAATGTCTCTTCTATCGGAAT3’) that anneals to the carboxy terminus of CAN1 gene (ectopic BIR assay). Control PCR was performed with 1 minute extension time for 20 cycles using Forward primer F3 (5’TCCATGCTAGATTAGCACACAGTAA3’) and reverse primer R4 (5’CTTTTGTAGGTGTCCTTAATTTCCA3’). PCR products were resolved in a 1% agarose gel and stained with ethidium bromide. The gel pictures were captured by GEL-LOGIC system and PCR bands were quantified by Carestream software. Inversed images were showed in the Fig. 6 and S11 Fig. A fluctuation-based test was performed to determine gross chromosomal rearrangements [70]. In brief, 3–5 ml of 11 confluent cultures in YEPD from single colonies were plated on SC medium lacking arginine and supplemented with 60 mg/L L-canavanine and 1 g/L 5-FOA for scoring GCRs. Cells were also diluted and plated on YEPD plates for counting total number of cells. GCR plates were cultured for 5–10 days at 30°C and colony numbers counted. Calculation of GCR rate and determination of junction type (de novo telomere formation or translocation) were as previously described [70]. | Chromatin poses a barrier to the recombination process. Chromatin modification is therefore a prerequisite factor for the efficient execution of the recombination event. Chromatin remodeling and several unique histone modifications at or near DNA double strand breaks (DSBs) facilitate early recombination processes, but little is known how chromatin state impinges on post-invasion steps of recombination, such as repair synthesis through homologous template, particularly recombination subtypes such as break-induced replication (BIR) involving extensive repair synthesis. Here, we investigated the effect of deletions in chromatin modification and remodeling genes on BIR and discovered that hyper-acetylation of H3K56 selectively impairs BIR and gene conversion associated with long DNA gap synthesis. We also found that hyper-acetylation of H3K56 interferes with the recovery from replication stress in checkpoint deficient cells and induces translocation-type gross chromosomal rearrangements (GCRs). The results provide a basic understanding of how histone modification facilitates efficient fork progression in recombination, controls the types of the repair products and sustains chromosome integrity upon induction of genotoxic stress. | Abstract
Introduction
Results
Discussion
Materials and Methods | 2015 | Hyper-Acetylation of Histone H3K56 Limits Break-Induced Replication by Inhibiting Extensive Repair Synthesis | 13,454 | 291 |
|
Several regulators are involved in the control of cell cycle progression in the bacterial model system Caulobacter crescentus, which divides asymmetrically into a vegetative G1-phase (swarmer) cell and a replicative S-phase (stalked) cell. Here we report a novel functional interaction between the enigmatic cell cycle regulator GcrA and the N6-adenosine methyltransferase CcrM, both highly conserved proteins among Alphaproteobacteria, that are activated early and at the end of S-phase, respectively. As no direct biochemical and regulatory relationship between GcrA and CcrM were known, we used a combination of ChIP (chromatin-immunoprecipitation), biochemical and biophysical experimentation, and genetics to show that GcrA is a dimeric DNA–binding protein that preferentially targets promoters harbouring CcrM methylation sites. After tracing CcrM-dependent N6-methyl-adenosine promoter marks at a genome-wide scale, we show that these marks recruit GcrA in vitro and in vivo. Moreover, we found that, in the presence of a methylated target, GcrA recruits the RNA polymerase to the promoter, consistent with its role in transcriptional activation. Since methylation-dependent DNA binding is also observed with GcrA orthologs from other Alphaproteobacteria, we conclude that GcrA is the founding member of a new and conserved class of transcriptional regulators that function as molecular effectors of a methylation-dependent (non-heritable) epigenetic switch that regulates gene expression during the cell cycle.
Epigenetic signals, such as methylation of DNA, play an important role in the regulation of gene expression in eukaryotes. Methylation of adenines in the N6 position (m6A) has been described in Bacteria, Archaea, Protists and Fungi. Though known for its protective role in bacterial restriction/modification systems [1], m6A also fulfills cellular functions in Gammaproteobacteria, including the initiation of DNA replication, transposition, mismatch repair, and virulence gene expression [2]–[4]. In the Alphaproteobacteria such as Caulobacter crescentus, Sinorhizobium meliloti, Brucella abortus and Agrobacterium tumefaciens, the solitary methyltranferase CcrM is required for efficient growth, presumably through gene expression control of critical cell cycle genes [5]–[8]. The cell cycle role of CcrM was originally described in C. crescentus [5], [7]. At each cell division, C. crescentus generates two different cells (stalked and swarmer) committed to specific stages of the cell cycle [9]. The stalked cell is able to replicate the DNA (S-phase) and it possesses an extension of the external envelope and membranes, called stalk. The swarmer cell is instead motile and non-replicative (G1-like) possessing a polar flagellum and several pili. Upon nutrient availability the swarmer cell differentiates in a stalked cell, resembling the eukaryotic G1→S transition. In this cyclical progression, a crucial role is played by CtrA, an essential transcriptional regulator that targets many cell cycle genes [10]. Its activity and abundance are precisely regulated in time and space through phosphorylation, proteolysis and transcription. In G1, CtrA∼P inhibits DNA replication by repression of the origin of replication [11] and only upon CtrA proteolysis or dephosphorylation, DnaA-mediated chromosome replication initiation occurs [12] committing cells to the S phase. The re-synthesis of CtrA requires transcription of ctrA that relies on the methylation-sensitive ctrAP1 promoter [13] whose activation depends on GcrA, an enigmatic factor that is encoded in the genomes of Alphaproeobacteria and several Caulophages [14], [15]. While in Caulobacter GcrA accumulates in early S phase and is confined to stalked cells [16] for the activation of ctrAP1, a second, auto-regulatory promoter, ctrAP2, reinforces ctrA transcription later in S-phase. Upon CtrA synthesis, an essential phosphorelay, composed of CckA and ChpT [17], phosphorylates CtrA that in turn activates transcription of the ccrM gene. After the introduction of m6A marks in the context of GAnTC sequences [5] CcrM is proteolyzed prior to cell division by the Lon protease [7]. How the m6A marks, introduced by CcrM, affect transcription is unclear, but the marks are transient as DNA replication converts the (full) methylation on both DNA strands to the hemi-methylated state, until strands are re-methylated in a distributive manner [18] once CcrM has accumulated at the end of S-phase. The time a given genomic locus spends in the hemi-methylated state is thus pre-determined by its physical proximity to the origin of replication [19], a feature that might be exploited to couple activation of certain promoters, such as ctrAP1, with replication progression [19]. GcrA and CcrM are implicated in the transcriptional regulation of ctrAP1, suggesting linked roles. While an underlying biochemical relationship is also hinted by the analysis of the gene occurrence pattern in the Alphaproteobacterial genomes [20], this link remains experimentally untested. Here we use chromatin-immunoprecipitation, biochemical, genetic and biophysical methods to explore the basis of transcriptional activation by GcrA. We uncovered that GcrA binds preferentially m6A-marked DNA and that it associates with the RNA polymerase, presumably to facilitate transcription initiation at methylated promoters. To assess if this mechanism is specific for Caulobacter or instead evolutionarily conserved, we performed experiments with GcrA orthologs in other Alphaproteobacteria, observing essentially the same behaviour. We conclude that GcrA and CcrM define an important regulatory pair, which is evolutionarily conserved and has been appropriated for epigenetic control of cell cycle transcription in Alphaproteobacteria.
Because GcrA is a conserved protein that lacks primary structural resemblances to known functional domains, we investigated the features of the primary structure of GcrA by bioinformatic prediction using SMART [21]; first, a non-significant homology (E-value equal to 734) to helix-turn-helix domains was detected (13–55 aa). Also GcrA has a high content of positively charged residues such as arginine and lysine mostly located in the central region (45–80 aa). Those features may support the ability of GcrA to bind DNA directly (see next sections) through the N-terminal part. Consistently, the N-terminal part is also the region of GcrA that is more conserved at the evolutionary level across orthologs of GcrA shown in the Figure S1. This conservation suggests an important functional role of the region, for example in the specific DNA binding and also putative interactions with other cellular factors. With the goal of investigating the interactions of GcrA with DNA and its targets in vivo we purified an N-terminally hexa-histidine tagged variant of GcrA (His6-GcrA) from an E. coli overexpression strain by sequential affinity and gel filtration chromatography and characterized its biophysical properties (see Materials and Methods). SDS-PAGE (Figure S2) and dynamic light scattering (DLS) analysis (data not shown) indicated a highly pure (>95%) and monodisperse preparation of His6-GcrA. Prediction of unfolded regions using RONN suggest that regions 41–105 aa and 145–173 aa of GcrA are disordered, while the software SOPMA [22] suggested that GcrA is partially structured in the N-terminal region (Figure S3) with an organization in three predicted alpha helices suggesting a folded structure. To gain insight into the spatial organization, we conducted Small Angle X-ray Scattering (SAXS) analysis (Protocol S1 and Table S1) that allows the determination of shape, size and oligomerization status of macromolecules in solution (Figure S4A). SAXS estimates the molecular mass of His6-GcrA at 42 kDa, which corresponds to a dimeric organization. Also the dimensions of His6-GcrA, by using computed radius of gyration (Rg) and maximum dimension (Dmax) values, combined with the pair-distance distribution function, P (R), shape and the Kratky plot representation, are consistent with an elongated form and partially disordered conformation of His6-GcrA dimers in solution (See legend of Figure S4A for more technical details). Possibly the interaction of GcrA with other proteins and with DNA can induce a stabilization of the disordered regions. Next, we performed limited proteolysis followed by MALDI-TOF mass spectrometry (MS) analysis in order to identify regions of His6-GcrA that were more resistant to proteolysis indicating its more compact (structured) nature. Two different proteases, Thermolysin and V8 (see Materials and Methods) were used and the most resistant fragments to proteolysis were analyzed by MS (Figure S4B). This analysis suggests that the N-terminal part of GcrA (from 1 to 115 ca.), although containing proteolytic sites for both proteases, was more stable than the C-terminal part, as also indicated by the prediction of alpha helical structures in the N-terminal portion of GcrA. In light of these structural features suggesting that the N-terminal domain of GcrA binds DNA, we sought specific in vivo targets of GcrA that could be used to probe DNA-binding of GcrA in vitro. Previous non-quantitative chromatin-immunoprecipitation (ChIP) experiments using polyclonal antibodies to GcrA, provided support for the notion that GcrA affects the transcription of cell cycle genes by, directly or indirectly, associating with specific chromosomal sites [16]. To illuminate the basis for this selectivity and the underlying mechanism of transcriptional regulation by GcrA in Alphaproteobacteria, we subjected ChIP samples from NA1000 wild-type cells to deep sequencing (ChIP-Seq) [23] (Protocol S2). By mapping the reads onto the genome, we determined the binding profile of GcrA on genomic regions. First, we used a peak finding strategy to identify regions bound by GcrA with high affinity (Protocol S2); this analysis allowed to identify 218 peaks that were subsequently associated to the closest genes (Table S2). Inspection of the top 50 targets (Figure 1A) revealed wide peaks (ca. 1 kbp wide, data not shown). We found that half of these GcrA-bound sequences were significantly associated with CcrM methylation sites (GAnTC). To explore this association more in details, we calculated the average number of methylation sites in 1 kbp windows centered on the peaks and found it to be close to 2 (1. 78), in comparison with 0. 58 sites per random 1 kbp genomic regions (Figure 1B). These results clearly indicate a significant enrichment of methylation sites in genomic regions bound in vivo by GcrA, raising the possibility that methylation enhances GcrA binding to its targets. Next, in all promoter regions, defined from 300 bp upstream to 100 bp downstream a gene' s start codon, we calculated the number of ChIp-Seq reads (see Protocol S2) (Table S3). We obtained (Z-score ≥2) 161 GcrA-bound promoter regions, 89 of which also contained a GAnTC methylation site (data not shown). This list contained many known GcrA-controlled targets such as mipZ, encoding a division regulator [24], podJ encoding a polarity factor [25], [26] and ctrA. We observed a remarkably small overlap with the genes previously identified as GcrA-dependent by DNA microarrays [16]. Only 5 genes passed the threshold (Z score ≥2), including those encoding CCNA_01542 (ice nucleation protein), CCNA_01556 (LacI family transcriptional regulator), CCNA_01766 (hypothetical protein), CCNA_02005 (inosine-uridine preferring nucleoside hydrolase), CCNA_02086 (sporulation domain containing protein), CCNA_02246 (mipZ), CCNA_02401 (encoding a transcriptional regulator) and CCNA_03325 (encoding a hypothetical protein). Since microarrays detect both direct and indirect mRNA abundance changes, it is likely that many genes whose expression was affected by GcrA depletion were, in fact, indirect targets of GcrA presumably under the control of other transcription factors, such as CtrA. In order to test if GcrA binds DNA in vitro, we set up an electrophoretic mobility shift assay (EMSA) with His6-GcrA and regions identified as in vivo targets of GcrA by the ChIP-seq experiment described above. We selected 5 regions as EMSA probes, each with distinct features: 1) the preferred target (with the maximum number of reads in Table S2) corresponding to the N-terminal coding sequence of gene CCNA_00697 that has three GAnTC sites; 2) the intergenic sequence between CCNA_00278 and CCNA_00279 that is efficiently bound by GcrA in vivo, but has no predicted GAnTC methylation site; 3) the promoter of the gene mipZ, which was also discovered previously by microarray as a GcrA-dependent gene and has two juxtaposed GAnTC sites; 4), the P1 promoter of ctrA (ctrAP1) that has one GAnTC site (there is another GAnTC sequence after the transcription start site of the ctrAP2 promoter) and is thought to be activated by GcrA [16]; 5) a negative control, corresponding to the intergenic region between CCNA_01926 and CCNA_01927 which GcrA binds non-specifically in vivo based on ChIP-seq data. Probes were designed as 70-mer double stranded oligo-nucleotides, in which the central part corresponds to the genomic region with the highest number of ChIP-seq reads. The EMSA (Figure 2) showed that His6-GcrA shifted four out of five probes, indicating the formation of a His6-GcrA•DNA complex with sequences identified by ChIP-Seq, except for the intergenic sequence between CCNA_01926 and CCNA_01927 (i. e. , the negative control). All positive probes gave rise to two distinct His6-GcrA•DNA complexes with similar migration properties, possibly reflecting different oligomeric states of His6-GcrA with different apparent affinities for the DNA (see below). In particular, probe CCNA_00697 was the most efficiently bound by His6-GcrA (Kd = 4±0. 5 µM); probes ctrA (Kd = 6. 5±0. 5 µM) and mipZ (Kd = 8. 5±0. 5 µM) also showed DNA binding however the complex forms only at a higher concentration of His6-GcrA, mirroring, with the exception of the intergenic region between CCNA_00278 and CCNA_00279 (Kd>9 µM), the binding preference ChIP-seq in vivo. The EMSA results also demonstrate that His6-GcrA binds DNA in a specific fashion in vitro. Considering also that the conserved GcrA protein has no homology with known DNA binding proteins at the primary structure level, we conclude that GcrA defines a new class of alphaproteobacterial DNA binding proteins that directly interacts with target promoters to control transcription of many Caulobacter S-phase genes, including the gene encoding the master regulator CtrA. Although methylation sites are associated with GcrA-bound regions, GcrA apparently can also bind sequences that do not contain GAnTC methylation sites, based on the methylation-dependent binding experiments described below, we suggest that multiple DNA constrains exist in the GcrA specificity, perhaps involving m6A marks in different sequences contexts or a different type of methylation mark altogether. In this context, it is noteworthy that two putative cytosine methyltransferases are encoded in the C. crescentus genome [27]. To test if GAnTC methylation modulates the binding of His6-GcrA to some of its targets in vitro, we conducted EMSA competition experiments with the un-methylated region of PmipZ, CCNA_00697 and ctrAP1 as biotinylated probes and double stranded synthetic oligos harboring N6-adenosine methylation at GAnTC on either one or both strands as competitors. In these experiments, His6-GcrA was pre-incubated with the unlabeled competitor DNA, followed by the addition of the biotinylated probe. The more the unlabeled competitor DNA reduces the abundance of the shifted His6-GcrA•DNA complex, the higher the affinity of His6-GcrA is for the unlabeled competitor. As shown in Figure 3, we observed a clear preference of GcrA for the methylated competitors over the un-methylated one, with those carrying methylation on both strands (“full”-methylation) competing better than either one harbouring the methylation on one of the two strands (“hemi”-methylation). Remarkably, in the case of the CCNA_00697 and mipZ competitors, hemi-methylation on the “sense” strand is a better competitor than the hemi-methylated competitor with the modification on the other strand. For promoters of ctrA and mipZ, the calculated Kds provided quantitative confirmation of the results shown in Figure 3 (Figure S9B and S9C). In order to assess if methylation alters the disposition of GcrA on its target DNA, we conducted DNase I protection assay using fully and hemi- (GAnTC) methylated fluorescently-labeled ctrAP1 promoter probes. As shown in Figure 4A, GcrA protects specific regions of the probe in a methylation-dependent manner, giving rise to a larger region of protection spanning the −35 to the −10 of the ctrAP1 promoter with the fully-methylated (i. e. on both strands) probe. By contrast, the protection of the hemi-methylated (on the plus strand) probe was confined to a region adjacent to the methylation site itself. Importantly, the un-methylated probe and the hemi-methylated probe carrying the modification on the minus strand did not show protection by His6-GcrA at any concentration. We conclude that methylation induces different oligomerization or conformational states in strand-specific manner. Next, we complemented the DNase I protection experiments of the target, with protection experiments of His6-GcrA by limited proteolysis in the presence or absence of the various methylated ctrAP1 probes (Figure 4B). We found that efficiency of proteolysis is accelerated in the presence of methylated probes, suggesting that conformational rearrangements are induced by the methylated probe to expose protease hypersensitive sites, akin to the DNase I hypersensitive sites of the target DNA that become exposed only when GcrA associates with a methylated target, but not in the presence of the un-methylated site. We also ruled out the possibility that the oligos affected the DNase I activity by incubating another protein (His6-ChpT) [28], which was proteolyzed identically with or without the DNAs (data not shown). CcrM-dependent methylation of ctrAP1 was previously proposed as an essential element of a transcriptional regulatory switch, culminating in the GcrA-dependent activation of ctrAP1 upon the conversion from full- to hemi-methylation [19]. Intriguingly, our results reveal that His6-GcrA binds hemi-methylated versus fully methylated ctrAP1 in strikingly different manner, with the latter covering a much larger area. This raises the possibility that cooperative interactions, induced by the transition from hemi- to full-methylation mediated by CcrM, can lead to a wider and stronger association of GcrA with the target DNA. As His6-GcrA wraps the entire region from −35 to +7 of fully methylated ctrAP1, it may interfere with RNA polymerase holo-enzyme (RNAP) at ctrAP1; a possibility that must be explored in future work. By contrast, on the hemi-methylated plus strand of ctrAP1, His6-GcrA occupies a 12 nt stretch overlapping the −35 region and adjacent GAnTC site and with lower affinity a 12 nt region from +5 onwards. GcrA could compete with RNAP for binding to ctrAP1 or alternatively tether it at the promoter, preventing promoter clearance, i. e. the switch from transcription initiation to the elongation phase. Furthermore the methylation strand-specificity opens the possibility of an “asymmetric” mechanism of gene regulation, in which only one of the two replicated loci is preferentially bound and transcriptionally regulated by GcrA before re-methylation by CcrM in the pre-divisional cell. Such, a regulatory bias could have far reaching consequences in all forms of spatiotemporal and/or of gene-dosage regulation for all living cells, as it has been suggested before for PapI-promoted Lrp binding to hemi-methylated sites in uropathogenic E. coli [29]. To explore the models described above, we tested whether GcrA can directly or indirectly associate with RNAP. To this end, we passed a soluble C. crescentus cell lysate, in which DNA was fragmented by a mild DNase I treatment, over a nickel-NTA column that had been pre-loaded or not with His6-GcrA. Following extensive washes with buffer containing up to 1 M NaCl, we eluted His6-GcrA and associated proteins with buffer containing imidazole (see Materials and Methods). Blots harbouring these samples were then probed with antibodies to the β subunit of core RNAP, revealing that RNAP β in the eluate from the His6-GcrA pre-loaded column only (Figure 5A). We extended these findings by showing that E. coli RNAP core enzyme can associate with the DNA•GcrA complex in an EMSA using ctrAP1 promoter. Increasing concentration of RNA polymerase clearly showed the formation of lower mobility complex whose formation was dependent on the presence of GcrA (Figure 5B). This interaction of RNAP with GcrA bound to its target was also observed with the mipZ promoter (Figure S5). Taken together, these results indicate that GcrA binds components of the RNA polymerase core complex and they provide a mechanistic explanation for how GcrA might affect gene transcription. The connection between methylation by CcrM and DNA-binding of GcrA, seen in vitro, together with the association of GcrA to RNAP, prompted us to explore if other GcrA target promoters are regulated in a methylation-dependent manner in vivo. To this end, we fused several promoters that have methylation sites and that emerged as in vivo targets of GcrA in the ChIP-seq experiments to the promoter-less lacZ reporter gene. We first confirmed the GcrA-dependence of these promoters by measuring lacZ-encoded β–galactosidase activities under GcrA-replete and deplete conditions using a ΔgcrA: : Ω; xylX: : Pxyl-gcrA strain [16] in which GcrA expression is induced in the presence of xylose and repressed in the presence of glucose (Figure 6A). After 5 h of depletion of GcrA in glucose, the β–galactosidase activities of the PmipZ-, PpodJ-, PflaY - and PpleC-lacZ reporters dropped by ca. 60% compared to the WT grown in glucose or the Pxyl-gcrA strain grown in xylose, and immunoblotting revealed a strong reduction in MipZ and PodJ abundance (Figure S6). It was previously also shown that activation of the ctrAP1 promoter requires GcrA [19]. By contrast, the PCCNA_00697-lacZ reporter only exhibited a 31% reduction in β–galactosidase activity under the same condition, possibly because as the preferred in vivo target of GcrA, residual GcrA that remains in the cell clings to the CCNA_00697 promoter (Figure 6B). Next, we asked if a mutation of the GAnTC influences the promoter activity. To this end, all GAnTC sites in a promoter fragment of the lacZ reporter construct were mutated to GCnTC and the activity of the mutant promoters assayed by β–galactosidase measurements. The mutant promoters were crippled by 60–80% and immunoblotting showed the PodJ and MipZ failed to accumulate in cells lacking CcrM (Figure S6). Interestingly, the mutant PCCNA_00697-lacZ reporter showed a different response, retaining WT (100%) activity. Interestingly, the effect on PCCNA_00697-lacZ is mirrored for ctrAP1 whose activity was also unchanged by mutation of the GAnTC site to prevent CcrM-dependent methylation (Figure 6C) [19]. To test if this response was typical of GcrA-dependent promoters that are distal to the replication terminus, we analysed the methylation/GcrA dependency of another promoter, tipF (CCNA_00747), at a comparable location with respect to the origin of replication (). Unlike ctrAP1 and the region at CCNA_00697, tipF promoter activity requires GcrA and an intact GAnTC methylation site (see below). To explore if methylation at GAnTC is required for GcrA to associate with its target sites in vivo, we compared the genome-wide promoter occupancy of GcrA in WT and ΔccrM cells by ChIP-seq (Figure 7A). Analysis of the two data sets (Table S3) unearthed a major role of GAnTC methylation in directing GcrA to target promoters, with 80 loci requiring CcrM to be efficiently bound by GcrA, including CCNA_00697, mipZ, podJ, flaY (encoding a putative flagellar regulator), pleC (encoding a developmental histidine kinase/phosphatase) and to a lesser extent ctrA. However for ctrAP1, detailed analysis of the ChIP-Seq traces (Figure S7) revealed that GcrA binding dropped in proximity to the GAnTC methylation sites. Immunoblotting confirmed that no apparent difference in the GcrA steady-state levels were discernible in WT and ΔccrM cells (Figure 7A). To corroborate the ChIP-seq data, we performed ChIP analysis of WT and ΔccrM cells and measured the abundance of precipitated (GcrA-bound) mipZ and podJ promoters by quantitative real-time PCR (qChIP). As shown in Figure 7B, in the absence of CcrM, GcrA occupancy is reduced by 70% and 60%, respectively. If CcrM-dependent GAnTC methylation is required for GcrA binding to its targets, then the corresponding promoters should not be methylated in ΔccrM cells. Because other methyltransferases might also contribute to adenosine methylation at the N6 position (m6A), we first determined the abundance of m6A across the genomes of WT and ΔccrM cells by ChIP-Seq analysis using an m6A-specific polyclonal antibody (Figure 7C). This analysis revealed that chromosomal loci, particularly towards the replication terminus, carry abundant CcrM-dependent m6A marks (Table S3). We validated this conclusion by qChIP experiments for two promoters proximal to the terminus, PmipZ and PpodJ, (Figure 7D) and a distal one, PtipF (Figure S10). To explore if GcrA-controlled functions are conserved across the Alphaproteobacteria, we introduced the GcrA ortholog [20] from Brucella melitensis biovar abortus 2308 (BAB1_0329) and Sinorhizobium meliloti Rm1021 (SMc02139) under the control of an xylose-inducible promoter on a low-copy plasmid [30] in C. crescentus, harbouring a temperature sensitive allele of gcrA with a Thr→Pro mutation at position 10 and evaluated their ability to support growth at the restrictive temperature [16]. As shown in Figure 8A, both B. abortus and S. meliloti gcrA orthologs are able to support viability of the strain gcrAts at the restrictive temperature (37°C) following induction with xylose. Orthologs of GcrA from S. meliloti and B. abortus, although able to complement GcrA functions, revealed morphological diversities in C. crescentus (Figure 8B), likely owing to differences in abundance, activity and/or target specificity of these GcrA versions. Regardless, the complementation of the temperature-sensitive phenotype indicates that the function and target site specificity of GcrA orthologs are similar. We confirmed this result by testing directly the ability of GcrA orthologs to bind the Caulobacter target promoters. Therefore B. abortus and S. meliloti GcrA with an N-terminal His6 tag were purified from E. coli overexpression strains (Figure S2). EMSA experiments using target sites of C. crescentus GcrA revealed that these GcrAs are able to bind DNA efficiently and with the same apparent specificity (Figure 9A). Surprisingly the B. abortus and S. meliloti GcrA orthologs are able to form multiple complexes with different migration properties, likely due to structural and/or charge differences. Finally we tested if binding of B. abortus and S. meliloti GcrAs is also stimulated by GAnTC methylation. EMSAs showed that methylation still affects GcrA binding, as fully methylated probes showed a stronger binding affinity in comparison with hemi-methylated and even more with non-methylated DNA. However the asymmetry in binding efficiency found for certain regions of DNA (Figure 9B) appeared different in other GcrAs with respect to the C. crescentus one. Despite the pervasive effects of adenosine methylation on transcription in various bacterial lineages, our understanding of the underlying operating principles in these systems is still limited. With the identification and genetic/biochemical characterizations of the m6A-marked promoters and the transcriptional effector (s) recognizing them, we elucidated a crucial first step towards understanding the physiological underpinnings and the evolution of these epigenetic control systems in Alphaproteobacteria. Studies with the methylation-sensitive ctrAP1 promoter of C. crescentus as model led to the appealing model that replication of a given locus by DNA polymerase converts the promoter from the fully to the hemi-methylated state at a specific time in the cell-cycle that is dictated by the relative distance of the promoter from the origin of replication (ori). For ctrAP1, the hemi-methylated state was thought to be a prerequisite for GcrA-mediated activation, while full (re) -methylation by CcrM at the end of S-phase was viewed as the event leading to promoter silencing. Our experiments not only establish GcrA as a methylation-dependent transcription factor binding ctrAP1 and other promoters in vivo and in vitro (Figure 10), but they may suggest an elegant explanation for the methylation-induced regulation of expression. While activation of the hemi-methylated plus strand of ctrAP1 correlates with localized binding of GcrA to 13 nt over the −35 region, in the fully methylated state more than 40 nt of ctrAP1, are covered. Once the DNA replication fork moves through the fully methylated ctrA locus in the ensuing cell cycle the binding state for hemi-methylated DNA is reinstated. Methylation seems to help recruiting GcrA to promoters but GcrA might interact with RNAP even in the absence of target DNA. Perhaps in the hemi-methylated state this binding allows the initiation of transcription and release of the polymerase, while in the fully methylated state GcrA could sequester RNAP, preventing its movement through the coding sequence. It is likely that the partially unstructured dimeric GcrA adopts compacter structure upon interacting with methylated target DNA or possibly RNAP, thus conferring these properties. Contrary to the view that CcrM- and GcrA-dependent control of ctrAP1 in response to DNA replication applies to all GcrA target promoters, we note that many ori-distal promoters (such as PmipZ, PpodJ, and PpleC), but also terminus-distal promoters (such as PtipF) that fire in early S-phase, also carry CcrM-dependent m6A marks that are required for the recruitment of GcrA. Promoters near the terminus will be replicated late in S-phase and are, thus, thought to reside in the hemi-methylated state only during a short window before the synthesis of CcrM. This begs the question what purpose of m6A marks at these promoters may serve, since methylation change by replication should not be temporally correlated with promoter activation. As many promoters of cell division (e. g. , mipZ), motility (e. g. , flaY) and polarity genes (e. g. , podJ) carry m6A marks that are recognized by GcrA (Figure 10), CcrM-dependent methylation might serve as a quality control function or coupling mechanism to prepare these promoters for activation in the ensuing cell division cycle once GcrA is expressed. In the gammaproteobacterium Vibrio cholerae, the origin-binding protein of chromosome II RctB is recruited to sites carrying m6A marks that have been introduced by the GATC-specific Dam methyltransferase [4]. Thus, while full methylation also has been adopted for regulatory purposes, different effectors and processes have been selected during evolution.
C. crescentus strains were grown in peptone-yeast extract (PYE, rich medium) at 30°C [31] or 37°C as necessary, tetracycline (1 µg/ml), kanamycin (25 µg/ml), spectinomycin/streptomycin (100-5 µg/ml) 0. 1% glucose, or 0. 1% xylose, as required. E. coli strains were grown at 20°C or 37°C in LB broth supplemented with ampicillin (100 µg/ml), as necessary. Plasmids were transformed into C. crescentus and E. coli BL21 (DE3) by electroporation. Plasmids and strains are listed in Table S5. To construct the ΔccrM mutant UG2212, the ΔccrM: : Ω mutation was transduced by ϕCr30-mediated generalized transduction from LS2144 [6] into NA1000. One transductant was selected on spectinomycin/streptomycin-containing medium and subjected to whole genome Illumina sequencing by Fasteris SA (Geneva, Switzerland). The genome sequence of UG2212 failed to reveal point mutations or insertions/deletions compared to the parent. DNA fragments from C. crescentus, S. meliloti and B. abortus were amplified by PCR using cell lysates or genomic DNA for Brucella using Pfu-Turbo (Life Technologies, www. lifetechnologies. com/) following a protocol as recommended by the manufacturer. Primers are listed in the Figure S8. PCR products were then transferred in pENTR by Directional TOPO Cloning (Life technologies, www. lifetechnologies. com/), sequence verified and then transferred in pET derivatives His6-tagged destination vectors for E. coli BL21 expression, or pMR20 destination vector for xylose inducible expression in C. crescentus strains [30]. Transcriptional reporters were made by cloning PCR-amplified promoters fragments (Figure S8) into plac290 using EcoRI/XbaI [32]. The full-length DNA fragment of GcrA was cloned into the pET15 derivative, pML375 vector [30], obtaining a recombinant protein with a His6-Thrombin cleavage site tag in N-terminus of the protein. Overexpression of GcrA was induced in E. coli BL21 (DE3) at OD (600 nm) 0. 6–0. 8 by 500 µM isopropyl-b-D-thiogalactoside (IPTG) O/N at 20°C. The cells were harvested by centrifugation and then resuspended in lysis buffer (PBS 1× pH 7. 5,0. 2 M NaCl, 1 mM DTT, 0. 1% Triton, supplemented with Complete Protease Inhibitor Cocktail (Roche, http: //www. roche. com/) and DNase I (Euromedex, www. euromedex. com/) and lysed by Emulsiflex (Avestin, www. avestin. com/) at 10°C. From the supernatant, GcrA was purified in two steps of purification, first, using Ni2+-nitrilotriacetate affinity resin (Ni-NTA) (Qiagen, www. qiagen. com/) equilibrated with lysis buffer and eluted by PBS 1× (pH 7. 5), 0. 5 M Imidazole, followed by Gel filtration step using HiLoad 16/60 Superdex 75 prep grade (GE Healthcare, www. gehealthcare. com/) equilibrated with running buffer (0. 1 M Tris pH 8. 5,0. 2 M NaCl, 5% Glycerol) optimized after DLS (See below the experimental procedure). DLS measurements by the Zetasizer nano ZS (Malvern, www. malvern. com/) with an accuracy of 0. 1°C were performed immediately after both the size exclusion step and the concentration step in order to find the best buffer composition. DLS was employed to estimate the thermo-stability of protein samples in different buffer solutions from to 15°C to 64°C, one degree steps. DLS was also used for the estimation of monodispersity of GcrA preparation. Purified His6-GcrA was digested with proteases Thermolysin (Sigma-Aldrich, www. sigmaaldrich. com/) and Endoproteinase GluC V8 (New England Biolabs, www. neb. com/) (25°C with 0. 5 mg/ml GcrA in 20 mM TRIS pH 8,150 mM NaCl for digestion with Thermolysin and 20 mM Tris (pH 7. 6), 1 mM CaCl2 in case of digestion with V8). The protease/substrate ratio was 1∶100 (w/w) in each case. At different time intervals, aliquots of the proteolysis reactions were stopped with loading buffer. The protein samples were then analyzed by SDS-PAGE and the fragments analyzed by Trypsin digestion and mass spectrometry. Proteolysis control of His6-ChpT [28] in presence of differentially methylated DNAs was performed as described above. Nickel columns loaded by His6-GcrA were also used for affinity chromatography showed in Figure 5B. A 1 ml HisPur-Ni-NTA Chromatography Cartridge (Qiagen, www. qiagen. com/), equilibrated with running buffer (0. 1 M Tris pH 8. 5,0. 15 M NaCl, 5% Glycerol) was loaded at 15°C with 23 mg of histidine-tagged C. crescentus GcrA (His6-GcrA) that was prepared as previously described, and washed with 15 volumes of running buffer. Meanwhile, 2 liters of C. crescentus cells (OD 600 nm of 0. 6) were harvested by centrifugation (5000 rpm, 20 min, 4°C) and resuspended in 30 ml of lysis buffer (0. 1 M Tris pH 8. 5,0. 15 M NaCl, 1 mM DTT, 0. 1% Triton, supplemented with Complete Protease Inhibitor Cocktail (Roche, www. roche. com/) and DNase I (Euromedex, www. euromedex. com/) ) and lysed by Emulsiflex (Avestin, www. avestin. com/) at 10°C. The lysate was then centrifuged at 9500 rpm, 20 min, 4°C and the supernatant obtained was applied to the column. The column was eluted with running buffer NaCl gradient from 0. 15 M and 1 M of NaCl. A last wash was done in presence of Imidazole (0,1 M Tris pH 8,5, 0,15 M NaCl, 5% Glycerol, 0. 5 M Imidazole) in order to remove the His6-GcrA and proteins still bound at the column. The negative control to this experiment was performed doing the same procedure with a 1 ml HisPur Ni-NTA Chromatography Cartridge without His6-GcrA. The eluted samples were run in SDS-PAGE gel and transferred to nitrocellulose membrane. The membrane was blocked with PBS, 0. 1% NP-40 and 3% dry milk for 1 hour at room temp. The membrane was incubated with anti-RNA polymerase B-subunit antibody (Thermo Scientific, www. pierce-antibodies. com/) against the β-subunit (1∶5000) at 4°C overnight. Each membrane was washed 5 times each for 10 min with PBS containing 0. 1% NP-40, followed by incubation with the secondary antibody (1∶50,000) for 45 min. The membrane was developed following the procedure described under immunoblot section. EMSAs were performed using the LightShift Chemiluminescent EMSA Kit (Thermo Scientific). Briefly, different versions of GcrA were incubated at room temperature in 10 mM Tris pH 7. 5,100 mM KCl, 0. 5 mM DTT, 50 ng/µl poly (dI-dC), and 0. 05% Nonidet P-40 binding buffer with 5 fmol of a biotin-labeled DNA fragment for 25 minutes. After 25 min incubation at room temperature, samples were resolved by a 10% non-denaturing polyacrylamide gel prepared in TBE buffer (450 mM Tris, 450 mM boric acid and 0. 01 mM EDTA). The samples were blotted onto a 0. 45-µm Biodyne B nylon membrane (Thermo Scientific, www. piercenet. com/) at constant current of 300 mA for 45 min at 4°C, and then cross-linked to the membrane using a 312 nm UV Transilluminator (Uvitec, www. uvitec. com.) for 10 min. Membranes were processed as recommended in the Chemiluminescent Nucleic Acid Detection Module Kit (Thermo Scientific, www. piercenet. com/). Competitive EMSAs were performed as described above, adding a preincubation step of 20 min at room temperature of GcrA and competitor DNAs before the usual 25 min GcrA/biotin-labeled DNA fragment incubation. EMSA in presence of RNA polymerase core enzyme (Epicentre, www. epibio. com/) was performed by pre-incubating GcrA in presence of RNAP for 20 min at room temperature before the usual incubation with biotin-labeled DNA. For detecting the binding region of GcrA, a 120 bp probe from ctrAP1 (Figure S8) was synthesized and labeled with Fam-6 (Eurogentec, www. eurogentec. com/). Single stranded probes containing m6A were also synthesized, which were later assembled into double stranded probes in different combinations. Five fmoles of probes were incubated at room temperature with increasing concentrations of purified GcrA as done with EMSA for 30 min. The samples were digested with approximately 7U of DNaseI (Euromedex, www. euromedex. com/) at room temp for 3 min. DNaseI was inactivated by adding 0. 1 M EDTA followed by incubation at 75°C for 10 min. The digested fragments were eluted using the mini-elute columns (Qiagen, www. qiagen. com/). The samples were run in a 3130 Genetic Analyzer (Life Technologies) as described before [33], analyzed by GelQuest (SequentiX, www. sequentix. de/). Sequencing reactions were also performed using Thermo Sequenase Dye Primer Manual Cycle Sequencing Kit (Affymetrix, www. affymetrix. com/) using the probe region as a template and a sequencing primer labeled with FAM at the 5 primes. β-Galactosidase assays were performed at 30°C as described earlier [34], [35]. PVDF (polyvinylidenfluoride) membranes (Merck-Millipore, www. merckmillipore. com) were blocked with PBS, 0. 05% tween 20 and 5% dry milk for 1 h and then incubated for 1 h with the primary antibodies diluted in PBS, 0. 05% tween 20,5% dry milk. The membranes were washed 4 times for 5 min in PBS and incubated 1 h with the specific secondary antibody diluted in PBS, 0. 05% tween 20 and 5% dry milk. The membranes were finally washed again 4 times for 5 min in PBS and revealed with Immobilon Western Blotting Chemoluminescence HRP substrate (Merck Millipore, www. merckmillipore. com/). The different antisera were used at the following dilutions: anti-CcrM (1∶10; 000), anti-DnaA (1∶10; 000), anti-GcrA (1∶10,000), anti-CtrA (1∶10,000). Anti-RNA polymerase beta antibodies (Abcam, www. abcam. com/) were used using the protocol described in “Affinity chromatography for RNAP detection”. Mid-log phase cells were cross-linked in 10 mM sodium phosphate (pH 7. 6) and 1% formaldehyde at room temperature for 10 min and on ice for 30 min thereafter, washed thrice in phosphate buffered saline (PBS) and lysed in a Ready-Lyse lysozyme solution (Epicentre, Madison, WI) according to the manufacturer' s instructions. Lysates were sonicated (Sonifier Cell Disruptor B-30) (Branson Sonic Power. Co. , www. bransonic. com/) on ice using 10 bursts of 20 sec at output level 4. 5 to shear DNA fragments to an average length of 0. 3–0. 5 kbp and cleared by centrifugation at 14,000 rpm for 2 min at 4°C. Lysates were normalized by protein content, diluted to 1 mL using ChIP buffer (0. 01% SDS, 1. 1% Triton X-100,1. 2 mM EDTA, 16. 7 mM Tris-HCl [pH 8. 1], 167 mM NaCl plus protease inhibitors (Roche, www. roche. com/) and pre-cleared with 80 µL of protein-A agarose (Roche, www. roche. com/) and 100 µg BSA. Ten % of the supernatant was removed and used as total chromatin input DNA. Polyclonal antibodies to GcrA [16] and m6A (Synaptic Systems GmbH, www. sysy. com/) were added to the remains of the supernatant (1∶1,000 dilution), incubated overnight at 4°C with 80 µL of protein-A agarose beads pre-saturated with BSA, washed once with low salt buffer (0. 1% SDS, 1% Triton X-100,2 mM EDTA, 20 mM Tris-HCl (pH 8. 1), 150 mM NaCl), high salt buffer (0. 1% SDS, 1% Triton X-100,2 mM EDTA, 20 mM Tris-HCl (pH 8. 1), 500 mM NaCl) and LiCl buffer (0. 25 M LiCl, 1% NP-40,1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris-HCl (pH 8. 1) and twice with TE buffer (10 mM Tris-HCl (pH 8. 1) and 1 mM EDTA). The protein•DNA complexes were eluted in 500 µL freshly prepared elution buffer (1% SDS, 0. 1 M NaHCO3), supplemented with NaCl to a final concentration of 300 mM and incubated overnight at 65°C to reverse the crosslinks. The samples were treated with 2 µg of Proteinase K for 2 h at 45°C in 40 mM EDTA and 40 mM Tris-HCl (pH 6. 5). DNA was extracted using phenol∶chloroform∶isoamyl alcohol (25∶24∶1), ethanol-precipitated using 20 µg of glycogen as carrier and resuspended in 100 µl of water. Real-time PCR was performed using a Step-One Real-Time PCR system (Applied Biosystems, www. appliedbiosystems. com/) using 5% of each ChIP sample (5 µL), 12. 5 µL of SYBR green PCR master mix (Quanta Biosciences, www. quantabio. com/), 0. 5 µL of primers (10 µM) and 6. 5 µL of water per reaction. Standard curve generated from the cycle threshold (Ct) value of the serially diluted chromatin input was used to calculate the percentage input value of each sample. Average values are from triplicate measurements done per culture. The final data was generated from three independent cultures. The DNA regions analyzed by real-time PCR were from nucleotide −167 to +43 relative to the start codon of podJ, from −208 to +9 relative to the start codon of mipZ, from −185 to −16 relative to the start codon of ctrA. | Methylation of genomic DNA at a specific regulatory site can impact a myriad of processes in eukaryotic cells. In bacteria, methylation at the N6 position of adenosine (m6A) is known to mediate a non-adaptive immunity response to protect cells from foreign DNA. While m6A marks are not known to govern expression of cell cycle genes in Gammaproteobacteria, cell cycle transcription in the model alphaproteobacterium Caulobacter crescentus requires the m6A methyltransferase CcrM that introduces m6A marks at GAnTC sequences and the enigmatic factor GcrA. Investigating if a functional and biochemical relationship exists between CcrM and GcrA, we found that CcrM-dependent m6A marks recruit GcrA to the promoters of cell cycle genes in vitro and in vivo and is required for efficient transcription. GcrA interacts with RNA polymerase, explaining how cell cycle transcription is affected. Importantly, m6A-dependent binding is also seen in GcrA orthologs, indicating that this transcriptional regulatory mechanism by CcrM and GcrA is conserved in Alphaproteobacteria. | Abstract
Introduction
Results/Discussion
Materials and Methods | bacteriology
prokaryotic models
model organisms
caulobacter crescentus
gene expression
genetics
epigenetics
biology
dna modification
microbiology
gene networks
dna transcription | 2013 | DNA Binding of the Cell Cycle Transcriptional Regulator GcrA Depends on N6-Adenosine Methylation in Caulobacter crescentus and Other Alphaproteobacteria | 13,005 | 301 |
Netherton syndrome (NS) is a severe skin disease caused by the loss of protease inhibitor LEKTI, which leads to the dysregulation of epidermal proteases and severe skin-barrier defects. KLK5 was proposed as a major protease in NS pathology, however its inactivation is not sufficient to rescue the lethal phenotype of LEKTI-deficient mice. In this study, we further elucidated the in vivo roles of the epidermal proteases in NS using a set of mouse models individually or simultaneously deficient for KLK5 and KLK7 on the genetic background of a novel NS-mouse model. We show that although the ablation of KLK5 or KLK7 is not sufficient to rescue the lethal effect of LEKTI-deficiency simultaneous deficiency of both KLKs completely rescues the epidermal barrier and the postnatal lethality allowing mice to reach adulthood with fully functional skin and normal hair growth. We report that not only KLK5 but also KLK7 plays an important role in the inflammation and defective differentiation in NS and KLK7 activity is not solely dependent on activation by KLK5. Altogether, these findings show that unregulated activities of KLK5 and KLK7 are responsible for NS development and both proteases should become targets for NS therapy.
Netherton syndrome (NS) is a life-threatening autosomal recessive disorder that affects approximately one in 200 000 newborn children [1,2]. Newborns suffering from NS exhibit congenital ichthyosiform erythroderma with scaly and peeling skin, resulting in severe disruption of epidermal barrier, which in some cases is fatal. These conditions may improve with age and older patients often show less severe ichthyosis exhibiting erythematous plaques with double-edged scales at the periphery [1–3]. The hair of NS patients is usually thin, fragile and the patients often develops “bamboo hair”, a hair shaft defect where the distal part of the hair shaft is invaginated into its proximal part [4]. NS may be also associated with growth retardation, asthma, food allergies, and elevated serum levels of IgE [1,5]. NS is caused by mutations in SPINK5 gene (serine protease inhibitor Kazal-type 5) that encodes LEKTI (lympho-epithelial Kazal-type related inhibitor), an inhibitor of serine proteases expressed in the epidermis and other stratified epithelia [6]. Full length LEKTI consists of 15 inhibitory domains (D1—D15) and upon synthesis undergoes proteolytic processing into multiple bioactive fragments containing one to six domains with distinct inhibitory specificities [7,8]. LEKTI has been reported to inhibit several proteases including plasmin, trypsin, subtilisin A, cathepsin G, elastase, caspase-14 [9–11] as well as members of the family of kallikrein-related peptidases (KLK), mainly KLK5, KLK7 and KLK14 [12–14]. Unregulated activity of KLK5 and possibly also KLK7 is considered a major source of pathology in NS. Spink5 deficient mice show increased proteolytic activities of KLK5 and KLK7 [15], which corresponds to elevated tryptic and chymotryptic activities described in NS patients [16,17]. KLK5 also initiates a proteolytic cascade by proteolytic activation of KLK7 and KLK14, that leads to degradation of corneodesmosomal proteins desmoglein1 (DSG1), desmocollin1 (DSC1), and corneodesmosin (CDSN) [18]. Premature degradation of corneodesmosomes results in detachment of the stratum corneum (SC) and disruption of the epidermal barrier in NS patients [17]. Upregulated proteolytic activity can further contribute to skin barrier defects by abnormal processing of profilaggrin, a precursor protein which is proteolytically converted into physiologically active filaggrin monomers. Filaggrin is one of key players in maintaining skin hydratation and water retention of the epidermis [19]. Recently, KLK5 was shown to promote profilaggrin processing either via proteolytic activation of elastase 2, which cleaves filaggrin precursor proteins [20] or by direct degradation of profilaggrin [21]. In addition, previous studies using mouse models and human NS patients suggest that the unregulated activity of KLK5 contributes to the inflammatory response in the LEKTI-deficient epidermis by activation of protease-activated receptor 2 (PAR2) [22–24]. In this study, we revealed the in vivo roles of KLK5 and KLK7 using a set of mouse models that are simultaneously deficient for KLK5 and KLK7 on the genetic background of Netherton syndrome-like mouse model based on a mutation found in human patients. The close proximity of these genes (on the same locus) has so far prevented the generation of suitable animal models and therefore the in vivo roles of KLK5 and KLK7 could not be studied concurrently. Our study shows that individual functional ablation of KLK5 or KLK7 is not sufficient to rescue the lethal effect of Spink5 mutation. In contrast, simultaneous deficiency of both KLK5 and KLK7 completely rescues the lethality allowing adult mice to survive to adulthood with a fully functional skin barrier.
To study NS pathology in vivo, we generated a new mouse model mimicking a causative mutation of SPINK5 gene (398delTG; p. A134X) previously described in human patients [25]. Due to the similarity between the human and murine SPINK5 nucleotide sequences, deletion of TG nucleotides at positions 402 and 403 of murine Spink5 (402delTG) produces the premature termination codon (PTC) at a similar position as described in human patients (p. A135X) (Fig 1A). To introduce the mutation into mouse genomic DNA, we prepared TALE nucleases (TALENs) specific for the critical region of Spink5 in combination with a single stranded oligonucleotide (ssODN) carrying the desired mutation (Fig 1B). Founders were screened for targeted incorporation of ssODN by RFLP analysis using XbaI restriction site as a marker (Fig 1C). Heterozygous mice carrying A135X mutation (hereafter referred to as Sp5+/A135X) did not show any obvious phenotype and were used to obtain Sp5A135X/A135X mice, which showed dramatic downregulation of Spink5-RNA expression (Fig 1D). The presence of PTC in the Spink5 transcript was confirmed by sequencing (S1 Fig). Sp5A135X/A135X mice were born in normal Mendelian ratios, however they exhibited severe skin phenotype with exfoliating epidermis, predominantly localized in the abdominal and facial area (Fig 1E) and died within 12 hours after delivery. These phenotypical features mimic NS characteristics and correspond to previously published mouse models of Netherton syndrome [15,26–28]. To elucidate the roles of KLK5 and KLK7 in NS, we generated a set of KLK mutants (Klk5-/-, Klk7-/-, Klk5-/-Klk7-/-), which were crossed with Spink5+/A135X line (Fig 2A). Klk5-/- mice were generated by substitution of exon2 of Klk5 gene with a tm1a-type targeting vector [29] (S2A Fig). As Klk5 and Klk7 are located within close proximity on mouse chromosome 7, generation of Klk5 and Klk7 double-deficient mouse by cross-breeding of individual KO lines is not possible. We therefore applied TALEN-mediated mutagenesis to disrupt the Klk7 gene directly on the genetic background of Klk5-/- mice by introduction of a frame-shift mutation in exon3 of Klk7 gene (S3A Fig). Positively targeted mice were analysed by sequencing of genomic DNA and founder Klk7ex3-A containing 20 bp deletion in Klk7 coding sequence (S3B Fig) was used to establish Klk5-/-Klk7-/- double-deficient line. Targeting of Klk7 was confirmed by cDNA sequencing (S3C Fig) and ablation of KLK7 protein was verified by western blot analysis (S3D Fig). Klk5-/-Klk7-/- mice were further crossed to FLPe expressing mouse line to remove Klk5 KO cassette and re-constitute expression of Klk5 (S2B Fig), thus generating Klk7-/- mutant line (Fig 2A). Klk5-/-, Klk7-/- and Klk5-/-Klk7-/- newborn P0 mice were phenotypically indistinguishable from their control littermates and did not shown any obvious cutaneous phenotype. KLK-deficient mutants were further bred to Sp5+/A135X line in order to generate double Klk5-/-Sp5A135X/A135X, Klk7-/-Sp5A135X/A135X, and triple Klk5-/-Klk7-/-Sp5A135X/A135X mutant mice (Fig 2A). Analysis of mRNA levels confirmed the loss of the targeted gene’s expression (Fig 2B). Similarly to Sp5A135X/A135X mutants, Klk5-/-Sp5A135X/A135X, Klk7-/-Sp5A135X/A135X, and Klk5-/-Klk7-/-Sp5A135X/A135X mice showed no embryonic lethality and the pups exhibited normal Mendelian ratio. Both Sp5A135X/A135X and Klk7-/-Sp5A135X/A135X pups had a fragile epidermis with numerous epidermal lesions located mainly on the abdomen and head (Fig 3A). Cutaneous defects were strongly improved in Klk5-/-Sp5A135X/A135X mice, which exhibited only minor skin lesions, while Klk5-/-Klk7-/-Sp5A135X/A135X mice showed no skin phenotype and the pups were visually indistinguishable from wt (Fig 3A). Sp5A135X/A135X P0 pups also showed underdeveloped or completely absent vibrissae hairs. These hair defects were partially rescued in Klk5-/-Sp5A135X/A135X and to a greater extent in Klk7-/-Sp5A135X/A135X newborn pups, which exhibited slightly shorter and irregularly distributed vibrissae hairs. Klk5-/-Klk7-/-Sp5A135X/A135X did not show any major abnormalities of whiskers (S4 Fig). Inactivation of KLK7 did not affect the survival of Sp5A135X/A135X mice as Klk7-/-Sp5A135X/A135X pups died within 12 hours after birth. Interestingly, Klk5-/-Sp5A135X/A135X mice survived until postnatal day 5 (P5) when they exhibited reduced body-size and dry skin with severe scaling throughout the body surface (Fig 3B). In contrast, simultaneous inactivation of both KLK5 and KLK7 fully rescued the lethality and Klk5-/-Klk7-/-Sp5A135X/A135X survive to adulthood. At P5, the skin of Klk5-/-Klk7-/-Sp5A135X/A135X appears to be more stretched and shiny in comparison to wt mice, however they show no signs of scaling (Fig 3B). Nevertheless, Klk5-/-Klk7-/-Sp5A135X/A135X showed alopecia and growth retardation 2–4 weeks after birth (Fig 3C), which disappeared with age. Interestingly, scanning electron microscopy (SEM) analysis of vibrissae and pelage hairs revealed that Klk5-/-Klk7-/-Sp5A135X/A135X pups from P4 –P28 develop a specific hair shaft defect that strongly resembles bamboo hair in NS patients (Fig 3E). The hair defects and growth retardation in Klk5-/-Klk7-/-Sp5A135X/A135X improve with age (Fig 3D and 3F) and no major cutaneous phenotype was seen in adulthood, with the exception of minor scaling on the ears, shorter tail, and a lower body weight (Fig 3D and 3F). Analysis of newborn P0 mice revealed that Sp5A135X/A135X and Klk7-/-Sp5A135X/A135X have significantly decreased weight in comparison to other mutant and wt lines (S5A Fig) while no differences where observed in E18. 5 dpc embryos (S5B Fig). This suggests that the weight loss in Sp5A135X/A135X and Klk7-/-Sp5A135X/A135X lines is caused by severe epidermal barrier disruption followed by rapid dehydration. The integrity of epidermal barrier in Sp5A135X/A135Xnewborn P0 pups was analysed using the toluidine blue (TB) penetration assay and showed severe skin barrier disruption marked by penetration of TB through large areas of the body, mainly the abdomen, paws, and head. Ablation of KLK7 on the Sp5A135X/A135X background did not improve the barrier and Klk7-/-Sp5A135X/A135X newborns showed similar barrier disruption to Sp5A135X/A135X. In contrast, Klk5-/-Sp5A135X/A135X newborns developed less severe barrier phenotype characterised by multiple small stained patches and in Klk5-/-Klk7-/-Sp5A135X/A135X pups, the barrier integrity was almost completely recovered and the mice showed TB staining only in the area of nostrils (Fig 4A). To confirm that disruption of skin barrier leads to the dehydration of newborn mice, we assessed the trans-epidermal water loss (TEWL) in P0 pups over time. Consistently with previous analyses of epidermal barrier properties, we found significantly impaired water retention in Sp5A135X/A135X and Klk7-/-Sp5A135X/A135X. Water barrier was partially rescued in Klk5-/-Sp5A135X/A135X pups and completely restored in Klk5-/-Klk7-/-Sp5A135X/A135X (Fig 4B). To understand why Klk5-/-Sp5A135X/A135X died at P5, mice were stained with TB at P5. Interestingly, Klk5-/-Sp5A135X/A135X mice showed clear penetration of the dye in the epidermis adjacent to hair shafts whereas this barrier defect was completely rescued in Klk5-/-Klk7-/-Sp5A135X/A135X (Fig 4C). Detailed analysis of the epidermis using SEM revealed that P5 Klk5-/-Sp5A135X/A135X mice had dramatic epidermal defects manifested by defective separation of hair shafts from the surrounding tissues and subsequent loss of infundibular epidermis in the upper part of hair follicles. These defects were not observed in Klk5-/-Klk7-/-Sp5A135X/A135X (Fig 4D) thus suggesting that KLK7 activity is responsible for the epidermal barrier defects contributing to the lethality of Klk5-/-Sp5A135X/A135X at P5. No barrier abnormalities were observed in P0 or P5 pups from control lines Klk5-/-, Klk7-/- and Klk5-/-Klk7-/-. Histological analysis of the epidermis from NS patients together with previously published data on LEKTI-deficient models describe an abnormally differentiated epidermis [15]. In line with these observations, analysis of non-lesional skin of P0 pups showed a reduced granular layer, acanthosis and sporadic SC detachment and parakeratosis in Sp5A135X/A135X pups (Fig 5A and S6 Fig). Although Klk7-/-Sp5A135X/A135X pups showed a similar phenotype to Sp5A135X/A135X mice, no differentiation defects were observed in the epidermis of newborn pups apart from the occasional focal detachment of SC (Fig 5A and S6B Fig). Klk5-/-Sp5A135X/A135X and Klk5-/-Klk7-/-Sp5A135X/A135X newborn pups exhibited well differentiated epidermal layers. To address the characteristics of epidermal differentiation defects, we analysed the expression of several differentiation markers in E18. 5 dpc embryos to avoid the contribution of secondary effects following barrier disruption and exposure to the environment (Fig 5B). In accordance with previous results, Sp5A135X/A135X embryos exhibited a poorly defined basal layer, abnormal expression of keratin 14 (Krt14) and markedly increased expression of keratin 6 (Krt6), suggesting hyperproliferation of keratinocytes in Sp5A135X/A135X embryos (Fig 5B). Krt14 and Krt6 expression in Klk5-/-Sp5A135X/A135X, Klk7-/-Sp5A135X/A135X, and Klk5-/-Klk7-/-Sp5A135X/A135X embryos exhibited a similar pattern to wt animals, indicating that the differentiation defects of LEKTI-deficient epidermis are fully dependent on concurrent activities of both KLK5 and KLK7. In the epidermis of both Sp5A135X/A135X and Klk7-/-Sp5A135X/A135X embryos profilaggrin granules were absent whereas they were present in Klk5-/-Sp5A135X/A135X, Klk5-/-Klk7-/-Sp5A135X/A135X and in wt mice (Fig 5B). As Klk5-/-Sp5A135X/A135X exhibited drastic epidermal defects before they die at P5 (Fig 4C and 4D), we performed histological analysis of skin from Klk5-/-Sp5A135X/A135X, Klk5-/-Klk7-/-Sp5A135X/A135X and wt mice at P5. We found that although Klk5-/-Sp5A135X/A135X pups showed normal differentiation at P0, over time they developed hyperplastic epidermis with acanthosis, severe intrafollicular hyperkeratosis and the skin was infiltrated by mast cells (Fig 6A and S7 Fig). In contrast, Klk5-/-Klk7-/-Sp5A135X/A135X showed no such defects and the epidermis was comparable to wt (Fig 5C). We also observed an increased expression of keratin6 in the epidermis of Klk5-/-Sp5A135X/A135X P5 pups, which was rescued in Klk5-/-Klk7-/-Sp5A135X/A135X (Fig 6B). Analysis of corneodesmosomal proteins revealed markedly decreased expression of CDSN at the stratum corneum/stratum granulosum interface in Klk5-/-Sp5A135X/A135X, but not in Klk5-/-Klk7-/-Sp5A135X/A135X pups (S8 Fig). These data suggest that abnormal epidermal differentiation of Klk5-/-Sp5A135X/A135X P5 pups is caused by KLK7 activity. As the aggravated inflammatory response and allergic manifestations are symptomatic for NS, we assayed the expression of pro-inflammatory and pro-TH2 cytokines in the skin isolated from Spink5- and Klk-deficient E18. 5 dpc embryos. As expected, Sp5A135X/A135X embryos showed elevated expression levels of TNFα, TSLP, Il-33, Il-1β as well as ICAM1 (Fig 7A). In contrast, these cytokines were not upregulated in Klk5-/-Sp5A135X/A135X animals, which is in line with the recent study of Furio et al. [24] indicating that KLK5 is responsible for triggering the inflammation in LEKTI-deficient epidermis. However, expression levels of TNFα, TSLP, Il-33, Il-1β and ICAM1 were also completely normal in mice with ablated KLK7, i. e. in Klk7-/-Sp5A135X/A135X (Fig 7A). Full rescue of cutaneous inflammation was also observed in Klk5-/-Klk7-/-Sp5A135X/A135X embryos (Fig 7A). To analyse cutaneous inflammation in later stages of development, we assayed expression levels of TNFα, TSLP, Il-33, Il-1β and ICAM1 in the skin of P5 pups from surviving LEKTI-deficient mutant lines Klk5-/-Sp5A135X/A135X and Klk5-/-Klk7-/-Sp5A135X/A135X and wt controls. Although no signs of cutaneous inflammation were found in Klk5-/-Sp5A135X/A135X E18. 5 dpc embryos, P5 pups showed significant upregulation of TNFα, TSLP, Il-33, Il-1β and ICAM1 (Fig 7B). Expression of these cytokines was normalized by inactivation of KLK7, in Klk5-/-Klk7-/-Sp5A135X/A135X P5 pups (Fig 7B). NS is also associated with systemic inflammation, allergy and elevation of IgE levels. We examined the serum levels of TNFα, Il-1β, IL-9 and IL-17 in Klk5-/-Klk7-/-Sp5A135X/A135X 6 weeks old mice, however no signs of systemic inflammation were found when compared to wt mice (Fig 7C). These observations are in line with the rescue of other NS-like symptoms in adult Klk5-/-Klk7-/-Sp5A135X/A135X mice.
Netherton syndrome is a severe genetic disorder associated with unregulated proteolytic activity, caused by the absence of functional LEKTI, a protease inhibitor encoded by SPINK5 gene. In this study, we elucidated the roles of KLK5 and KLK7 proteases in the disease by genetic inactivation of these proteases on the background of a mouse model for NS. This novel model was generated by mimicking a SPINK5 p. A134X mutation found in human patients [25] and recapitulates the phenotype of previously described Spink5-deficient mouse models [15,26,27]. However, our Sp5A135X/A135X mice in combination with the ablation of KLK5 and KLK7 reveal the complexity of the LEKTI-KLK network. We showed, that although single inactivation of KLK5 or KLK7 rescues a number of NS-like pathological manifestations, only simultaneous ablation of both proteases fully rescues the lethal phenotype of Sp5A135X/A135X mice. It has been proposed that the barrier defects observed in LEKTI-deficient skin are caused by proteolytic hyperactivity leading to premature degradation of corneodesmosomal proteins. In vitro assays showed that three putative LEKTI targets are able to promote corneodesmosome degradation namely KLK5, KLK7, and KLK14 [18]. The in vitro study of the KLK proteolytic activation cascade proposed that KLK5 acts upstream from KLK7 and KLK14 and therefore, KLK5 hyperactivity should contribute to barrier defects either directly or indirectly via activation of the remaining KLKs. Indeed, significant improvement of skin-barrier defects by inactivation of KLK5 in Sp5A135X/A135X mice was observed, however the rescue was incomplete as toluidine blue staining in Klk5-/-Sp5A135X/A135X mice revealed patches of disrupted-barrier distributed all over the body surface. This observation implicates the role of another protease whose activity contributes to barrier defects in the absence of LEKTI and does not depend on KLK5. This was identified as KLK7, since Klk5-/-Klk7-/-Sp5A135X/A135X newborn mice did not show any major barrier defects of epidermis. Interestingly, single inactivation of KLK7 on Sp5A135X/A135X background did not significantly improve the barrier defects. Therefore we assume that barrier properties of LEKTI-deficient neonatal epidermis are compromised mainly by direct activity of KLK5 and only to a lesser extent by KLK5-mediated activation of KLK7. The significant contribution of KLK5 to NS pathology is in line with a recent study of Furio et al. showing amelioration of skin barrier-phenotype in Spink5-deficent newborns upon KLK5 inactivation [24]. Nevertheless, the remaining activity of KLK7 still contributes to the defective barrier and further intensifies with age as Klk5-/-Sp5A135X/A135X show severe epidermal defects manifested by loss of infundibular epidermis at P5. Klk5-/-Klk7-/-Sp5A135X/A135X mice exhibit no skin-barrier defects at P5 and most importantly, in contrast to Klk5-/-Sp5A135X/A135X mice, the triple mutants survive to adulthood. The skin defects in Klk5-/-Sp5A135X/A135X P5 pups markedly resemble those observed in mice deficient for corneodesmosomal proteins CDSN and DSC1 [30,31]. Indeed, Klk5-/-Sp5A135X/A135X show reduced CDSN expression at P5, which indicates that unregulated activity of KLK7 results in degradation of corneodesmosomes. The mechanism of pro-KLK7 activation in the absence of KLK5 remains unclear. KLK7 can be activated by matriptase [32] and a recent study also suggests a role of mesotrypsin in pro-KLK7 activation [33]. Although the lethal phenotype is fully rescued in Klk5-/-Klk7-/-Sp5A135X/A135X mutants and the mice do not show any signs of skin barrier-defects leading to dehydration, we observed minor barrier disruptions in the nostril area of newborn pups, which suggests the activity of another protease physiologically inhibited by LEKTI. As the toluidine blue -stained area overlaps with the expression of KLK14 in late embryonic development (S9A Fig), we propose that KLK14 could be responsible for the remaining pathology of LEKTI-deficient mice even in the absence of KLK5 and KLK7. Moreover, we and others also observed expression of KLK14 in hair follicles (S9B Fig) [34], which makes KLK14 a candidate protease responsible for the development of the bamboo hair defect in Klk5-/-Klk7-/-Sp5A135X/A135X animals up to the age of 3 weeks. As reported, the defects of cell adhesion proteins in hair follicles result in “lanceolate hair”–a hair shaft phenotype in mice that strongly resembles the bamboo hairs of Klk5-/-Klk7-/-Sp5A135X/A135X mutants and NS patients [35,36]. This further supports a possible role of KLK14 in the formation of bamboo-hair, as KLK14 is linked to the degradation of desmosomal proteins in LEKTI-deficient epidermis [8]. Nevertheless, any targets of LEKTI inhibition present in hair follicles, such as caspase-14 [37] or other, currently unidentified proteases, should be considered as a potential cause of bamboo hairs. Association of NS with abnormal epidermal differentiation accompanied by acanthosis, parakeratosis, and hyperproliferation of keratinocytes was previously reported in Spink5-deficient mouse models [15,26,27] and our Sp5A135X/A135X confirms the previous findings. We found clear overexpression of keratin6 in Sp5A135X/A135X E18. 5 dpc epidermis, suggesting that events leading to hyperproliferation of keratinocytes are triggered prior to the exposure to the external environment and are a result of unregulated proteolytic activity in the epidermis. In light of the fact that single inactivation of either KLK5 or KLK7 completely rescues the differentiation defects in LEKTI-deficient embryos as well as in newborn mice, we believe that the signalling events resulting in keratinocyte hyperproliferation in neonates depend on the presence of both, KLK5 and KLK7 together. Moreover, we observed that aggravated cutaneous inflammation, which is found in E18. 5 Sp5A135X/A135X embryos, fully depends on simultaneous activity of both, KLK5 and KLK7. KLK5 was previously shown to initiate inflammation in LEKTI-deficient epidermis by activation of PAR2, which results in the induction of pro-Th2 and pro-inflammatory cytokines [22–24]. In this study we show that KLK7 is also required for the induction of inflammation in LEKTI-deficient mice as P5 Klk5-/-Sp5A135X/A135X pups developed severe acanthosis together with significantly increased expression of TNFα, TSLP, Il-33, Il-1β and ICAM1 while Klk5-/-Klk7-/-Sp5A135X/A135X showed no major defects in the epidermis and increased levels of pro-inflammatory cytokines. Altogether, this suggests that inflammation and differentiation changes in older LEKTI-deficient pups (P5) are initiated by KLK7 activity which is independent of KLK5. Indeed, KLK7 was previously shown to induce inflammation and keratinocyte proliferation in the epidermis [38,39] and a recent study identified KLK7 as a proliferative factor in a mouse model of colon cancer and in human cells in vitro [40]. The mechanism by which KLK7 induces inflammation and differentiation changes remains to be elucidated. In contrast to KLK5, KLK7 cannot directly activate PAR2 as shown in vitro [41] and thus, the inflammation is likely to be triggered by a different mechanism. One possible pathway is the KLK7-mediated conversion of pro-IL1β to active IL1β [42], which could affect the inflammatory phenotype of NS- epidermis. In summary, we show that the individual inactivation of KLK5 or KLK7 only partially rescues the defective skin barrier but not the lethal phenotype of Sp5A135X/A135X. Only the concurrent ablation of both KLK5 and KLK7 can fully rescue the lethal phenotype of Sp5A135X/A135X mice, therefore both proteases should be investigated as clinical targets. We show that KLK7 plays an important role in the inflammation and defective differentiation in NS and its activity is not dependent on activation by KLK5. We also show that the pathological effects of unregulated KLK activities are remarkably age dependent. Altogether, this study expounds the complexity of the proteolytic network and its regulation, which are especially important to understand Netherton syndrome and its treatment.
All animal studies were ethically reviewed and performed in accordance with European directive 2010/63/EU and were approved by the Czech Central Commission for Animal Welfare. Knock-out first allele of Klk5 was produced by introduction of targeting construct (vector PRPGS00082_A_A10 obtained from NIH Knock-out Mouse Program, KOMP) via homologous recombination in embryonic stem cells (ESC). Positively targeted ESC were injected into developing wt embryos, to produce chimeric mice, which were used to establish Klk5-/- line. TALENs targeting exon5 of Spink5 gene were designed using TAL Effector Nucleotide Targeter 2. 0 (https: //tale-nt. cac. cornell. edu/) [43,44], assembled using the Golden Gate Cloning system[43], and cloned into the ELD-KKR backbone plasmid as described previously [45]. DNA binding domains of TALENs specific for the desired target site within Spink5 gene (Fig 1B) consisted of following repeats: NI-HD-HD-NN-HD-NI-NN-NG-NI-NN-NI-NG-NN-NG-NN-NI-NI-HD-NG (5´ TALEN-Spink5) and HD-NG-NN-HD-NG-NG-NG-NI-NG-NI-NN-NN-NN-NG-NI-HD-NG-HD-NI-HD (3´ TALEN-Spink5). Both TALEN plasmids were used for production of TALEN encoding mRNA as described previously [46]. 5 μl of TALEN mRNA (with total RNA concentration of 40 ng/μl) was mixed with 100 μM of targeting single-stranded oligonucleotide (Sigma-Aldrich; Fig 1B) and the final solution was microinjected into C57BL6/N-derived zygotes. Genomic DNA isolated from tail biopsies of newborn mice were screened by PCR (primers F1: 5´-CCTGTCTCTGCCTTCAGACC-3´ and R1: 5´-GGCTGTGGTAACTGTCCAAAA-3´) and subsequent RFLP analysis using XbaI restriction enzyme (Thermo-Scientific). TALENs were designed and synthesized as described above. DNA binding domains of TALENs specific for exon3 of murine Klk7 contained following repeats: NN-NG-NI-NI-NI-NN-NI-NI-NN-NN-HD-NG-HD-NN-HD (5´ TALEN-Klk7) and NN-NI-NG-NG-NN-HD-HD-NG-NG-NG-NN-NI-NN-HD-NI-NN (3´ TALEN-Klk7). TALEN mRNA with total RNA concentration of 40 ng/μl was microinjected into Klk5-/-oocytes. Genomic DNA isolated from tail biopsies of newborn mice was screened by PCR (primers F3: 5´- GGAGAAGGCCAGGGTCTGAA-3´ and R3: 5´- TGGTCAGAAACCCACGGAGA-3´) and subsequently analyzed by RFLP using NcoI restriction enzyme (Thermo-Scientific). Newborn pups from at least two independent litters were separated from mothers to prevent fluid intake. The rate of water loss was analyzed by measuring the reduction of initial body weight at 1h, 2h, 3h and 4 h. Newborn mice were euthanized and then dehydrated by incubation for 5 min in 25,50,75, and 100% methanol. After rehydration in PBS, mice were incubated for 4 hours in 0. 1% toluidine blue (Sigma-Aldrich), washed in PBS and imaged. Newborn pups or skin tissues were fixed in 3. 6% formaldehyde for 24 h and embedded in paraffin. 5-μm sections were prepared using microtome were stained by hematoxylin/eosin (H&E) or by 0. 5% toluidine blue using standard protocols. Images were obtained using Zeiss Axioscan Z1 (Carl Zeiss AG). Dorsal skin of P5 pups was fixed in 3. 6% formaldehyde for 24 hours and embedded in paraffin. 5-μm paraffin sections were used for antigen retrieval with Discovery Ultra automated IHC/ISH system (Ventana) and stained with antibodies against Dsg-1 (Santa Cruz, 1: 100 dilution, retrieval at pH6) and CDSN (Abcam, 1: 100 dilution, retrieval at pH6). After 1 hour incubation at room temperature, anti-rabbit peroxidase conjugated polymer (Zytomed GmBH) was applied for 30 min and the reaction was developed using DAB (DAKO) as a chromogen. Images were obtained using Zeiss Axioscan Z1 (Carl Zeiss AG). In order to stain the cryosections, dorsal skin of E18. 5 dpc embryos or P5 pups was isolated, embedded in Tissue-Tek O. C. T (Sakura), and frozen at -80°C. 6 μm-sections were stained as described previously [47], using antibodies against Keratin6 (Covance, 1: 1000 dilution), Keratin14 (Covance, 1: 2000 dilution) and Filaggrin (Covance, 1: 1000 dilution). Nuclei were stained using DAPI (Roche). Images were obtained using Zeiss Axioimager Z2 (Carl Zeiss AG). Dorsal skin was obtained from newborn pups, crushed in liquid nitrogen and total RNA was isolated using TRIzol (Thermo-Scientific) according to the manufacturer' s instructions. Residuals of genomic DNA were removed using 1 U of DNAse I (Roche) per 1 μg of RNA by 15 min incubation at 37°C. 1 μg of total RNA was used for reverse transcription by M-MLV Reverse Transcriptase (Promega) using oligo (dT) primers. RT-PCR was performed in a 20-μl reaction mixture containing SYBR Green JumpStart™ Taq ReadyMix with MgCl2 (Sigma-Aldrich) and 0. 25mM of each primer. Respective gene expression was normalized to the expression of TATA-binding protein (TBP). Normalized expression levels were then re-expressed relative to the mean expression level of the respective target in the wt mice. Primer sequences are detailed in S1 Table. The samples on cellulose filter paper strips were fixed with 3% glutaraldehyde in cacodylate buffer overnight at 4°C. After fixation, extensively washed samples were dehydrated through ascending alcohol concentrations followed by absolute acetone and critical point drying from liquid CO2 in a K 850 unit (Quorum Technologies Ltd). The dried samples were sputter-coated with 20 nm of gold in a Polaron Sputter-Coater (E5100) (Quorum Technologies Ltd). The final samples were examined in a FEI Nova NanoSem 450 scanning electron microscope (FEI) at 5 kV using secondary electron detector. The levels of TNFα, IL-1β, IL-9, and IL-17 in mouse serum were analyzed using Bio-Plex Pro Mouse Cytokine Assay (Bio-Rad Laboratories) with high sensitivity range standard settings according to manufacturer’s instructions. | Netherton syndrome (NS) is a genetic skin disorder caused by the loss of protease inhibitor LEKTI, which leads to the dysregulation of epidermal proteases and severe skin-barrier defects. In this work, we aimed to explore the molecular mechanisms underlying this disease using a novel mutant mouse model for NS, which is based on mimicking a causative mutation known from human patients. This novel model reproduces the symptoms of NS and thus provides a useful tool to study the NS pathology in a complex in vivo environment. Most importantly, by combination of this NS-mouse model with mutant mice individually or simultaneously deficient for proteases KLK5 and KLK7, we elucidated the complex proteolytic networks that are dysregulated in the absence of LEKTI. We show that although the single ablation of KLK5 or KLK7 is not sufficient to rescue the lethal effect of LEKTI-deficiency, simultaneous deficiency of both KLKs completely rescues the epidermal barrier and the postnatal lethality. Our results also provide novel insights into the roles of KLK5 and KLK7 in the inflammation and differentiation defects that are associated with NS. Based on these findings, we propose that both, KLK5 and KLK7 should become targets for NS therapy. | Abstract
Introduction
Results
Discussion
Materials and Methods | skin
medicine and health sciences
integumentary system
pathology and laboratory medicine
enzymes
immunology
enzymology
animal models
developmental biology
model organisms
signs and symptoms
experimental organism systems
epidermis
embryos
research and analysis methods
embryology
hair
inflammation
proteins
mouse models
immune response
biochemistry
diagnostic medicine
anatomy
phenotypes
genetics
biology and life sciences
proteases
neonates | 2017 | KLK5 and KLK7 Ablation Fully Rescues Lethality of Netherton Syndrome-Like Phenotype | 9,981 | 303 |
Differentiation of CD4+ T cells into effector or regulatory phenotypes is tightly controlled by the cytokine milieu, complex intracellular signaling networks and numerous transcriptional regulators. We combined experimental approaches and computational modeling to investigate the mechanisms controlling differentiation and plasticity of CD4+ T cells in the gut of mice. Our computational model encompasses the major intracellular pathways involved in CD4+ T cell differentiation into T helper 1 (Th1), Th2, Th17 and induced regulatory T cells (iTreg). Our modeling efforts predicted a critical role for peroxisome proliferator-activated receptor gamma (PPARγ) in modulating plasticity between Th17 and iTreg cells. PPARγ regulates differentiation, activation and cytokine production, thereby controlling the induction of effector and regulatory responses, and is a promising therapeutic target for dysregulated immune responses and inflammation. Our modeling efforts predict that following PPARγ activation, Th17 cells undergo phenotype switch and become iTreg cells. This prediction was validated by results of adoptive transfer studies showing an increase of colonic iTreg and a decrease of Th17 cells in the gut mucosa of mice with colitis following pharmacological activation of PPARγ. Deletion of PPARγ in CD4+ T cells impaired mucosal iTreg and enhanced colitogenic Th17 responses in mice with CD4+ T cell-induced colitis. Thus, for the first time we provide novel molecular evidence in vivo demonstrating that PPARγ in addition to regulating CD4+ T cell differentiation also plays a major role controlling Th17 and iTreg plasticity in the gut mucosa.
The CD4+ T cell differentiation process activates the transcriptional and secretory cellular machinery that helps orchestrate immune modulation in infectious, allergic and immune-mediated diseases. Upon antigen presentation, naïve CD4+ T cells become activated and undergo a differentiation process controlled by the cytokine milieu in the tissue environment. For instance, interleukin-6 (IL-6) in combination with transforming growing factor β (TGF-β) trigger a naive CD4+ T cell to become a T helper 17 (Th17) cell [1], [2]. In contrast, TGF-β alone can activate regulatory pathways leading to differentiation of naive CD4+ T cells into an induced regulatory CD4+ T cell (iTreg) phenotype, which in turn tightly dampens effector and inflammatory responses. CD4+ T cell differentiation was once viewed as a rigid process whereby a naive cell differentiated into terminal phenotypes. However, mounting evidence supports the tissue environment-dependent plasticity of CD4+ T cell subsets and suggests the emergence of new phenotypes [3]–[5]. At the molecular level, plasticity is achieved by a cytokine-driven reprogramming of signaling pathways and targeted activation of master regulator transcription factors which results in gene expression changes [6]. Antigen presenting cells (APCs) influence T cell differentiation through antigen presentation, co-stimulation and cytokine secretion [7]. The crosstalk between T cell phenotypes has been fully characterized in terms of classical Th1 versus Th2 differentiation [8]–[11]. Indeed, a logical network model of CD4+ T cell differentiation process centered around Th1 versus Th2 differentiation was published by Mendoza [12]. However, this logical model did not consider the Th17 or iTreg cell subsets. In the last decade, Th17 has emerged as an extremely plastic phenotype [6], [13]–[15] that can acquire regulatory functions following changes in the local cytokine milieu [16]–[19]. Furthermore, human iTreg cells become interleukin-17 (IL-17) -producing Th17 cells [20], thereby supporting the concept that Th17 plasticity is a two-way process. However, the molecular mechanisms underlying these processes are incompletely understood. Retinoic acid receptor-related orphan receptor gamma (RORγt) is a master regulator transcription factor required for Th17 differentiation [21], [22] and it has been proposed as a potential therapeutic target to suppress Th17 responses in autoimmune diseases [23], [24]. Similar to RORγt, the peroxisome proliferator-activated receptors (PPARs) are ligand-activated transcription factors and members of the nuclear receptor superfamily. PPARγ is highly expressed in CD4+ T cells and it has been reported to modulate Th1 and natural Treg (nTreg) function [25]–[27], but limited information is available regarding its role in modulating the Th17 and iTreg phenotypes. The loss of PPARγ in CD4+ T cells enhanced antigen-specific proliferation and overproduction of interferon γ (IFN-γ) in response to IL-12 [28]. In addition, the deficiency of PPARγ in nTreg cells impairs their ability to prevent effector T cell-induced colitis following transfer of naïve CD4+ T cells into SCID recipients [28]. Furthermore, pharmacologic activation of PPARγ prevents removal of the silencing mediator for retinoid and thyroid hormone receptors' co-repressor from the RORγt promoter in T cells, thus interfering with RORγt transcription [29]. While previous studies shed some light on the role of PPARγ in Th17 differentiation, this is the first study to investigate the role of PPARγ in controlling Th17 to iTreg cell plasticity in the gut mucosa. Computational approaches have become a powerful tool that allows concurrent multiparametric analysis of dynamic biological processes and diseases. The emerging use of systems modeling in combination with experimental immunology studies in vivo can help integrate existing knowledge and provide novel insights on rising trends and behaviors in biological processes such as CD4+ T cell differentiation and function. Of note, bioengineering studies demonstrated the predictive value of a whole-cell computational model of the life cycle of Mycoplasma genitalium [30]. These multi-mode calibrated models demonstrate an emerging strategy to answer questions about fundamental cell-based processes in silico and help focus experimental designs of animal pre-clinical and human clinical studies. We combined computational modeling and mouse adoptive transfer studies to gain a better mechanistic understanding of the modulation of CD4+ T cell differentiation and plasticity at the intestinal mucosa of mice. Our sensitivity analyses highlighted the importance of PPARγ in the regulation of Th17 to iTreg plasticity. Indeed, in vivo evidence demonstrates that PPARγ is required for the plasticity of Th17 promoting a functional shift towards an iTreg phenotype. More specifically, PPARγ activation is associated with upregulation of FOXP3 and suppression IL-17A and RORγt expression in colonic lamina propria CD4+ T cells. Conversely, the loss of PPARγ in T cells results in colonic immunopathology driven by Th17 cells in adoptive transfer studies.
Cytokines are small molecules secreted in response to external stimuli, which are key in cell-to-cell communication. Cytokine signaling is fast and canonical, consisting of 1) binding to cytokine cell surface receptor, 2) activation of receptor-associated kinase, 3) STAT phosphorylation and translocation into the nucleus and 4) activation of gene expression. In naïve CD4+ T cells cytokine signaling leads to the expression of transcription factors that upregulate gene subsets that shape cell phenotype and function. As an output of this process, differentiated cells preferentially secrete phenotype–associated cytokines, such as IL-17 produced by Th17 cells or IFNy produced by Th1 (Figure S1). To facilitate a comprehensive representation of the dynamics associated with the major pathways activated by cytokines which control CD4+ T cell differentiation and plasticity, we constructed an ordinary differential equation (ODE) -based computational model including cytokines, membrane receptors and transcription factors (Figure 1). Knowledge discovery involved an iterative process that fully integrated computational modeling and in vivo experimentation in mice (Figure S2). The CD4+ T cell differentiation model consists of 60 ODEs, 52 reactions and 93 species (Figure S3). The mathematical model was engineered to ensure proper modulation of intracellular pathways and cell phenotypes via external cytokines representing the cytokine milieu. The Hill Function and mass action equations were used [31]. While the Hill Coefficient allowed us to quantify the effect of a ligand binding a macromolecule through cooperative binding, mass action laws can represent dynamic equilibriums for elementary reactions, considering products as a proportion of the participating molecules in the reaction. Experimental data (Table S1) was used to calibrate and adjust model parameters to ensure correct dynamics (Table S2 and S3, Figure S4). A list of modeling assumptions can be found in Table S4. Among the four possible phenotypes in this mathematical model, to induce Th17 differentiation from a naïve CD4+ T cell, external IL-6 and external TGF-β were added in combination and demonstrated upregulation of RORγt, IL-17 and STAT-3 (Figure S5) as followed by our table of initialization fates (Table S5). Sensitivity analyses identified PPARγ as an essential regulator of CD4+ T cell differentiation and plasticity (Figure S6). Based on the results of the sensitivity analysis we performed computer simulations aimed to further characterize the role of PPARγ on Th cell differentiation in silico. Following induction of the computational model towards a Th17 phenotype by adding external TGF-β and external IL-6 in silico, modeling efforts predicted that increasing concentrations of PPARγ in Th17 cells led to downregulation of RORγt and IL-17 and upregulation of FOXP3 (Figure 2A), thus, displaying a phenotype switch from Th17 to iTreg. A list of computational modeling derived predictions can be found in Table S6. To validate the results of our computational simulations, we first isolated and sorted naïve CD4+ T cells from spleens of wild-type and T cell-specific PPARγ null mice. Deletion of PPARγ via a transgenic expression of Cre under control of the CD4 promoter (PPARγfl/fl; CD4-Cre+) allowed us to use loss-of-function approaches to characterize the role of PPARγ in Th17 differentiation. Cells were polarized towards a Th17 phenotype with recombinant mouse IL-6 and TGF-β. IFNγ and IL-4 were eliminated to block Th1 and Th2 differentiation respectively with neutralizing antibodies. After 60 hours of culture, cells were treated with increasing amounts of pioglitazone (PIO), a synthetic PPARγ agonist of the thiazolidinedione (TZD) class of anti-diabetic drugs. Before starting pioglitazone treatment, at t = 60 h, IL-17 and RORγt expression were significantly upregulated in PPARγ null when compared to wild-type cells (Figure 2B). Following pioglitazone treatment for 24 h. , Th17 cells from wild-type mice showed increasing levels of FOXP3 and downregulation of RORγt and IL-17A with increased concentration of the exogenous PPARγ agonist in wild-type (Figure 2C), but this effect was not observed in PPARγ null Th17 cells (Figure 2D), suggesting the role of PPARγ in the modulation of these molecules. The same study was repeated three times with very similar trends on these behaviors (Figure S7). These results provide in vitro evidence that PPARγ significantly dampens Th17 differentiation and slightly enhances FOXP3 expression. Interestingly, uncoupling between suppressed Th17 responses and enhanced iTreg cells suggests that a T cell-extrinsic mechanism (i. e. , APC-derived signals) might be contributing to this Th17 plasticity in vivo. To determine whether the loss of T cell PPARγ favors Th17 and impairs iTreg cell differentiation and also to assess whether T cell-extrinsic mechanisms might be affecting iTreg upregulation we conducted computational simulations and in vivo studies of PPARγ deletion in T cells. Chronologically, a PPARγ-deficient naïve CD4+ T cell was created in silico by blocking PPARγ downstream signaling. The loss of PPARγ in silico caused upregulation of RORγt and IL-17 in Th17 cells (Figure 3B) and down-regulation of FOXP3 in iTreg cells (Figure 3D) compared to wild-type CD4+ T cells (Figure 3A and 3C). These results demonstrate that PPARγ exerts a regulatory role in CD4+ T cell differentiation from a naïve state to Th17 or iTreg cells. Next, to validate this computational prediction, we sorted CD4+CD25-CD45RBhigh naïve T cells from spleens of donor wild-type and T cell-specific PPARγ null mice and adoptively transferred 4×105 viable cells to SCID recipients (Figure S8). Cells isolated from the colonic lamina propria (LP), spleen and mesenteric lymph nodes (MLN) of recipient mice were assayed for expression of FOXP3, RORγt and IL-17A by intracellular flow cytometry. The transfer of CD4+ T cells lacking PPARγ resulted in significantly greater accumulation of IL-17-producing Th17 cells and lower levels of FOXP3+ iTreg cells in spleen, MLN and colonic LP of recipient mice (Figure 3E and 3F and Figure S9). Recipients of PPARγ null cells showed a significantly more severe and earlier onset of disease when compared to recipients of wild-type cells (Figure 4A). Histological examination demonstrated that colons recovered from recipients of PPARγ null CD4+ T cells had significantly greater lymphocytic infiltration and crypt hyperplasia than those recovered from recipients of wild-type CD4+ T cells (Figure 4B). To determine whether PPARγ activation played an essential role in converting fully differentiated Th17 cells into iTreg cells, the computational model was induced to Th17 with the addition of IL-6 and TGFβ and PPARγ was activated when the cell was a fully differentiated Th17. Results show that following induction of Th17 and subsequent PPARγ activation, IL-17, STAT-3 and RORγt were dramatically downregulated, whereas FOXP3 was upregulated, thereby demonstrating a phenotypic switch from a Th17 to an iTreg phenotype (Figure 5A). To ensure that parameter space scan and time-course were linked and the changes in PPARγ were being observed in a time-dependent manner, a combination of both was run, reiterating the phenotype switch with increasing concentrations of PPARγ over time observing an upregulation of FOXP3 and a downregulation of IL-17, RORγt and STAT3-P (Figure 5B). To address this hypothesis, we sorted CD4+ CD25- CD45RBhigh naïve T cells from spleens of donor wild-type mice and transferred 4×105 viable cells to RAG2−/− recipients. When clinical signs of disease and colitis appeared, a subset of mice was sacrificed and spleen, MLN and colons were extracted to examine Th17 and Treg levels (baseline results). After verifying the presence of Th17 cells in colon, MLN and spleen, half of the mice were received a daily treatment of 70 mg/kg of pioglitazone given orally to activate PPARγ (Figure 5C). During the treatment period, mice treated with pioglitazone recovered weight and their disease activity scores dropped significantly (Figure S10) compared to mice treated with PBS (Figure S11). Histopathological examinations also showed that colons from recipient mice treated with pioglitazone had a significantly lower lymphocytic infiltration and crypt hyperplasia than those from non-treated recipients (Figure S12). Untreated mice maintained a predominant Th17 response characterized by increased levels of CD4+ T cells expressing RORγt and IL-17A. In contrast, pioglitazone-treated mice not only recovered from colitis and its associated weight loss, but also showed a switch from a predominant Th17 into an iTreg phenotype characterized by increased expression of FOXP3 and decreased IL17-A and RORγt in CD4+ T cells of the colonic LP and MLN (Figure 5D and 5E and Figure S13). This data supports the in silico prediction that activation of PPARγ in Th17 cells favors differentiation into iTreg cells, which facilitates colonic tissue reconstitution and recovery from disease.
Computational models can help to synthesize and integrate existing knowledge and narrow the experimental design prior to costly in vivo experimentation. To gain a more comprehensive understanding of the mechanisms controlling CD4+ T cell differentiation, we first compiled and integrated existing literature knowledge and data related to the cytokines and intracellular signaling pathways involved in the differentiation of a naïve CD4+ T cell into effector and regulatory cell subsets. To determine whether the model predictions regarding novel mechanisms of immunoregulation in Th17 and Treg cells were sensitive to the model parameters we performed a sensitivity analysis of the signaling pathways controlling Th17 and iTreg phenotypes. Our simulations reproduced known CD4+ T cell differentiation behaviors for Th1, Th2, Th17 and iTreg, and predicted novel mechanisms of T cell-mediated immunoregulation. By simulating the cytokine milieu that surrounds a CD4+ T cell in silico, we dissected crucial signaling pathways and their transcriptional regulation programs involved in differentiation and plasticity of CD4+ T cells. While computational predictions carry certain uncertainty given by the topology of the network, computational modeling approaches applied to CD4+ T cell differentiation have proven useful in characterizing the importance of dual waves of expression of T-bet and sequentially acting positive feedback loops of TCR-IFNγ-STAT1-Tbet and IL-12-STAT4-Tbet signaling in Th1 differentiation [32]. A central question in T cell biology involves improving the understanding of instructive versus selective factors that regulate the differentiation process. Selective factors include competition for cytokines by competing clones of CD4+ T cells in an expanding population. For example, regulatory T cells are able to outcompete for IL-2 and deprive effector T cells of this survival signal [33]. While the computational model presented herein comprehensively addresses the instructive factors (i. e. , the impact of cytokine combinations on T cell phenotypes), stochastic simulations and multiscale modeling are needed to adequately model selective factors by linking molecular-level intracellular signaling sub-models and tissue-level cell-cell interaction models. Some studies have addressed selective factors by focusing on the crosstalk in molecular pathways in an expanding Th1 population using in vitro data [34] but only one phenotype has been computed and with a limited scope. The study presented here is the first to comprehensively investigate at the systems level the mechanisms controlling CD4+ T cell differentiation and plasticity between Th17 and iTreg cells, presenting a model that computes not only one but four of the CD4+ T cell phenotypes. Several distinct signals regulate CD4+ T cell activation and differentiation: a signal from the T cell receptor (TCR) interacting with MHC, a co-stimulatory signal (i. e. , CD28 interacting with B7. 1 or B7. 2 on antigen presenting cells), and a cytokine-driven signal. Other studies have more narrowly focused on CD4+ T cell proliferation [35], TCR signaling [36] or co-stimulatory signals [37]. Our mathematical approach more comprehensively studies the non-cognate interactions (i. e. , cytokine milieu) and instructive factors controlling CD4+ T cell differentiation. Future studies will leverage the modeling efforts described here to construct multi-scale hybrid models driven by high-performance computing strategies that integrate sub-models of intracellular signaling pathways such as the CD4+ T cell model and tissue-level models that can simulate cell-cell interactions. These integrative approaches will provide an avenue for incorporating stochasticity as well as the modulation of phenotype and function of immune cells at sites of inflammation or infection by selective and instructive factors. Sensitivity analyses and computational simulations using the CD4+ T cell differentiation model predicted that the nuclear receptor PPARγ modulates the balance between Th17 and iTreg cells, by controlling both the initial differentiation from a naïve CD4+ T cell as well as plasticity between phenotypes. Activation of PPARγ in silico favored differentiation of iTreg and antagonized Th17 differentiation by down-modulating RORγt and IL-17. These findings are in line with previous reports demonstrating that the pharmacologic activation of PPARγ selectively controls Th17 differentiation in mice and humans by interfering with RORγt transcription [38]. Furthermore, ciglitazone, a PPARγ agonist, significantly enhanced generation of iTreg cells [26] and PPARγ induced potent and stable FOXP3 expression [27] resulting in the suppression of effector CD4+ T cell responses [28]. Our in silico results demonstrate that the upregulation of FOXP3 and downregulation of RORγt and IL-17 in CD4+ T cells is modulated by PPARγ and behaves in a dose-dependent manner. Indeed, our in vitro results support the dose-dependent effect in the suppression of Th17, although not accompanied by a similar increase in FOXP3+ iTreg cells. However, our in vivo findings further demonstrate that pioglitazone treatment favors a switch of fully differentiated Th17 cells into an iTreg phenotype by increasing activation of PPARγ. Thus, our plasticity modeling efforts are more predictive of in vivo than in vitro behaviors of CD4+ T cells, suggesting a missing component, possibly provided by APCs in the widely utilized in vitro system. For instance, all trans retinoic acid, which in vivo is produced by APC-derived, increased and maintained FOXP3 expression [39]. Conclusively, the mechanisms by which T cell extrinsic factors modulate CD4+ T cell plasticity are yet not fully understood. Here, however, we propose PPARγ as a novel candidate for such modulation. The CD4+ T cell mathematical model predicted an upregulation of RORγt and IL-17 in Th17 cells lacking PPARγ when compared to the wild-type counterparts. In complete correspondence to this modeling prediction, our in vitro results show that following Th17 differentiation, CD4+ T cells lacking PPARγ exhibit a more dramatic upregulation of RORγt and IL-17A than wild-type cells. Also, we have also observed a marginal upregulation of FOXP3 in wild-type cells. The uncoupling between the dramatic downregulation of RORγt and the more limited upregulation of FOXP3 observed in vitro could be attributed to external factors that play an important role in this process, which are not fully mechanistically understood or not included in the in vitro system used (i. e. , APCs). As opposed to the in vitro results, the in vivo findings in mice with CD4+ T cell-induced colitis were more consistent with the modeling predictions. Recent studies show that changes in the cytokine environment mediate the conversion of iTreg into Th17 cells [6]. Notably, different subsets of myeloid cells in humans can orchestrate the differentiation of naïve CD4+ T cells into either effector or regulatory phenotypes [7]. Myeloid APCs are essential for the induction of IL-17A+ FOXP3+ T cells from memory CCR6+ T cells or Treg cells [40]. At the colonic mucosa, numbers and functions of IL-17-producing cells are tightly controlled by PPARγ, and its modulation of the dual roles of Th17 cells as effectors of pathogenic, tissue-damaging versus pathogen-clearing responses has been investigated in the context of Clostridium difficile and enteroaggregative Escherichia coli infections [41], [42]. However, the mechanisms controlling CD4+ T cell plasticity at the gut mucosa remain largely unknown, including the essential and dispensable regulators of these processes. Herein, we combined computational and experimental approaches to investigate for the first time the role of PPARγ in the re-programming of fully differentiated Th17 cells into an iTreg phenotype in the gut mucosa. Of note, the presence of FOXP3 RORγt double-positive cells with suppressive actions on effector CD4+ T cell subsets has been associated with the plasticity of Th17 and iTreg [18]. TGF-β is a common inductor of Th17 and iTreg that can upregulate FOXP3, but in combination with IL-6, it upregulates IL-17 and dramatically downregulates FOXP3 expression [2]. Other cytokines, such as IL-23, modulate plasticity by restraining FOXP3+ Treg activity [14]. Clinically, inhibition of IL-17 promotes differentiation of stable iTreg cells in patients with autoimmune hepatitis [43]. However, IL-17+FOXP3+ cells were identified in inflamed intestinal mucosa of patients with Crohn' s disease (CD), but not in patients with ulcerative colitis (UC) [44], the two clinical manifestations of inflammatory bowel disease. Furthermore, in line with our sensitivity analysis and computer simulations, results of our adoptive transfer studies in mice indicate that activation of PPARγ by oral pioglitazone administration favors a switch from Th17 to iTreg in MLN and colonic LP of mice with CD4+ T cell-induced colitis, thereby demonstrating that PPARγ is implicated in the modulation of CD4+ T cell plasticity in vivo. The loss of PPARγ favored Th17 differentiation and reduced the conversion of IL-17A-producing Th17 cells into CD4+FOXP3+ T cells in vivo. Adoptive transfer studies using T cell-specific PPARγ null naïve T cells demonstrate that PPARγ is needed for suppressing effector responses at sites of inflammation such as the colonic LP in a mouse model of chronic colitis. Interestingly, FOXP3 inhibits Th17 by antagonizing the function of the transcription factors RORγt and RORα [6], [19]. This suggests a potential interaction of RORγt with FOXP3 in larger transcriptional complexes, which could explain why RORγt is more rapidly down-regulated than FOXP3 is increased. More specifically, the decrease of RORγt could result from a synergism between the inhibition exerted by PPARγ and the parallel inhibition caused by FOXP3, which in turn is enhanced when PPARγ is activated. The observation that PPARγ may interact with FOXP3 and RORγt suggests a cross-talk between transcriptional programs of crucial importance to the regulation of immune responses and clinical outcomes during infectious and immune-mediated diseases. In summary, we demonstrate for the first time that activation of PPARγ results in reprogramming of the CD4+ T cell molecular pathways that control the Th17 phenotype, leading to the induction of an iTreg phenotype. This phenotype switch is associated with protection from CD4+ T cell-induced colitis during adoptive transfer experiments in mice. Thus, the balance between Th17 and Treg cells helps delineate the outcome of immunological processes from effector inflammation to regulatory tolerance. Our modeling approaches allowed us to narrow the design of experiments and to better understand the molecular mechanisms of action controlling CD4+ differentiation. This new mechanistic knowledge is broadly applicable to the development of immune therapeutics for infectious, allergic and immune-mediated diseases. More specifically, we propose that PPARγ is a promising therapeutic target for chronic inflammatory and infectious diseases where Th17 cells contribute to the gut immunopathogenesis.
All experimental protocols were approved by the Virginia Tech institutional animal care and use committee (IACUC) (Protocol Number: 10-087VBI) and met or exceeded guidelines of the National Institutes of Health Office of Laboratory Animal Welfare and Public Health Service policy. Animals were under strict monitoring throughout the duration of the project and all efforts were made to minimize unnecessary pain and distress. Mice were euthanized by carbon dioxide narcosis followed by secondary cervical dislocation. To facilitate a comprehensive representation of the dynamics associated with the major non-cognate pathways controlling CD4+ T cell differentiation and plasticity, we constructed an ordinary differential equation (ODE) -based computational model of the cytokines, receptors and transcription factors controlling CD4+ T cell differentiation and plasticity (Figure 1, Text S1). The mathematical model was engineered to ensure proper modulation of intracellular pathways and cell phenotypes via external cytokines representing the cytokine milieu. The mathematical model constructed was based on experimental findings and illustrates intracellular pathways controlling a naïve T cell differentiation into Th1, Th2, Th17 or iTreg phenotypes. The model comprises 60 differential equations representing 52 reactions and 93 species (Figure S3). The COmplex PAthway SImulator software [45] (COPASI; http: //www. modelingimmunity. org/) was used for model development, sensitivity analysis, and calibration. Sensitivities of the steady-state fluxes of reactions were derived with respect to the reaction rates in the system. These sensitivities were normalized and represented flux control coefficients according to Metabolic Control Analysis (MCA) [46], [47]. In this case, sensitivities were performed with respect to PPARγ pathway-controlling parameters and levels of different species were assessed. The model was calibrated to experimental data (Table S1), which varied external concentration of cytokines and resulted in different phenotypes described by varying levels of transcription factors and proteins. We used the ParticleSwarm algorithm implemented in COPASI to determine unknown model parameter values and fully calibrate the model (Table S2 and S3, Figure S4). The resulting model adequately computes the differentiation of CD4+ T cells into the four phenotypes: Th1 with external IFNγ, IL-12, IL-18 and αIL-4 addition, Th2 with IL-4 and αIFNγ addition and iTreg with IL-2 and external TGFβ addition (Figure S5). Also, to induce Th17 differentiation from a naïve CD4+ T cell, external IL-6 and external TGF-β were added in combination and demonstrated upregulation of RORγt, IL-17 and STAT-3. In silico simulation consisted of time-courses or parameter scans. Also, the combination of both was performed. In this last case, each plotted line has an incremented concentration of the parameter being scanned. Thus, differential patterns of expression of molecules, either upregulated or downregulation, over time can be observed by looking at the arrows in each molecule. This model is available at www. modelingimmunity. org and model assumptions and model predictions are available in the supplementary materials (Table S4 and Table S6 respectively). Also a complete table with all the numerical values of all parameters of the model is provided in the supplementary materials (Table S7). B6. CB17-Prkdcscid/SzJ (SCID), B6. 129P2 (Cg) -Rorctm2Litt/J, C57BL/6J and B6 (Cg) -Rag2tm1. 1Cgn/J were purchased from The Jackson Laboratory and housed under specific pathogen-free conditions in ventilated racks. The mice were maintained in the animal facilities at Virginia Tech. All experimental protocols were approved by the institutional animal care and use committee at Virginia Tech and met or exceeded guidelines of the National Institutes of Health Office of Laboratory Animal Welfare and Public Health Service policy. Spleens and mesenteric lymph nodes (MLN) were excised and crushed in 1×PBS/5% FBS using the frosted ends of two sterile microscope slides. Single cell suspensions were centrifuged at 300× g for 10 min and washed once with 1×PBS. Red blood cells were removed by osmotic lysis prior to the washing step. All cell pellets were resuspended in FACS buffer (1×PBS supplemented with 5% FBS and 0. 09% sodium azide) and subjected to flow cytometric analysis. Paralelly, colons were excised and lamina propria leukocytes (LPL) were isolated. Tissue pieces were washed in CMF (1× HBSS/10% FBS/25 mM Hepes), and tissue was incubated twice with CMF/5 mM EDTA for 15 min at 37°C while stirring. After washing with 1×PBS, tissue was further digested in CMF supplemented with 300 U/ml type VIII collagenase and 50 U/ml DNAse I (both Sigma-Aldrich) for 1. 5 hs at 37°C while stirring. After filtering the supernatants, cells were washed once in 1×PBS, pellets were resuspended in FACS buffer and subjected to flow cytometric analysis. For fluorescent staining of immune cell subsets 4–6×105 cells were incubated for 20 min with fluorochrome-conjugated primary mouse specific antibodies: anti-CD3 PE-Cy5 clone 145-2C11 (eBioscience), anti-CD4 PE-Cy7 clone GK1. 5 (eBioscience), anti-CD4 APC clone RM4-5 and anti-CD25 Biotin clone 7D4 (BD Biosciences). Cells were washed with FACS buffer (1×PBS supplemented with 5% FBS and 0. 09% sodium azide). For intracellular staining of transcription factors and cytokines, cells were fixed and permeabilized using a commercial kit according to the manufacturer' s instructions (eBioscience). Briefly, cells were fixed and permeabilized for 20 minutes, Fc receptors were blocked with mouse anti-CD16/CD32 FcBlock (BD Biosciences) and cells were stained with fluorochrome-conjugated antibodies towards anti-mouse, FOXP3 FITC clone FJK-16s, anti-mouse ROR gamma (t) PE, clone B2B and anti-mouse IL17-A APC, clone eBio17B7 (eBioscience). All samples were stored fixed at 4°C in the dark until acquisition on a LSR II flow cytometer (BD Biosciences). A live cell gate (FSC-A, SSC-A) was applied to all samples followed by single cell gating (FSC-H, FSC-W) before cells were analyzed for the expression of specific markers. Data analysis was performed with FACS Diva (BD Biosciences) and Flow Jo (Tree Star Inc.). Six-week-old SCID and RAG2-/- mice were administered intraperitoneally (i. p.) 4×105 CD4+ CD45RBhigh CD25- from either CD4 null PPAR γ fl/fl or C57BL/6J (wild-type), or B6. 129P2 (Cg) -Rorctm2Litt/J mice. Mice were weighed on a weekly basis and clinical signs of disease were recorded daily for 14 wk. Mice that developed severe signs of wasting disease were sacrificed. Otherwise, mice were sacrificed 90 days after transfer. Splenocytes obtained from CD4 null PPAR-γ fl/fl or C57BL/6J (wild-type) mice were enriched in CD4+ T cells by magnetic negative sorting using the I-Mag cell separation system (BD Pharmingen). Cells were incubated with a mixture of biotinylated Abs followed by a second incubation with streptavidin particles and exposed to a magnet to remove unwanted cells. The purity of the CD4+-enriched cell suspension was between 93 and 96%. CD4-enriched cells were used for adoptive transfer, or further purified by FACS. For FACS sorting, cells were labeled with CD45RB, CD4, and CD25 and separated into CD4+ CD45RBhigh CD25- cells (i. e. , effector T cells) in a FACSAria cell sorter (BD Biosciences). The purity of the FACS-sorted CD4+ subsets was ≥98%. CD4+CD62L+ cells from either wild-type or T PPARγ null (CD4Cre+) mice were sorted using magnetic activated cell sorting (MACS, Miltenyi Biotec) and stimulated with plate bound anti-CD3 (5 µg/ml, BD Biosciences) under Th17 conditions with 2. 5 ng/ml hTGF-β1 (R&D Systems), 25 ng/ml IL-6 (Peprotech), 10 µg/ml anti-IL-4 (clone 11B11, R&D Systems), and 10 µg/ml anti-IFN-γ (clone XMG1. 2, R&D Systems). 60 hours after activation, an aliquot was obtained to check purity and DMSO-diluted pioglitazone (PIO, Cayman Chemicals) was added to the media at 0,0. 1,1, 10,40 or 80 µM. Control (0 µM PIO) was treated with DMSO only. 24 hours after treatment Th17 cells were restimulated with PMA (50 ng/mL, Acros Organics) and ionomycin (500 ng/mL, Sigma) in the presence of BD GolgiStop (BD Biosciences) for 6 h, after which intracellular staining was performed. The experiment was repeated three times for consistency. Co-stimulation of with CD28 has been described to downregulate Th17 development [37], [48]. We also performed optimization studies for Th17 differentiation using CD28 as a co-stimulatory signal and the addition of recombinant IL-23 in the cytokine cocktail, however, no differences were observed. Co-stimulation signaling optimization studies were run adding either 0 or 2. 5 µg/mL of αCD28 in the media. No differences were found. Thus, the data presented are with αCD3 stimulation only. Colonic sections were fixed in 10% buffered neutral formalin, later embedded in paraffin and then sectioned (5 µm) and stained with H&E stain for histological examination. Colons were graded with a compounded histological score including the extent of (1) leukocyte infiltration, (2) mucosal thickening and (3) epithelial cell erosion. The sections were graded with a score of 0–4 for each of the previous categories, and data were analyzed as a normalized compounded score. Parametric data were analyzed using the ANOVA followed by Scheffe' s multiple comparison method. Nonparametric data were analyzed by using the Mann-Whitney' s U test followed by a Dunn' s multiple comparisons test. ANOVA was performed by using the general linear model procedure of SAS, release 6. 0. 3 (SAS Institute). Statistical significance was assessed at a P≤0. 05. | CD4+ T cells can differentiate into different phenotypes depending on the cytokine milieu. Due to the complexity of this process, we have constructed a computational and mathematical model with sixty ordinary differential equations representing a CD4+ T cell differentiating into either Th1, Th2, Th17 or iTreg cells. The model includes cytokines, nuclear receptors and transcription factors that define fate and function of CD4+ T cells. Computational simulations illustrate how a proinflammatory Th17 cell can undergo reprogramming into an anti-inflammatory iTreg phenotype following PPARγ activation. This modeling-derived hypothesis has been validated with in vitro and in vivo experiments. Experimental data support the modeling-derived prediction and demonstrate that the loss of PPARγ enhances a proinflammatory response characterized by Th17 in colitis-induced mice. Moreover, pharmacological activation of PPARγ in vivo can affect the Th17/iTreg balance by upregulating FOXP3 and downregulating IL-17A and RORγt. In summary, we demonstrate that computational simulations using our CD4+ T cell model provide novel unforeseen hypotheses related to the molecular mechanisms controlling differentiation and function of CD4+ T cells. In vivo findings validated the modeling prediction that PPARγ modulates differentiation and plasticity of CD4+ T cells in mice. | Abstract
Introduction
Results
Discussion
Materials and Methods | medicine
computer science
adaptive immunity
computer modeling
immune cells
immunity
gastroenterology and hepatology
t cells
inflammatory bowel disease
immunology
biology
immunoregulation
immunomodulation | 2013 | Systems Modeling of Molecular Mechanisms Controlling Cytokine-driven CD4+ T Cell Differentiation and Phenotype Plasticity | 9,518 | 320 |
In the short-germ beetle Tribolium castaneum, waves of pair-rule gene expression propagate from the posterior end of the embryo towards the anterior and eventually freeze into stable stripes, partitioning the anterior-posterior axis into segments. Similar waves in vertebrates are assumed to arise due to the modulation of a molecular clock by a posterior-to-anterior frequency gradient. However, neither a molecular candidate nor a functional role has been identified to date for such a frequency gradient, either in vertebrates or elsewhere. Here we provide evidence that the posterior gradient of Tc-caudal expression regulates the oscillation frequency of pair-rule gene expression in Tribolium. We show this by analyzing the spatiotemporal dynamics of Tc-even-skipped expression in strong and mild knockdown of Tc-caudal, and by correlating the extension, level and slope of the Tc-caudal expression gradient to the spatiotemporal dynamics of Tc-even-skipped expression in wild type as well as in different RNAi knockdowns of Tc-caudal regulators. Further, we show that besides its absolute importance for stripe generation in the static phase of the Tribolium blastoderm, a frequency gradient might serve as a buffer against noise during axis elongation phase in Tribolium as well as vertebrates. Our results highlight the role of frequency gradients in pattern formation.
The anterior-posterior (AP) axis of arthropods, annelids, and vertebrates is partitioned into segments. The French flag model, in which threshold concentrations of morphogen gradients are interpreted by downstream genes to partition a developing tissue [1], [2], provides the main theoretical framework explaining segmentation in Drosophila. Specifically, gradients of maternal factors span the AP axis of Drosophila providing positional information to downstream gap genes, which in turn diffuse in the syncytial blastoderm to form more localized morphogen gradients. Both maternal and gap gene gradients provide further positional information to the pair-rule genes whose striped expression is the first indication of segmentation in the embryo [3]. In Drosophila, all segments form more or less simultaneously in a syncytial blastoderm of fixed AP axis length. In contrast, vertebrate segmentation (somitogenesis) takes place sequentially in an elongating and cellularized embryo. A different model, the ‘clock and wavefront’ explains segmentation in vertebrates [4], [5]. Multiple genes (hairy/enhancer-of-split and genes of Notch, Wnt and FGF signaling pathways) show oscillatory expression in the presomitic mesoderm (PSM) of the vertebrate embryo and are thought to be constituents of a molecular clock [6], [7]. In cells located anterior to a wavefront, oscillations are arrested into stable stripes. The wavefront is thought to be defined by a moving threshold that forms within the overlapping posterior gradients of Wnt and FGF [8], [9] and an opposing retinoic acid gradient [10]. Oscillations seem to arrest gradually (i. e. they are modulated by a frequency gradient) as evidenced by kinematic expression waves that sweep the PSM from posterior to anterior [7]. In most short-germ arthropods, anterior segments form in a blastoderm, as in Drosophila, while posterior segments form subsequently during the germband stage out of a population of cells at the posterior end of the embryo (termed the ‘growth zone’) [11], reminiscent of somitogenesis in vertebrates. Although it is conceivable that short-germ arthropods utilize a ‘French flag’-based segmentation mechanism in the blastoderm and a ‘clock and wavefront’ mechanism in the germband, it has recently been shown that a segmentation clock operates in both the germband [12] and blastoderm [13] of the short-germ insect Tribolium castaneum, where waves of pair-rule gene expression (specifically Tc-even-skipped (Tc-eve) ) propagate from posterior to anterior [13]. The identification of factors that provide positional information for segmentation in the blastoderm of short-germ arthropods has been controversial [14]–[18]. Demonstration of the clock-based nature of short-germ segmentation fuels this debate as attention now turns to the search for factors functioning as a wavefront. The homeodomain transcription factor Caudal (Cad) has been implicated as playing a prominent role in AP patterning in arthropods since its expression overlaps with the newly forming stripes [19]. Cad is required for segmentation in the Drosophila abdomen [20], and for posterior patterning in other species [21], [22]. It plays an even more prominent role in non-diptran insect segmentation; it is required for trunk segmentation in Nasonia vitripennis [23] and for both trunk and gnathal segmentation in Tribolium castaneum [24] and Gryllus bimaculatus [25]. However, the exact role of Cad in segmentation is still not known. Here we test the hypothesis that the posterior gradient of Tribolium cad (Tc-cad) expression regulates the oscillation frequency of pair-rule gene expression to produce kinematic waves in the Tribolium blastoderm. We found that the expression of Tc-eve was abolished in strong Tc-cad RNAi knock-down embryos, but in weak Tc-cad knock-down embryos, the Tc-eve expression domain was posteriorly shifted and its oscillation frequency reduced. Perturbing the Tc-cad gradient in different ways by knocking-down its regulators further demonstrated that the extension, intensity, and slope of the Tc-cad gradient correlated with the extension, frequency, and width of Tc-eve expression waves, respectively. As shown by computer simulations, these observations are consistent with the hypothesis that Tc-cad functions as a frequency gradient regulating the spatiotemporal dynamics of pair-rule gene oscillation in Tribolium. These observations, combined with the continued expression of Tc-cad in a posterior gradient suggest that Tc-cad also acts as a wavefront in the elongating germband. Our study highlights the concept of a frequency gradient as a pattern formation mechanism. Using computer modeling, we also showed that a graded frequency profile might even be essential within the clock-and-wavefront model as a buffer against noise.
The wave dynamics of Tc-eve in Tribolium can be explained by assuming a posterior-to-anterior gradient that positively regulates the frequency of Tc-eve oscillations [13]. Tc-cad is an obvious candidate to encode such a frequency gradient because its mRNA expression forms a posterior-to-anterior gradient that overlaps the Tc-eve expression waves arising at the posterior throughout Tribolium segmentation (Figure 1 A–D). Since studying segmentation in the germband phase of Tribolium development is hindered by the truncation phenotype generated by most segmentation gene knock-downs, we largely restricted our analysis to the stripes that form during the blastoderm stage. The expression of Tc-cad in the blastoderm (Figure 1 E) is approximated with reasonable accuracy by a posterior-to-anterior linear gradient that plateaus at the posterior end (Figure 1 F; Text S3). We used three descriptors to characterize this gradient: maximum posterior (plateau) value, position of anterior border, and slope (Figure 1 F). We analyzed the temporal dynamics of the Tc-cad gradient by calculating its three descriptors at 14–17 and 17–20 hours after egg lay (AEL) (Figure 1 G), spanning the formation of the first and second Tc-eve expression stripes in wild type (WT) [13] (analysis of later times was precluded by primitive pit formation, asterisk in Figure 1 C). As shown in Figure 1 G, the anterior border of Tc-cad expression gradient did not experience a significant shift during the formation of the first and second Tc-eve stripes, (which is also evident in Figure 1 A, B). However, both the maximum posterior value and the slope of the Tc-cad gradient increased over time. This indicates that the Tc-cad gradient was building up during the formation of the first and second Tc-eve stripes, but did not undergo a substantial shift along the AP axis (Figure 1 H). Characterizing Tc-cad gradient dynamics with higher temporal resolution (Figure S1) indicates that this buildup phase occurred between 14 to 16 hours AEL (i. e. before completion of the first Tc-eve stripe), after which the gradient was more or less static. This argues against a substantial influence of Tc-cad temporal dynamics on the wave dynamics of Tc-eve expression in the blastoderm. By the time the third stripe formed in the germ rudiment, the Tc-cad gradient had retreated toward posterior (Figure 1 C). The spatial distribution of Tc-cad renders it a probable wavefront candidate in a clock-and-wavefront model. In the traditional model, a wavefront should retract posteriorly (like Tc-cad expression during the germband stage). However, a static but smooth gradient (like Tc-cad expression during the formation of first and second Tc-eve stripes in the blastoderm) that modulates the frequency of Tc-eve oscillation is, in principle, capable of forming a striped expression pattern (Movies S1, lower panel) [13], [26]. Taking the initial buildup phase of the Tc-cad gradient into consideration (Movies S1, upper panel) yields similar results. However, this buildup phase is expected to slow down the formation of the first stripe (Figure S2). This agrees with experiment, since the first cycle of Tc-eve oscillations starts at 13. 5 hours AEL and ends at 17 hours AEL (i. e. the first stripe takes 3. 5 hours to form), while the second cycle starts at 17 hours AEL and ends at 20 hours (i. e. the second stripe takes 3 hours to form) [13]. However, this is not obvious in the timing results presented here (see below), since we chose to start our analysis at 14 hours AEL. In both vertebrates and arthropods, canonical Wnt is a positive regulator of cdx/cad [25], [27]–[29]. Once bound by Wnt ligand, the receptor Frizzled recruits the β-catenin destruction complex (comprising Axin, APC, and other factors), rendering β-catenin free to enter the nucleus and bind Pangolin (TCF) with the help of Legless (Lgs), Pygopous (Pygo) and other coactivators [30] to activate Wnt targets. In Tribolium, wnt1 and wnt8 are expressed at the posterior pole of the blastoderm, and at the posterior end of the growth-zone in the germband [31], which is expected to produce a posterior gradient of Wnt activity, the formation of which is enhanced by the anterior localization of Wnt repressors in the blastoderm [28], [32]. Manipulating Wnt activity affected Tc-cad expression in the Tribolium blastoderm. Knocking down Tc-lgs (a positive Wnt regulator) by means of maternal RNAi (Methods) shifted the Tc-cad expression gradient posteriorly (compare Figure 2 C–C′ to Figure 2 A–A′). In addition, the posterior maximum value of Tc-cad and slope of the gradient were reduced in Tc-lgs RNAi embryos compared to WT (Figure 2 D–D″). Knocking down Tc-apc1 (a negative Wnt regulator) repositioned the Tc-cad gradient anteriorly (Figure 2 G–H″). Interestingly, the maximum posterior value of the Tc-cad expression gradient at 14–17 hours AEL was lower in Tc-apc1 RNAi embryos than in WT embryos (Figure 2 H), but eventually reached WT levels by 17–20 hours AEL (Figure 2 H′). Thus, it appears that the Tc-cad expression gradient takes longer to mature in Tc-apc1 RNAi than in WT embryos, which might be indicative of early negative Wnt regulation of Tc-cad. Knocking down another Wnt regulator, Tc-pan, also perturbed the Tc-cad expression gradient. Pan, a component of the activator complex, also acts as a repressor in the absence of nuclear β-catenin [33]. Hence, we expected Wnt activity to be reduced posteriorly but increased anteriorly in Tc-pan RNAi embryos compared to WT, resulting in a shallower Wnt gradient across the blastoderm, and consequently a shallower Tc-cad gradient. As expected, the border of the Tc-cad gradient in Tc-pan RNAi embryos shifted anteriorly, the gradient reached a lower maximum posterior value, and the slope was lower compared to WT (Figure 2 E–F″). In Drosophila, two Hox3 type genes are involved in early patterning: bicoid (bcd), which is expressed anteriorly and plays a major role in AP patterning, and zerknüllt (zen), which is expressed dorsally and specifies the amnioserosa [34]. Tribolium lacks bcd [17] but one of its zen homologs, Tc-zen1, is expressed both anteriorly and dorsally [35]. Anterior expression precedes dorsal expression and is suspected to play a role in AP patterning [36]. As shown in Figure 2 I–J″, the Tc-cad gradient in Tc-zen1 RNAi embryos shifted anteriorly, but had the same slope and maximum posterior expression level as WT, indicating that Tc-zen1 represses Tc-cad anteriorly (see Figure 2 B for a summary of Tc-cad regulation). In Tribolium, Tc-eve is expressed in waves that shrink while propagating from posterior to anterior (Figure 3 A) [13]. Tc-eve and Tc-cad RNAi embryo display similar phenotypes lacking all post oral segments, and previous studies implicate cad in the regulation of eve in arthropods [24], [25]. Axis elongation is an essential component of the clock-and-wavefront model. We have previously shown that blastoderm segmentation in Tribolium seems to be clock-based [13]. Despite the lack of axis elongation at the blastoderm stage, we did not exclude the possible existence of a retreating frequency gradient (wavefront). In the current study, we provide evidence that Tc-cad expression acts as a frequency gradient that modulates pair-rule gene oscillations in the blastoderm. Although a static step frequency gradient (i. e. suddenly dropping from non-zero to zero frequency) does not possess any patterning capacity, a static but gradually decreasing frequency gradient can generate a striped pattern [26]. Indeed, the first two stripes of Tc-eve form during a time period when the Tc-cad gradient is largely static. After the formation of the first two stripes, Tc-cad expression then abruptly retreats to the prospective growth zone (Figure 1 C). Later during axis elongation in the germband stage, Tc-cad expression retreats posteriorly with every newly forming Tc-eve stripe (Figure 1 D). However, in principle, a step frequency gradient is capable of generating a striped pattern during the germband retraction phase. In vertebrates, a transition from high to low frequency (termed the ‘arrest front’) is thought to be determined by a threshold within a retracting posterior gradient. Such a mechanism might be very sensitive to the location of the threshold. Uncertainty in threshold location due to noise might lead to the generation of noisy patterns. On the other hand, gradually arresting oscillations would average out the noise and make the mechanism independent of precise threshold location. To investigate this, we developed two computer models for the clock-and-wavefront mechanism: one that utilizes a step frequency gradient by applying a threshold on a retracting smooth gradient (threshold-based model), and the other utilizes a smooth retracting frequency gradient without applying any thresholds (threshold-free model). Both generated similar striped patterns in the absence of noise (Figures 6 A–D; Movies S7 and S8). We then investigated the performance of both models after introducing random fluctuations in the intensity of the posterior gradient at each cell. This is expected to result in independent random shifts in threshold locations across the lateral axis of the embryo, which ultimately leads to salt-and-pepper noise at the stripe borders; however, the threshold-free model is more robust to this type of noise than the threshold-based model (Figures 6 E–H; Movies S9 and S10).
In Drosophila, maternal cad mRNA (Dm-cad) is ubiquitously expressed in the early blastoderm [37]. A posterior-to-anterior protein gradient of Dm-Cad forms due to translational repression by a reciprocal gradient of Dm-Bicoid [38]. Dm-Cad acts as an activator of posterior gap [39] and pair-rule genes [40] and binds to the enhancers of these genes [41], [42]. However, the mild segmentation defects in embryos in which the shape of Dm-Cad gradient has been altered argues against its function as a morphogen gradient [20], [43]. In the wasp Nasonia vitripennis, Nv-cad plays a more prominent role in activating gap and pair-rule genes, and a limited positioning role [23]. In the cricket Gryllus bimaculatus, Gb-cad was found to activate the pair-rule gene Gb-eve, and activate and position gap gene domains. This indicates that cad might act as a morphogen gradient in non-dipteran insects. In this study, we described similar results in Tribolium. We showed that in strong Tc-cad RNAi, expression of Tc-eve was abolished (Figure S2 A); while in weak Tc-cad RNAi, Tc-eve expression was posteriorly shifted (Figure 3 B). However, a morphogen gradient acting through concentration thresholds is less likely to act in positioning the highly dynamic pair-rule gene expression domains in Tribolium. Instead, we argue that Tc-cad regulates the frequency of a pair-rule clock to produce the observed wave dynamics. Three cad homologs are found in mouse: Cdx1, Cdx2, and Cdx4. They are expressed in nested domains in the posterior end of the embryo. The Cdx1–Cdx2 double mutant exhibits fused somites [44], suggesting a role in somitogenesis. However, the Cdx1–Cdx2 double mutant also shows down-regulation of some caudalizing factors involved in somitogenesis (such as wnt3a) that are themselves Cdx regulators [45], [46]. Cdx genes also directly regulate Hox genes in a dose dependent manner [47], [48], and even regulate their activation times [49]. In summary, cad (-related) genes are involved in posterior patterning in many species. While it is not clear whether they play a permissive or instructive role, there is evidence that they might act as a morphogen gradient for gap genes in basal insects (like in Gryllus) and for Hox genes in vertebrates. In this study, we showed that Tc-cad regulates the spatiotemporal dynamics of Tribolium pair-rule genes in a dose dependent manner, stressing the instructive role of cad in the development of a non-dipteran insect. However, we cannot exclude the possibility that Tc-cad regulates pair-rule genes indirectly. Indeed, Tc-cad and Wnt might cross-regulate in a positive feedback loop to form identical gradients. In this case, it is hard to decide which is the direct regulator (or whether both Wnt and Tc-Cad are direct regulators) of Tc-eve expression without performing detailed cis-regulatory analysis of the Tc-eve locus. In the original formulation of the clock-and-wavefront model, the anterior-to-posterior movement of a step frequency profile (i. e. suddenly dropping from non-zero to zero frequency) over an oscillating field of cells sequentially generates a striped pattern in an anterior-to-posterior order [4]. Later, this mechanism was modified by assuming a graded frequency profile to accommodate the observation that oscillations organize into kinematic waves in the chick PSM [7]. Several efforts have been made to identify molecular gradient (s) that regulate the frequency of the vertebrate segmentation clock. A posterior-to-anterior Wnt activity gradient was found to define the PSM oscillation domain in the mouse [50], [51]. Furthermore, down-regulation of Wnt activity reduced the clock frequency in both mouse and chick [52]. However, elevated and flattened constitutive stabilization of β-catenin in the mouse PSM only extended the oscillation domain, arguing against a role for the shape of Wnt activity gradient in segmentation [50]. A posterior-to-anterior FGF gradient in the PSM was found to define where oscillations arrest [9], [53], [54], but manipulating the level of FGF signaling does not alter the clock period [9], [52]. A gradient of Her13. 2 in zebrafish was suggested to modulate clock frequency through heterodimerization with other zebrafish clock constituents: Her1 and Her7 [55], [56]. However, this idea was recently challenged and an alternative model of gradual switching between two oscillatory modules was suggested [57]. It is not known whether the gradual arrest of oscillations and the resulting kinematic waves in vertebrates have any functional role or are a mere peculiarity, since, based on computer simulations of the clock-and-wavefront model, stripe widths depend only on the wavefront velocity and the maximum clock period, not on the shape of the frequency profile [5]. Although used for cosmetic means within the clock-and-wavefront model, a graded frequency profile (even a static one) by itself has a patterning capacity [26]; kinematic waves were observed in an oscillating Zhabotinskii chemical reaction, where a reactant controlling the frequency of oscillation is distributed in a gradient [58], [59]. Since a static step frequency profile is unable to generate any stripes, the patterning capacity of a graded frequency profile might explain how blastodermal Tc-eve stripes in Tribolium form in the absence of axis elongation. Although the possibility of a yet unidentified frequency gradient that sweeps across the blastoderm still exists, we showed in this study that a strong candidate for the frequency gradient in Tribolium, Tc-cad, does not appreciably shift during the formation of the first two Tc-eve stripes (Figure 1 G, H). In addition to its necessity in the absence of axis elongation, a graded frequency profile renders the clock-and-wavefront robust against noise in wavefront gene expression, as shown by computer simulations (Figure 6 and Movies S7, S8, S9, S10). This improvement in robustness might be due to the distributed nature by which oscillations are arrested in a graded frequency profile, in contrast to the total reliance on a single threshold in a step frequency profile. This and other recent works reinforce the importance of the concept of a frequency (or phase) gradient in sequential patterning [60], [61]. In clock-based segmentation models that utilize a static frequency gradient, stripes continue to shrink and never stabilize (although stripe shrinkage rate decreases with time, Movie S1). Stripe stabilization can be achieved by the retraction of the frequency gradient, kick-starting another ‘stabilizing’ genetic program that completely freezes the stripes. Such a stabilizing program might further refine the stripes and/or split them into secondary stripes. Interestingly, in the germband stage (where Tc-cad retracts continuously along with germband elongation), once a Tc-eve stripe forms, it splits into two secondary (segmental) stripes [13], whereas in the blastoderm stage, the first Tc-eve stripe does not split until Tc-cad expression completely retreats towards the posterior, at which time the second Tc-eve stripe is already formed and the third stripe is starting to emerge (Figure 1 B–D). This suggests a link between Tc-cad retraction and Tc-eve splitting. Stabilizing and refinement/splitting strategies might rely on auto- and cross-regulatory interactions between pair-rule genes or on a reaction diffusion mechanism [62] or both.
In situ hybridization was performed using DIG-labeled RNA probes and anti-DIG: : AP antibody (Roche). Signal was developed using NBT/BCIP (BM Purple, Roche), or Fast Red/HNPP (Roche). Immunocytochemistry was performed using anti-Eve (mouse monoclonal antibody 2B8, hybridoma bank, University of Iowa) as primary, and anti-mouse: : POD as secondary antibody (ABC kit, Vector). AlexaFluor 488 tyramide (Invitrogen) was used to give green fluorescent signal. All expression analyses were performed using embryos from uninjected GA-1 strain (WT) or adult GA-1 females injected with double-stranded RNA (ds RNA) of the gene of interest. dsRNA was synthesized using the T7 megascript kit (Ambion) and mixed with injection buffer (5 mM KCl, 0. 1 mM KPO4, pH 6. 8) before injection. Used dsRNA concentrations: 200 ng/µl for severe Tc-cad, 7. 5 ng/µl for mild Tc-cad, 200 ng/µl for Tc-lgs, 200 ng/µl for Tc-pan, 1 µg/µl for Tc-apc1,1 µg/µl for Tc-zen1, and 200 ng/µl; 1 µg/µl for Tc-lgs; Tc-zen double RNAi. One hour developmental windows were generated by incubating one hour egg collections at 23–24°C for the desired length of time. For 3-hour developmental windows, eggs were collected after three hours instead of one hour. The beetles were reared in whole-wheat flour supplemented with 5% dried yeast. | One of the most popular problems in development is how the anterior-posterior axis of vertebrates, arthropods and annelids is partitioned into segments. In vertebrates, and recently shown in the beetle Tribolium castaneum, segments are demarcated by means of gene expression waves that propagate from posterior to anterior as the embryo elongates. These waves are assumed to arise due to the regulation of a molecular clock by a frequency gradient. However, to date, neither a candidate nor a functional role has been identified for such a frequency gradient. Here we provide evidence that a static expression gradient of caudal regulates pair-rule oscillations during blastoderm stage in Tribolium. In such a static setup, a frequency gradient is essential to convert clock oscillations into a striped pattern. We further show that a frequency gradient might be essential even in the presence of axis elongation as a buffer against noise. Our work also provides the best evidence to date that Caudal acts as a type of morphogen gradient in the blastoderm of short-germ arthropods; however, Caudal seems to convey positional information through frequency regulation of pair-rule oscillations, rather than through threshold concentration levels in the gradient. | Abstract
Introduction
Results
Discussion
Materials and Methods | developmental biology
genetic oscillators
animal genetics
invertebrate genetics
gene regulatory networks
genetics
biology and life sciences
gene disruption
molecular genetics
computational biology
evolutionary developmental biology | 2014 | Caudal Regulates the Spatiotemporal Dynamics of Pair-Rule Waves in Tribolium | 6,394 | 297 |
Many human diseases, arising from mutations of disease susceptibility genes (genetic diseases), are also associated with viral infections (virally implicated diseases), either in a directly causal manner or by indirect associations. Here we examine whether viral perturbations of host interactome may underlie such virally implicated disease relationships. Using as models two different human viruses, Epstein-Barr virus (EBV) and human papillomavirus (HPV), we find that host targets of viral proteins reside in network proximity to products of disease susceptibility genes. Expression changes in virally implicated disease tissues and comorbidity patterns cluster significantly in the network vicinity of viral targets. The topological proximity found between cellular targets of viral proteins and disease genes was exploited to uncover a novel pathway linking HPV to Fanconi anemia.
Functional interactions between cellular targets of viral proteins and disease susceptibility genes [1], [2], [3], [4], [5] might play key roles in disease etiology. Advances in the mapping of the human interactome network, as well as in the systematic identification of gene-disease associations, provide functional data that can be used to explore fundamental connections between viral targets and disease genes. Here we formulate a local impact hypothesis, stating that diseases that can be either genetic or virally implicated can be better understood from a network perspective [6]. By this hypothesis the products of disease susceptibility genes should reside in the network vicinity of the corresponding viral targets [7], [8]. To test this hypothesis we focused on Epstein-Barr virus (EBV) and human papillomavirus (HPV) type 16, two human viruses that differ in their host tropism, genome and proteome size, and disease etiology. We find that the disease susceptibility genes of known virally implicated diseases are in the immediate network vicinity of the host proteins that are targeted by these viruses. We could identify a viral disease module for EBV and HPV, representing a subnetwork of the interactome that contains key mechanistic pathways responsible for the observed virus-disease associations. A computational prioritization procedure, joined by large-scale comorbidity and expression pattern analyses, identified new potential mechanistic disease pathways. To validate several of these pathways, HPV16 E6 and E7 oncogenes were independently expressed in primary human fibroblast (IMR90) and keratinocyte (HFK) cell populations to identify disease-associated genes whose expression levels were significantly altered in these E6/E7-expressing cell populations. We could identify a novel pathway that links HPV to a specific form of Fanconi Anemia. The systematic network-based framework we applied works to decipher the interplay between viruses and disease phenotypes.
We define as “virally implicated diseases” those diseases whose association with a particular virus is supported by peer-reviewed publications in the literature. This list includes not only diseases for which there is universally accepted consensus that a virus is causal (such as cervical cancer for HPV16 and Burkitt' s lymphoma for EBV), but also diseases which have some reproducible evidence of viral association but for which the mechanistic pathways are not worked out. There is significant and legitimate controversy and subjectivity regarding which diseases are virus-associated or virally implicated, so to avoid infusing personal bias in the selection process, we turned to several recently published authoritative review articles [1], [2], [9], [10] as well as additional literature searches. From these sources we compiled a list of 17 and 14 diseases for which a disease etiology with EBV and HPV16 has been claimed. Most of the selected virally implicated diseases (13 for EBV and 9 for HPV16) are genetic diseases in that they have been associated with mutations in at least one human gene (Table 1), as compiled in the Online Mendelian Inheritance in Man (OMIM) Morbid Map repository [11], although there are notable exceptions. Infectious mononucleosis, a disease clearly linked to EBV infection, lacks any known susceptibility genes (Table 1a). Similarly, cervical carcinoma, known to be caused by HPV infections, does not have a known genetic association (Text S1). Whenever a given disease is universally associated with viral infection and not driven by genetic changes, our approach will not yield a link between these diseases and the corresponding virus. To explore the role of macromolecular networks in virus-disease associations we collected four categories of biological connections: 1) lists of previously published experimental virus-human protein-protein [12], [13], [14] and protein-DNA interactions [2], [15], [16]; 2) a newly generated dataset of EBV-human and HPV16-human protein-protein interactions (Tables S8, S9 in Text S1), with sets (1) and (2) together defining our set of “viral targets”; 3) previously published experimental human protein-protein interactions [17], [18], [19], [20], [21], experimental and predicted human protein-DNA interactions [22], [23], and predicted human metabolic coupling interactions [24], all of which together define our “host interactome”; and 4) human gene/disease associations [11] which define a set of human genes associated with human diseases (Text S1). To test our hypothesis that genes associated with virally implicated diseases are located in the network vicinity of viral targets (Figure 1A), we measured the shortest paths, defined as the minimum number of “hops” along the links of the host interactome from viral targets to genes associated with a given virally implicated disease (Figure 1B). For either EBV or HPV the average shortest path (averaged over the number of virally implicated diseases) is significantly shorter than when virally implicated diseases were replaced with randomly sampled human diseases in OMIM (Figures S2, S3 in Text S1; P = 2. 3×10−6 for EBV and P = 7×10−7 for HPV16, based on empirical calculation). That this shortest path was less than one for both EBV (0. 667) and HPV (0. 5) indicates that viral proteins preferentially target a disease associated protein directly (hop 0) or a protein that directly interacts with a disease associated protein (hop 1). To mitigate potential investigational biases that accompany literature-curated datasets [25], we also examined the average shortest path upon removal of small-scale protein-protein and protein-DNA interactions from the host interactome, leaving only interactions derived from high-throughput investigations. The average shortest path remained significantly shorter than random (average shortest path = 1. 0; P = 4. 9×10−5 for EBV and average shortest path = 1. 0; P = 3×10−4 for HPV16, based on empirical calculation; Figure 1C, D). The shortness of the path lengths between viral targets and genes associated with virally implicated diseases is mostly due to the tendency of the viral targets being hubs, and to a lesser degree to the properties of disease genes (Text S1). The relative shortness of the paths from viral targets to disease genes validates the hypothesis that genes in the “neighborhood” of viral targets are more likely associated with virally implicated diseases, compared to genes in distant regions of the host interactome. But still, given the small world nature of the interactome, large numbers of proteins are within a few hops of the viral targets, potentially implicating hundreds of diseases for which there is no known relationship to HPV or EBV. Accordingly, a procedure is needed to identify the set of host cellular components (genes, proteins, and metabolites) that are most likely impacted by the virus, representing the network neighborhood of viral targets. Do the three kinds of interactions used to build the interactome — protein-protein, metabolic and regulatory interactions — play a comparable role in linking viral targets to virally-implicated diseases, and how deep into the interactome should one go, keeping in mind that most proteins are approximately three links from the viral proteins. To find the optimal neighborhood responsible for the phenotypic impact of a virus, we tested several “configurations” that govern the maximum hops allowed from the viral targets for each type of biological interaction. The simplest configuration includes only viral targets, while the more extended configurations capture increasing number of hops along the links of the interactome network, connecting an increasing number of proteins. The best configuration, as measured by the odds ratio of the enrichment of virally implicated diseases, defined the optimal neighborhood as the viral targets themselves and the genes regulated by them, and was the same for both viruses (Figure S4A, B in Text S1; Tables S3, S4 in Text S1). This agrees with our finding that genes associated with virally implicated diseases are themselves viral targets or are the interaction partners of viral targets (local impact hypothesis). For both viruses protein-protein and metabolic interactions in the host interactome were of secondary importance in linking the viral targets to the diseases they cause. For other viruses, however, these interactions could prove to be important. Indeed, analyses restricted to high-throughput data suggested additional relevance of host protein-protein and metabolic interactions (Figure S4C, D in Text S1; Tables S3, S4 in Text S1). In the selected optimal configuration, 9 out of the 13 virally implicated diseases for EBV were associated with genes in the neighborhood of EBV targets (Table 1a), and 7 out of 9 virally implicated diseases for HPV were associated with genes in the neighborhood of HPV16 targets (Table 1b). Both of these numbers were significantly higher than random expectation (Figure 2A, B; P = 0. 0012 for EBV and P = 0. 0005 for HPV based on empirical calculation). We therefore chose this configuration to define the network neighborhood of the viral targets, representing the viral disease modules, leaving aside the host metabolic and protein interactions. According to the local impact hypothesis, the genes regulated by viral targets should have significantly altered expression levels in virally implicated disease tissues within the viral disease modules. To test this, we collected microarray gene expression data for two representative EBV-implicated diseases, Burkitt' s lymphoma and B cell lymphoma [26], and for two HPV16-implicated diseases, cervical cancer [27] and head and neck squamous cell carcinoma [28] (Methods). We compared gene expression levels between disease tissues to control (unaffected) tissues. We defined genes with significantly altered expression levels (“differentially expressed genes”) as those whose changes in expression level between disease and normal tissues were among the top or bottom 5% of all genes (conclusions were unaltered across a wide range of cutoffs) (Table S5 in Text S1 and Text S1). In disease samples there were significantly more differentially expressed genes in the neighborhood of viral targets of either EBV or HPV16 than in the neighborhood of randomly sampled host genes that are regulated by at least one transcription factor in the TRANSFAC database (Figure 2C, D; Figures S5, S6 and Table S5 in Text S1). Given the high interconnectivity of the host interactome, the number of all potential distinct paths linking viral targets to genes (or gene products) associated with virally implicated diseases exceeds 10200 for both viruses (Text S1). Yet, the local impact hypothesis argues that only paths within the neighborhood of viral targets might play a mechanistic role in virally implicated diseases. These paths, defined as the shortest paths between the set of viral targets and genes associated with virally implicated diseases, are much fewer (20 for EBV and 24 for HPV), and could be inspected individually to determine whether they may contribute to known disease mechanisms and whether they predict potentially novel links between viruses and virally implicated diseases. Several of these paths are already informative upon disease mechanisms (highlighted in Figure 2E, F): i) EBV protein EBNA-LP has been shown to bind to RB1, which in turn regulates MYC, a human gene associated with Burkitt' s lymphoma, an EBV-implicated disease [2], [10]; ii) EBV protein EBNA2 binds to host protein RBPJ [2] which regulates Bcl-3 [29], which is in turn associated with B cell lymphoma, an EBV-implicated disease [2], [10]; and iii) HPV E6 protein interacts with p53 which regulates TNFRSF10B which is associated with head and neck squamous carcinoma, an HPV16-implicated disease [30]. The many other suggestive paths uncovered between viral targets and genes associated with virally implicated diseases (Figure 2E, F) represent candidates for focused investigations into the molecular mechanisms of these diseases. The neighborhoods of viral targets in the host interactome, along with their disease associations, represent “viral disease networks” (Figure 3A, C). The viral proteins, their viral targets, the proteins in their local neighborhood and diseases associated with all the host genes are included in the disease network. In line with the local impact hypothesis, we expect that these neighborhoods contain most cellular components that play a role in the phenotypic impact of the virus on the host. The neighborhood of randomly chosen human proteins as viral targets yields a much sparser and smaller network (Figure 3B, D), indicating that the observed viral disease networks had not emerged by chance, but instead reflect the functional adaptation of viruses to the host interactome. Randomly chosen degree-controlled viral targets also yielded random disease networks with significantly smaller connected components (Figure S8 in Text S1). The uncovered viral disease networks contain several diseases that have not been previously associated with infection by the corresponding viruses (grey squares in Figure 3A, C). These diseases arise by mutations in cellular pathways that are targeted by these viruses. Some of these diseases might arise from infection with HPV or EBV. Given the large number of such disease candidates (128 for EBV and 141 for HPV), it is important to prioritize them based on their proximity to viral targets, inferring the likelihood that the virus-induced perturbations could contribute to the particular disease phenotype. We implemented a topology-based network flow algorithm [31] that simultaneously exploits the local modularity of the interactome and the non-random placement of the disease associated components in the network. Initially, only the viral targets have non-zero scores, and the score of other proteins in the entire interactome is zero. The algorithm iteratively distributes scores to host genes based on their potential association with viral perturbation, prioritizing the genes in the neighborhood of the viral targets. Using literature-derived virally implicated diseases (Table 1) as a positive reference set, we evaluated the precision-recall performance of the prioritization for both EBV and HPV16 (Figure S9 in Text S1) and found enrichment of virally implicated diseases among the high-ranking diseases (e. g. Burkitt' s lymphoma for EBV and bladder cancer for HPV16, Tables S6, S7 in Text S1), supporting the feasibility of the prioritization procedure. To independently benchmark the prioritization of candidate diseases, we turned to relative risk measurement [24], [32], [33], which provides population-based clinical associations between candidate diseases and viral infection in patients (Text S1). Using U. S. Medicare patient medical history data [24], [34] derived from 13 million patients, we found that higher-ranked diseases in the prioritization are more often associated with viral infection, for either EBV or HPV (Figure 3E). This comorbidity analysis indicates that diseases with associated genes in the network vicinity of viral targets are strong candidates for being virally implicated. The prioritized virally-implicated disease candidates (Tables S6, S7 in Text S1) indicate, for example, that malignant neoplasms of retina and bladder, ranked in the top three by the flow algorithm regarding their potential association with HPV, have relative risk 15. 7 and 2. 7 (Table S7 in Text S1), meaning that HPV patients have 15. 7 and 2. 7 times increased chance of developing these diseases. Several diseases that ranked high in our prioritization procedure are not commonly linked to the studied viruses, but their potential viral association was supported by recent suggestive reports, such as malignant neoplasm of thyroid gland association with EBV [35] and retinoblastoma association with HPV [36], [37]. To demonstrate the value of the network-based approach to generate new biological hypotheses, we explored whether the cellular perturbations induced by expression of individual viral proteins are similar to those seen in particular disease phenotypes. We generated primary human keratinocyte (HFK) populations with stable expression of the HPV16 E6 or E7 oncoproteins and analyzed the gene expression profiles of multiple independent samples for these cells in concert with expression data from IMR90 cells expressing HPV16 E6 or E7 proteins (Methods and Text S1). Of the 104 human genes regulated by the 15 human protein targets of E6 and E7 (i. e. , those two degrees away from E6 and E7 in Figure 3C), 22 were found to be differentially expressed in E6 or E7 induced IMR90 and/or HFK cell populations (Figure 3F; Table S14 in Text S1). Of these 22 genes 15 of them were also differentially expressed in cervical carcinoma tissues evaluated previously to test the local impact hypothesis (Text S1). These 22 genes have been linked to 39 diseases in OMIM, among which only six belong to known HPV-related diseases (Table 1B). We therefore asked if any of the remaining 33 diseases might be virally implicated (Figure 3F). Illustrative of our approach is ovarian cancer, which is linked to HPV via three lines of evidence: (i) the disease has significant comorbidity with HPV associated diseases; (ii) four of the ovarian cancer associated genes in the disease network are differentially expressed in E6 or E7 induced IMR90 or HFK cell populations; (iii) three of these, FN1, BRCA1, ERBB2, are differentially expressed in ovarian carcinoma tissues (GEO dataset: GDS3592; two-tailed t-test; P<0. 05) [38]. Seven out of 39 diseases have high relative risks among HPV patients (Figure 3F), of which four are previously unknown. Among these four diseases, neoplasm of peritoneum, benign neoplasm of skin, and diseases of sebaceous glands satisfied only two of the three criteria (Text S1). Fanconi anemia, the fourth disease on the list, satisfied all three in that (i) Fanconi anemia shows high relative risk with HPV; (ii) FANCC, a gene in the disease network mutated in Fanconi anemia, is up-regulated in the E6 exogenous expression IMR90 cell data, and (iii) FANCC is significantly up-regulated in low density bone marrow cells of Fanconi anemia patients (GEO dataset: GSE16334; two-tailed t-test; P = 0. 00069) [38]. In addition, HPV16 E7 was hypothesized to induce expression of FANCD2 through an E2F-dependent pathway [39], [40], a finding that is also supported by our analysis (Text S1). Our analysis predicts a novel potential link between HPV and Fanconi anemia, through the E6→TP53→FANCC pathway, which had not been previously established. FANCC has been reported to be expressed at higher levels upon activation of TP53 [41] but since E6 targets TP53 for degradation it is unlikely that the observed upregulation of FANCC expression in E6 expressing cells is solely modulated by p53. An additional connection to Fanconi anemia may be through interaction with BRCA1 [42] (Figure 2F). In addition to a physical interaction of E6 and E7 with BRCA1, BRCA1 expression is upregulated in E6 expressing cell lines (Table S14 in Text S1). BRCA1 has been shown to have a potential role in Fanconi anemia through its role in the colocalization of FANCD2 protein [43], [44]. The clinical connection between Fanconi anemia and HPV associated tumors has been subject to debate. Not debatable is that FA patients have a much-increased risk in developing squamous cell carcinomas (SCCs) at anatomical sites infected by HPVs. Our analysis does not necessarily mean that SCCs in Fanconi patients are caused by HPV, but that they arise by similar molecular mechanisms. The well-documented interplay between E7 and FA and our discovery of a possible connection between E6, FANCC and BRCA1 support this hypothesis. Moreover, we observe a relative risk of 3. 7 among female HPV patients (mostly cervical cancer patients) toward Fanconi anemia using the US-wide Medicare data, which further supports the identified molecular level relationship between Fanconi anemia and HPV (Methods and Text S1).
Given the large number of functional interactions present in human cells and the many possible paths among cellular components, uncovering the precise impact of a virus upon the host interactome is an enormously complicated task. Here we provide evidence that a large proportion of the effect of a virus can be accounted for locally in the network space, which allowed us to develop and test a general methodology designed to elucidate the consequences of viral impacts on the host interactome network, and to prioritize candidate diseases for potential viral implications. A predictive methodology should ideally take into account cell tropism. Tissue-specific gene expression data can be merged with our analysis (Text S1). We used tissue-specific expression data from BIOGPS [45] to narrow down the number of genes and their associated diseases from the diseasome map. If a gene in the neighborhood of the viral targets is not expressed or is not present in the tissue of interest, we removed the gene from the network. In this way, we obtain a tissue-specific viral network. By applying tissue specificity, the number of associated diseases for EBV was reduced from 128 to 89, and for HPV from 141 to 105, without losing any of the virally-implicated diseases (Text S1 for analysis details; Tables S15, S16 in Text S1 for the list of genes and diseases). The strategy developed here is not unique to EBV and HPV16. Although the strategy should work better for carcinogenic pathogens, given how well-studied proteins involved in cancer are, it is equally applicable to any pathogen for which protein interactions between the pathogen and the host proteome have been mapped. While still limited by the incompleteness of genome- and proteome-scale datasets [19], the usefulness of the method is likely to grow alongside the ongoing expansion of high-throughput functional genomics databases and gene-disease associations.
Yeast two-hybrid screens (Y2H) between EBV and HPV16 viral proteins and ∼12,200 human proteins encoded by a library of full length human open reading frame (ORFs) clones in Human ORFeome v3. 1 [46], [47] encompassing ∼10,200 human genes were carried out as before [17], [48]. The EBV-human library Y2H screen tested 86 out of 89 EBV proteins as fusions to the DNA binding domain of Gal4 (Gal4-DB) against ORFeome v3. 1 proteins fused to the activation domain of Gal4 (Gal4-AD), while the HPV-human library Y2H screen testing HPV16 proteins E4, E5, E6 and E7 was carried out in reciprocal fashion with HPV proteins as both Gal4-DB and Gal4-AD fusions against the corresponding human Gal4-AD and Gal4-DB fusions, respectively. Raw data of the gene expression datasets used (GSE2350, GSE2392 and GSE15156) was obtained from Gene Expression Omnibus (GEO) [49], normalized and log-transformed by RMA algorithm [50], and expression changes were calculated as the ratio of expression levels between virus-infected tissues and normal tissues. To obtain the disease associated genes that are differentially expressed in viral protein induced cell populations, HPV16 E6 and E7 oncogenes were independently transfected into primary human fibroblast (IMR90) and keratinocyte (HFK) cell populations. Affymetrix Human Gene1. 0 ST and Human Genome U133 Plus 2. 0 arrays, respectively, were used to measure gene expression profiles for five or more replicate samples in each of the cell types. Array data were normalized by RMA, batch effects were removed using ComBat, and the limma package in R/Bioconductor was used to identify differential expression. Relative risk (RR) was calculated as the ratio between the observed co-occurrence and probabilistically-inferred (assuming independence) co-occurrence of two diseases, based on the patient medical history data from United States (U. S.) Medicare, which contains the clinical diagnosis record of each hospital visit (in ICD-9 codes) of 13 million U. S. patients at age 65 or older [33]. Patients with viral infections were defined with the following diagnostic codes in U. S. Medicare database: 200 (B cell lymphoma) or 147 (nasopharyngeal carcinoma) for EBV infections; 078. 1,079. 4,180 or 795. 0 for HPV infections. The statistical significance of the average shortest path between viral targets and genes associated with a given virally implicated disease was calculated by randomly sampling human diseases from OMIM (full table of disease in Dataset S2). The number of virally implicated diseases associated with the proteins in the neighborhoods of random host targets was calculated by picking random proteins from the interactome space (n = 7,832). For both measurements, P values were calculated based on empirical data with 10,000 random configurations. For the analysis of GEO microarray data we used two-tailed t-test statistics. | Many “virally implicated human diseases” - diseases for which there is scientific consensus of viral involvement - are associated with genetic alterations in particular disease susceptibility genes. We proposed and demonstrated that for two human viruses, Epstein-Barr virus and human papillomavirus, topological proximity should exist between host targets of viruses and genes associated with virally implicated diseases on host interactome networks (local impact hypothesis). For representative EBV- and HPV16- implicated diseases, genes in the neighborhood of viral targets in the host interactome have significantly shifted expression levels in virally implicated disease tissues, in line with the local impact hypothesis. The viral neighborhoods in the host interactome, along with their disease associations, defined as “viral disease networks”, contain connections known to be informative upon disease mechanisms as well as diseases whose associations with viruses are not yet known. We prioritized these diseases for their candidacy as potential virally implicated diseases based on network topology, and benchmarked this prioritization of candidate diseases using relative risk measurement which depicts population-based clinical associations between candidate diseases and viral infection. Exogenous expression of HPV viral proteins in a human cell line offered evidence for a novel disease pathway that links HPV to Fanconi anemia. | Abstract
Introduction
Results
Discussion
Methods | microarrays
systems biology
biology
computational biology
metabolic networks
genetics and genomics | 2012 | Viral Perturbations of Host Networks Reflect Disease Etiology | 5,926 | 274 |
Candida parapsilosis and Candida albicans are human fungal pathogens that belong to the CTG clade in the Saccharomycotina. In contrast to C. albicans, relatively little is known about the virulence properties of C. parapsilosis, a pathogen particularly associated with infections of premature neonates. We describe here the construction of C. parapsilosis strains carrying double allele deletions of 100 transcription factors, protein kinases and species-specific genes. Two independent deletions were constructed for each target gene. Growth in >40 conditions was tested, including carbon source, temperature, and the presence of antifungal drugs. The phenotypes were compared to C. albicans strains with deletions of orthologous transcription factors. We found that many phenotypes are shared between the two species, such as the role of Upc2 as a regulator of azole resistance, and of CAP1 in the oxidative stress response. Others are unique to one species. For example, Cph2 plays a role in the hypoxic response in C. parapsilosis but not in C. albicans. We found extensive divergence between the biofilm regulators of the two species. We identified seven transcription factors and one protein kinase that are required for biofilm development in C. parapsilosis. Only three (Efg1, Bcr1 and Ace2) have similar effects on C. albicans biofilms, whereas Cph2, Czf1, Gzf3 and Ume6 have major roles in C. parapsilosis only. Two transcription factors (Brg1 and Tec1) with well-characterized roles in biofilm formation in C. albicans do not have the same function in C. parapsilosis. We also compared the transcription profile of C. parapsilosis and C. albicans biofilms. Our analysis suggests the processes shared between the two species are predominantly metabolic, and that Cph2 and Bcr1 are major biofilm regulators in C. parapsilosis.
More than 300 Candida species have been described to date [1]. Although all Candida species are Ascomycetes (belonging to the Saccharomycetales), they are paraphyletic, and do not share a recent common ancestor [2]. As a result, they have few shared characteristics. The term “Candida” suggests that they are asexual species, but sexual or cryptic sexual cycles are increasingly being identified [3], [4], [5], [6]. Most well studied Candida species belong to the monophyletic CTG clade, where the codon CTG is translated as serine rather than leucine [2], [7]. These include the major human fungal pathogens Candida albicans, Candida parapsilosis and Candida tropicalis [8], [9]. Whereas C. albicans is still the most common cause of candidiasis, C. parapsilosis and the non-CTG clade species Candida glabrata are increasing in frequency [8], [9], [10]. Properties of C. albicans associated with the ability to cause disease have been well characterized, and include growth in yeast and hyphal forms, epigenetic switching from white to opaque cells, secretion of hydrolases, and adhesion and biofilm development (reviewed in [11]). While some of these properties are likely to be shared with other CTG-clade species, many are species or lineage specific. For example, only C. albicans and its close relative Candida dubliniensis can grow in truly hyphal forms, and white-opaque switching and the associated parasexual cycle have only been described in C. albicans, C. dubliniensis and C. tropicalis [12], [13], [14]. C. parapsilosis, a major cause of infection in premature neonates [15], does not appear to have a sexual cycle [16], [17] and does not undergo white-opaque switching [18]. Unlike other Candida species, C. parapsilosis is often isolated from the hands of health care workers and has been responsible for causing outbreaks of infection [19], [20], [21], [22], [23], [24]. C. parapsilosis is responsible for approximately 20% of Candida infections particularly in infants less than 1 year old [25], [26]. One of the major factors of Candida species associated with pathogenicity is their ability to grow as biofilms on implanted medical devices [27]. Biofilms are composed of communities of microorganisms associated with a surface and embedded in an extracellular matrix, and are believed to be the major growth form of microorganisms in nature [28]. Biofilms are extremely refractory to antimicrobial therapy and treatment usually involves removal of the infected device. Biofilm formation in C. albicans has been well characterized and occurs in several stages (reviewed in [29], [30]). The first step involves yeast cells adhering to a substrate surface. This is followed by a period of cellular growth, or biofilm initiation. During the maturation stage, hyphae are produced and cells become encased in an extracellular matrix (ECM). The final stage is dispersal, when yeast cells break away from the biofilm structure and disseminate around the body to seed new sites of infection [31], [32]. Although many Candida species form biofilms, the structures are highly variable [28], [33]. C. albicans biofilms consist of a compact basal layer of yeast cells and a thicker less compact hyphal layer all surrounded by an ECM composed mainly of carbohydrate [34]. In contrast C. parapsilosis does not make true hyphae, and biofilms are composed of yeast and pseudohyphal cells only [27], [35], [36]. The ability of C. parapsilosis to produce biofilm is also highly strain dependent [28], [33]. Many of the key regulators of biofilm formation in C. albicans have been identified (reviewed in [37]). Hyphal formation is a pivotal step, and mutants blocked in filamentation are often impaired in biofilm development [38]. Nobile et al [39] identified a network of six transcription factors (BCR1, EFG1, TEC1, ROB1, NDT80 and BRG1) that play a major role in regulating C. albicans biofilm growth. In addition, Finkel et al [40] identified 30 transcription factors required for adhesion, some of which (such as Bcr1) are also necessary for mature biofilm development. We have previously shown that C. parapsilosis orthologs of BCR1 and EFG1 are required for biofilm formation in this species [18], [35]. However, even though the function of the transcription factors is at least partially conserved, many of the gene targets are different, and some conserved targets of Bcr1 have different functions [41]. For example the CFEM family of cell wall proteins is required for biofilm development in C. albicans, but not in C. parapsilosis [41]. We report here the construction of the first large-scale gene deletion collection in C. parapsilosis, targeting 100 genes representing transcription factors, protein kinases and species-specific genes. We carry out a detailed comparison of C. parapsilosis and C. albicans phenotypes, particularly in relation to biofilm development. We find that overall, the molecular function of orthologous genes is generally conserved between the two species. However, there are also important differences. BRG1 and TEC1, transcription factors required for biofilm development in C. albicans, do not have the same role in C. parapsilosis. CZF1, UME6, CPH2 and GZF3 are regulators of biofilm development in C. parapsilosis only.
Most of the available gene disruption collections in C. albicans target transcription factors or protein kinases [42], [43], [44], [45]. We therefore selected similar genes from C. parapsilosis (Figure 1, Tables S1, S2). In total, we chose genes encoding 73 transcription factors, 16 protein kinases, 1 putative RNA-binding protein, 1 putative tRNA-methyl transferase, 6 genes that are apparently unique to C. parapsilosis, and members of the CFEM family of transmembrane proteins [41], [46], [47]. We selectively deleted entire open reading frames rather than generating random insertions, to facilitate downstream analysis. The C. parapsilosis genome is diploid and therefore requires two rounds of gene disruption to create a homozygous mutant. Gene disruptions in C. parapsilosis have previously been carried out using a recyclable SAT1 (nourseothricin resistance) cassette [35], [36]. Although this approach is successful, it is slow and very inefficient. Instead, we adapted a fusion PCR method previously developed for gene deletion in C. albicans, in which each allele is replaced with a heterologous selectable marker [45] (Figure 1A, Figure S1). Deletions were constructed in the C. parapsilosis type strain, CLIB214. Auxotrophic mutations in LEU2 and HIS1 were first generated using a SAT1 cassette, yielding strain CPL2H1 (see methods) (Figure S1, Table S1) [35]. Candidate genes were then deleted by replacing one allele with HIS1 from C. dubliniensis, and one with LEU2 from C. maltosa (Figure 1A, Figure S1). All mutant strains were confirmed by PCR using primers inside CdHIS1 or CmLEU2 and a primer outside of the integration sites at both the 5′ and 3′ end of the gene. Deletion of the target open reading frame was also confirmed using PCR. A control strain (CPRI) was created by integrating CdHIS1 and CmLEU2 at the site of the original HIS1 alleles (Figure S1). Both CLIB214 and CPRI strains were used as controls for the majority of experiments. Two independent homozygous deletion mutants were generated for each targeted gene, which increases the probability that an observed phenotype in both strains is a result of the gene deletion, and not from a secondary effect. A unique 20 base pair sequence tag (barcode) was also incorporated into each mutant strain, which will facilitate future competition experiments. In total, 200 barcoded deletion strains were constructed (Table S1). Growth of the deletion collection was determined in 42 different conditions, designed to identify nutritional, cell wall, osmotic and oxidative stress phenotypes, and response to antifungal drugs (Table S2). Some strains had severe growth defects, and were not included in the phenotype analysis (Table S2). For the remainder, each independent strain was grown in YPD in 96 well plates (two replicates per candidate gene), and then replica-plated using a 48-pin replicator to selective media. YPD was used as the base media except where the effect of different carbon sources was tested. Several drugs and chemicals were tested at a range of concentrations (Table S2). Growth was scored using a simple scoring system (−4 to +1, where 0 is the same as the control strains). The scores were averaged between sister strains and replicate screens and were converted to heatmaps (Figure 1B, C). Of the 73 transcription factor deletions constructed in C. parapsilosis, 64 orthologous deletions were available from a large scale screen in C. albicans [43], and another (UME6) was obtained from Banerjee et al [48]. Growth of the C. albicans strains was monitored under similar conditions as for C. parapsilosis except that lower concentrations of ketoconazole, fluconazole and caspofungin and higher concentrations of CdCl2 were used, and utilization of heme was not tested (Table S2). Nine transcription factor deletions are available in C. parapsilosis only. The phenotypes of these, and of the C. parapsilosis protein kinase deletions, are described in the supporting information (Text S1, Table S2). Figure 1C shows a comparison of the phenotypic profiles of orthologous deletions in the two species. Thirty-five gene deletions are not shown because they have no, or little, effect on growth in any condition tested for either species (Table S2). Changes in colony morphology were not recorded. Another four strains (deletions of ACE2, NRG1, SSN6 and TUP1) were removed because the phenotypes are difficult to score in C. albicans or in both species, mostly because they significantly affect filamentation. Deleting RBF1 and RPN4 in C. albicans and NDT80 in C. parapsilosis results in dramatic reduction in growth; the NDT80 deletion is therefore not shown, and the RBF1 and RPN4 deletions are included with the C. parapsilosis-only data (Figure 1B). Many phenotypes previously described in C. albicans are shared in C. parapsilosis. These include the role of CAP1 in the oxidative stress response (sensitivity to cadmium chloride [49]), enhanced sensitivity of CSR1 deletion to metal chelators (EDTA; [43]) and the role of HAP2, HAP5, HAP43 and SEF1 in regulating the response to iron (deletions have reduced growth to low iron, resulting from addition of the iron chelator BPS [50], [51]), of RIM101 as a regulator of the response to alkaline conditions [42], and UPC2, which determines sensitivity to azole drugs [52], [53]. The function of many of these regulators is conserved across a wide evolutionary distance, at least since the common ancestor with S. cerevisiae [54], [55], [56], [57], [58], [59], [60]. Regulation of the iron and copper response is similar in both species though there are some subtle differences (Text S1). Some phenotypes conserved between the two species have not previously been reported. These include the sensitivity of the upc2 deletions to the presence of the iron chelator, BPS (Figure 1D). Upc2 is also required for growth on xylose as the main carbon source in both C. albicans and C. parapsilosis (Table S2). However, despite the overall similarity, there are also significant differences between the two species. Several deletion strains (e. g. BCR1, CPH2, GIS2, ISW2, MIG1, SEF2, STP4) have no shared phenotypes. Some gene deletions have pleiotropic effects in C. parapsilosis; for example, deleting SEF1 results in reduced growth in many conditions, whereas the equivalent deletion in C. albicans results in much fewer phenotypes (Table S2). One of the most obvious differences between C. albicans and C. parapsilosis occurs during growth in low oxygen (hypoxic) conditions (Table S2, Figure 1C, E). Deleting UPC2 confers sensitivity to hypoxia in both, as previously reported [43], [53], [61], [62]. No other C. albicans gene deletion tested affects hypoxic growth. However, in C. parapsilosis, deleting CPH2 also reduces growth in hypoxia, and growth is restored when the wildtype gene is reintroduced (Figure 1E). In addition, we have previously shown that expression of CpCPH2 is increased during growth in hypoxia [63]. However, deleting CpCPH2 does not affect sensitivity to azole drugs (Figure 1E). We have previously shown that in C. parapsilosis, similar to C. albicans, members of the CFEM family are important for heme utilization [41], [64], [65]. Whereas C. albicans has five family members, in C. parapsilosis the family has expanded to seven (CFEM1-7). Here, we deleted CFEM1-CFEM4 together, CFEM5 and CFEM6 together, and CFEM7 alone (Table S2). Similar to previous reports, strains missing CFEM1-CFEM4 or CFEM5-CFEM6 are unable to use heme as a sole source of iron (Figure 1B, Table S2, [41]). However, deleting CFEM7 alone had no effect (Table S2). We also find that deleting CTH1 (CPAR2_407950) renders cells sensitive to iron, and unable to use hemin as a sole source of iron (Figure 1B). Expression of CpCTH1 is induced in low iron conditions [41]. CpCTH1 is an ortholog of both members of the CTH1/CTH2 gene pair in S. cerevisiae, proteins that bind to RNA transcripts from iron metabolic genes, targeting them for degradation [66], [67]. Our results suggest that CTH1 forms part of the iron regulatory pathway in C. parapsilosis. The effect of the C. parapsilosis gene deletions on biofilm development on polystyrene surfaces was determined visually using crystal violet staining, and by measurement of biomass (dry weight). Several deletion strains (ADA2, MSS2, VPS34, NDT80 and YCK2) exhibited growth defects on YPD and were not included in the biofilm screen. Of the 95 unique deletions tested, eight had obvious visual defects and significantly reduced dry weight formation, including seven transcription factors (EFG1, CZF1, GZF3, UME6, CPH2, BCR1 and ACE2) and one protein kinase (MKC1) (Figure 2A, B). Some other deletion strains had minor effects on biofilm growth, but only these eight had significant and reproducible reductions in biofilm mass. Reintroducing the intact genes restored biofilm growth (Figure 2C; restoring BCR1 and EFG1 have been described previously [18], [35]). These mutants displayed no significant defects in growth in liquid culture in biofilm media, except that the ACE2 deletion has a cell-separation defect [68]. A recent study identified a network of six transcription factors (BCR1, EFG1, TEC1, NDT80, BRG1 and ROB1) that regulate biofilm formation in C. albicans [39]. In a separate study, ACE2 was shown to be required for adhesion and subsequent biofilm development in the same species [40]. To determine the overlap between the networks regulating biofilm growth in C. albicans and C. parapsilosis, we directly compared the effect of deleting the orthologous transcription factors in the two species (Figure 2D, E). Both species were grown in conditions that maximize biofilm development [35], [38], [41]. We confirmed that deleting BCR1, EFG1 and ACE2 reduces biofilm development in C. albicans, similar to C. parapsilosis (Figure 2D), whereas deleting CZF1, GZF3, UME6, or CPH2 has little effect on C. albicans biofilm growth in our assay (Figure 2D). From the remaining genes in the C. albicans biofilm regulatory network, orthologs of TEC1, NDT80, and BRG1 were not identified in our large-scale screen of biofilm-defective mutants in C. parapsilosis (Figure 2A). We confirmed that deleting TEC1 or BRG1 did not dramatically reduce C. parapsilosis biofilm formation in follow-up tests (Figure 2E). However, we could not determine the role of the NDT80 ortholog, because unlike in C. albicans, deleting NDT80 in C. parapsilosis results in a significant growth defect. There is no ortholog of the final member of the C. albicans network, ROB1, in the C. parapsilosis genome [39], [47]. We used confocal laser scanning microscopy (CLSM) to visualize the morphology and structure of C. parapsilosis biofilms growing on the surface of Thermanox slides and stained with concanavalin A [63] (Figure 3). The control strains produce a compact biofilm consisting of layers of yeast and pseudohyphal cells. The biofilm layer is thinner than previously reported (10–20 µm, [35]), but it is reproducible. Deleting CPH2, UME6, EFG1 and BCR1 in C. parapsilosis results in biofilms with very few layers that are consistently thinner than those produced by the control strains. Biofilms generated by CPH2, UME6 and CZF1 deletions are mainly composed of yeast cells, with few pseudohyphal cells present, whereas the BCR1 mutant produces a thin biofilm of both yeast and pseudohyphal cells. The ACE2 deletion generates patchy biofilm, consisting of clumps of pseudohyphal cells across the surface of the slide, probably due to the cell separation defect caused by this mutation [68]. Biofilms produced by the GZF3 and MKC1 deletion strains have both yeast and pseudohyphal cells present in several layers; however the biofilm produced is not as compact as the wild type (Figure 3). We have previously shown that deleting EFG1 increases morphological switching between “wrinkled” and “smooth” colonies, both of which have reduced biofilm development [18]. We did not differentiate between different colony morphologies in the assay presented here. However, CPH2, UME6 and CZF1 deletion strains all have reduced colony wrinkling compared to wildtype (not shown). Biofilm development in vivo is substantially different from the in vitro models used here. In particular, biofilms formed in catheters undergo stress from blood flow. We therefore used an established rat central venous catheter (CVC) model of infection [69] to test the effect of deleting regulatory genes in C. parapsilosis (Figure 4). We compared the deletion constructs to the control CPRI strain, because biofilm development by the wildtype strain was variable in this model. C. parapsilosis biofilms in vivo consist of yeast cells, matrix and host cells. We have previously shown that deleting BCR1 greatly reduces biofilm formation in vivo, whereas deleting EFG1 has a more minor effect [18], [41]. Here we show that deleting UME6 and CZF1 also greatly reduce biofilm formation (Figure 4). The biofilm from the UME6 deletion consists of a single layer of yeast cells with little obvious matrix, whereas the CZF1 deletion produces little obvious biofilm. Deleting ACE2 results in a clumpy biofilm. In contrast to the in vitro assay, deleting CPH2, GZF3 and MKC1 has little effect on biofilm growth in vivo. We previously used microarray analysis designed from a partial genome sequence to determine the transcriptional profile of C. parapsilosis biofilms growing under flow conditions in a fermenter [63]. In general, we found that the transcriptional response of biofilms is similar to that of cells growing in hypoxic conditions, and is associated with upregulation of fatty acid metabolism genes. Here, we used RNA-seq analysis to characterize the transcription profile of cells growing in static conditions, the same as those used to identify biofilm regulators. Compared to planktonic conditions, 777 genes are up-regulated and 662 genes are down-regulated in biofilms (log2FC+/− 1. 5, adjusted p-value <0. 001) (Table S3). Upregulated genes are enriched for processes associated with lipid/fatty acid oxidation and transmembrane transport, whereas downregulated genes are enriched in translation, macromolecule and amino acid biosynthetic processes and cellular component assembly (Table S4). We used Gene Set Enrichment Analysis (GSEA) to identify gene categories that are over-represented in the biofilm transcriptome. GSEA allows the identification of statistically significant overlaps between a ranked gene list (e. g. the C. parapsilosis biofilm transcriptome), and gene lists (or gene sets) identified in other experimental analyses. We used a collection of 8,852 gene sets derived from microarray and ChIP (chromatin immunoprecitation) experiments from C. albicans and from protein-protein interactions from S. cerevisiae that were collected and described by Sellam et al [70]. We extracted a set of C. albicans orthologs of our ranked list of genes differentially expressed in C. parapsilosis biofilms, and looked for similarities between this gene list and the gene sets described by Sellam et al [70]. The network of similar gene sets was visualized using Cytoscape [71] (Figure 5, Figure S2). In these figures, nodes represent gene sets, and edges connect nodes sharing a significant number of genes. Clustering algorithms in Cytoscape group highly interconnected and similar gene sets together. We have colored nodes that included genes upregulated in C. parapsilosis biofilms in red, and those that include downregulated genes in blue. Figure 5 shows that upregulated genes share similarities with gene sets associated with transport (including carboxylic acid and drug transport) in C. albicans. There is also a significant overlap with the C. albicans biofilm regulatory network described by Nobile et al [39]. The transport and biofilm networks are connected via Ndt80 (Figure 5). Downregulated genes are enriched for processes associated with translation and the ribosome. The similarity between the C. parapsilosis biofilm transcriptome and the C. albicans biofilm regulatory network prompted us to directly compare the biofilm transcriptional profiles of the two species, using the C. albicans data from Nobile et al [39] (Figure 6A). In C. albicans 785 genes are up-regulated and 300 genes down-regulated in biofilm compared to planktonic cells. There are 192 genes upregulated in biofilms in both species (Figure 6A). These are enriched in oxidoreductases and in pathways associated with carboxylic acid metabolism. Interestingly, genes upregulated in C. albicans only (and not in C. parapsilosis) are enriched in processes associated with biofilm formation and adhesion. Many are transcription factors, some of which we have shown have no role in biofilm development in C. parapsilosis (BRG1, WAR1 [40], CRZ2 [40] and ZNC1 [40]), and others that we have not tested, but which are not differentially expressed in biofilms e. g. GCN4 [72]. Genes upregulated only in C. parapsilosis are enriched in pathways associated with transmembrane and drug transport. This category includes a large number of genes with no annotation, suggesting that their function may be specific to C. parapsilosis. One of the difficulties comparing gene expression profiles between species is the reliance on the identification of orthologs present in both. This may underestimate the importance of genes that are found in one species only. In addition, it is very difficult to evaluate the roles of gene families that have different numbers or members in different Candida species [73]. We therefore categorized the gene families in C. albicans and C. parapsilosis using Markov Clustering (MCL; [74]) and looked for evidence of family enrichment among differentially expressed genes in biofilms (again using the data from [39] for C. albicans). Figure 6B shows the gene families ordered by average gene expression for both species (from highest to lowest expression; full data in Table S5). The CFEM family, the ATO family of putative ammonia transporters [75], the OPT family of oligopeptide transporters [76] and the TES family of acyl CoA-thioesterases are among the most enriched among upregulated genes in both species (Figure 6, Table S5). General amino acid permeases (GAP family) and secreted aspartyl proteases (SAP family) are also among the most highly expressed in C. parapsilosis, and are enriched among up-regulated genes in C. albicans (Figure 6B). The most downregulated genes enriched in both species include a family involved in Nucleic Acid Metabolic Processes (NAMP) (including TUP1, a major hyphal regulator in C. albicans that is required for biofilm growth [77]) and histone genes (HHT family). The PGA30/RHD3 family has the highest average expression in C. parapsilosis; these encode GPI-anchored proteins predicted to be localized to the cell wall [78]. To identify the genes that are regulated in C. parapsilosis biofilms, we used RNA-seq to compare the transcriptional profiles of biofilms from strains deleted for EFG1, CZF1, UME6, CPH2, BCR1 and ACE2 to biofilms from wildtype strains (Figure 7). We were unable to isolate RNA of sufficient quality from strains deleted for GZF3 or MKC1, suggesting that there may be changes in the extracellular matrix or cell wall that we cannot directly observe. Only seven genes are downregulated in biofilms of all deletion strains and upregulated in the wildtype (ARO10, ATO1, PUT4, STE18, HGT17, CPAR2_803700, CPAR2_805760). We have not yet characterized the roles of these genes in biofilm development. Processes associated with transmembrane transport, organic acid catabolism and carboxylic acid metabolism are upregulated in most of the deletion biofilms, compared to wildtype (Figure 7). The transcriptional profiles of biofilms from the BCR1 and CPH2 deletion cluster close together (Figure 7), suggesting that these transcription factors regulate a core set of genes. We also used GSEA to compare the gene sets regulated by some of the transcription factors with the biofilm network described in Figure 5. Figure 8 shows that there is a high degree of overlap between the gene sets enriched in the BCR1 and CPH2 deletion and in the wildtype biofilm/planktonic comparison. Gene sets associated with carboxylic acid transport and with carbohydrate transport are downregulated in both. The targets of Efg1, a known biofilm regulator in C. parapsilosis [18] have the smallest overlap with the biofilm network (Figure 8C). Again, the overlap includes gene sets associated with carboxylic acid transport.
The phenotype screen revealed a high conservation in the regulation of different biological processes between C. albicans and C. parapsilosis. Overall, 15 of the 23 transcription factors in Figure 1C share at least one major phenotype in C. albicans and C. parapsilosis. Some shared phenotypes, such as role of UPC2 in the response to low iron conditions, have not previously been highlighted. However, Homann et al [43] noticed that deleting UPC2 in C. albicans resulted in the accumulation of a pink color after several days of growth on BPS, possibly signaling formation of a ferrous-BPS complex. Deleting UPC2 also results in a growth defect of Yarrowia lipolytica in low iron conditions [79]. It is therefore likely that this transcription factor is involved in regulating the response to iron in several fungal species. Despite the overall similarity observed in the phenotypic screen there are some significant differences, including the role of CPH2 in the hypoxic response of C. parapsilosis. In C. albicans, CPH2 is a regulator of hyphal growth with no known hypoxic role [80], [81]. It has a basic helix-loop-helix (bHLH) DNA-binding domain, and is probably a remnant of the Sterol Regulatory Binding Proteins (SREBPs) that regulate sterol synthesis and the hypoxic response in filamentous fungi and in Schizosaccharomyces pombe; species that have no recognizable Upc2 [79], [82], [83]. Most species in the Saccharomycotina (including Saccharomyces and Candida species) have lost domains from their SREBP homologs, and the remaining remnants have no known role in regulating sterol synthesis, or in the hypoxic response [79], [83]. Y. lipolytica, an outgroup of the Saccharomycotina, is a notable exception in that it retains a full length SREBP, and has gained a Upc2 ortholog [79]. However, even in this species Upc2 is the main regulator of sterol synthesis, although SREBP has retained some role in regulating the hypoxic response, particularly in the control of filamentation [79]. In C. parapsilosis, deleting CPH2 does not affect sensitivity to ketoconazole (Figure 1E), indicating that its role in hypoxia is unlikely to be related to regulating expression of ergosterol genes. Elucidating the role of CPH2 in the hypoxic response of C. parapsilosis will require substantial further investigation. However it appears that Cph2 and SREBPs have retained a previously unsuspected role in regulating the hypoxic response in the Saccharomycotina. We identified seven transcription factors and one protein kinase that are important regulators of C. parapsilosis biofilm development in vitro (Figure 2). Deleting CPH2, GZF3 and the protein kinase MKC1 has no effect on biofilm development in the in vivo rat catheter model. This suggests that biofilm development may be context dependent. A similar phenomenon has been described in C. albicans; for example, CaBRG1 is required for biofilm formation in vitro but not in vivo [39], and deleting CaRHR2 reduces biofilm formation in the central venous catheter model but not in an oral pharyngeal model [84]. However, we note that in C. parapsilosis the GZF3 and MKC1 deletions strains have minor effects on the structure of biofilms on Thermanox slides (Figure 3), suggesting that their role as biofilm regulators may be restricted to growth in specific conditions. Two of the C. parapsilosis biofilm regulators (BCR1 and EFG1) are conserved with the well-characterized C. albicans biofilm circuit [39], and a third (ACE2) is also likely to regulate biofilm development in both species (Figure 9, [40], [68]). Five (CPH2, UME6, CZF1, GZF3 and MKC1) appear to be unique to C. parapsilosis. We previously showed that the role of Bcr1 as a regulator of biofilm development in C. parapsilosis shares some similarities with that of its ortholog in C. albicans [85]. However, there are significant differences in the two species. For example, some of the conserved targets (e. g. the CFEM family) are required for biofilm development in C. albicans and not in C. parapsilosis [41]. The role of Bcr1 is also strain dependent. In C. parapsilosis, strains which make relatively low levels of biofilm (like CLIB214, the isolate used throughout this study) are Bcr1-dependent, while those that generate high levels of biofilm are not [86]. In C. albicans, “sexual” or “pheromone-stimulated” biofilms are made by cells that are homozygous at the mating MTL locus, and they are distinguished from the more general pathogenic biofilms. There is some evidence that both kinds of biofilms require Bcr1 [87], whereas other studies suggest that Bcr1-dependent expression of CFEM genes is required for drug resistance in pathogenic biofilms only [88], [89]. “Sexual” biofilms have not been described in C. parapsilosis, where most (and possibly all) isolates are of a single mating type [17], [90]. There are therefore aspects of the roles of Bcr1 in both species than remain to be elucidated. Like Bcr1, Efg1 also regulates biofilm development in both C. albicans and C. parapsilosis [18], [39], [87], [91], [92]. In C. albicans Efg1 plays a central role in networks regulating high frequency epigenetic switching from white-to-opaque cells [93], and in filamentation of both cell types [12]. In addition, it is required for virulence [94] and for drug resistance [95]. In C. parapsilosis, Efg1 also regulates a high frequency switching system [18]. In both species, Efg1 directly bind to the promoters of a large number of transcription factors [18], [39]. Many of the promoters are exceptionally long, suggesting that they are regulated by several transcription factors [18], [39]. It is therefore likely that Efg1 is a member of several regulatory circuits in both species. Although Ace2 was not identified as a component of the C. albicans biofilm regulatory circuit by Nobile et al [39], we and others have shown that it is required for biofilm development in this species [40], [68], which we confirmed in the current assay (Figure 2D). Ace2 regulates expression of genes during late M/early G1 phase of the cell cycle in S. cerevisiae and C. albicans [68], [96], [97]. The transcription factor is part of the RAM network (Regulation of Ace2 and Morphogenesis) that controls exit from mitosis in many fungi (reviewed in [98]). Ace2 regulates adherence of C. albicans, leading to sparse biofilms, and is target of Snf5, a chromatin regulator [40]. Deleting CpACE2 results in a cell separation defect, and expression of many targets shared with CaACE2 is reduced (e. g. CHT1, CHT3, SCW11) (Table S3). The function of Ace2 in regulating biofilm development in Candida species is therefore likely to be conserved. Deleting the transcription factors UME6, CPH2, CZF1 and GZF3 reduces biofilm development in C. parapsilosis only (Figure 2A, D). We did not test the effect of deleting the MKC1 kinase ortholog in C. albicans. It remains possible that the orthologous transcription factors may play a more minor role in biofilm growth of C. albicans. For example, overexpressing UME6 enhances biofilm development in both species [48], [99]. However, we detected very little reduction in C. albicans biofilms formed by ume6 deletion strains (Figure 2D, and two deletions in a different genetic background that are not shown), and it is clear that the phenotype of the UME6 deletion in C. parapsilosis is considerably more severe. In S. cerevisiae, UME6 is an activator of gene expression during early meiosis, whereas in C. albicans, the ortholog regulates filamentation. Expression of UME6 is also increased during pseudohyphal growth in C. parapsilosis [99]. The CPH2 regulon in C. albicans is not known, but an early microarray study suggests that CaCPH2 regulates expression of filamentation genes, many of which are also targets of Efg1 [100]. Other regulators of filamentation, such as NRG1, also control biofilm formation in both C. albicans and C. parapsilosis [48]. It is likely that filamentation is more important for C. albicans biofilms; although both species can grow as filaments, only C. albicans make true hyphae, and many mutants locked in the yeast phase make poor biofilms [101], [102]. Ume6 and Cph2 may therefore have functions separate to filamentation that are important for biofilm formation in C. parapsilosis. Although deleting CZF1 does not reduce mature biofilm development in C. albicans, the transcription factor is required for early stage adherence [40]. The functional targets are not known [40]. In C. parapsilosis, CZF1 regulates expression of transporters (Table S3); their role in biofilm development remains to be elucidated. GZF3 encodes a GATA-type transcription factor that is induced during oxidative stress in C. albicans [103]. Expression is induced in biofilms, but there is no evidence that it plays a major role in C. albicans biofilm development [39]. We used RNA-seq and network analysis to characterize the transcriptome of C. parapsilosis biofilms. Gene sets enriched in genes downregulated in biofilms are associated with translation and with ribosome function, which probably reflects reduced growth in biofilms compared to planktonic cells (Figure 5). Genes up-regulated in C. parapsilosis biofilms are enriched for pathways associated with transport (including carboxylic acid transport), and are also enriched in gene sets that are upregulated in C. albicans biofilms [39]. A more direct comparison of the C. parapsilosis and C. albicans biofilm transcriptomes revealed that similarities between the species center on metabolic changes, and in particular monocarboxylic acid metabolism (Figure 6). Some of the shared response may be related to the fact that biofilms are hypoxic environments [32], [63]. For example, 43 genes that are up-regulated in C. parapsilosis and C. albicans [39] biofilms and in both species in hypoxic conditions [61], [104] are enriched for processes associated with carbohydrate and lipid metabolism, which may reflect changes occurring during adaptation to hypoxic environments. Other metabolic features, such as glycerol metabolism, are also increasingly being recognized as potential virulence factors [105]. Expression of RHR2, glycerol-3-phosphatase is important for development of C. albicans biofilms, and it is suggested that glycerol levels directly regulate expression of adhesins [105]. RHR2 expression is also increased in C. parapsilosis biofilms (Table S3). Analysis of gene family enrichment was also used to compare the biofilm transcriptional response of these two species. The ATO, TES, OPT and CFEM families are enriched in both C. parapsilosis and C. albicans biofilms (Figure 6B). Although the roles of the ATO and TES families in biofilm development of C. albicans has not been elucidated, increased oligopeptide transfer (OPT) has been associated with early biofilm stages [106]. Some members of the CFEM family are regulated by Bcr1 in both C. albicans and C. parapsilosis [41], [44], [107]. Surprisingly, although CFEM genes are required for biofilm development in C. albicans [108] they do not have a similar role in C. parapsilosis [41]. We further confirmed this here by deleting the CFEM genes in C. parapsilosis in clusters (CFEM1-4 together, CFEM5-6 together and CFEM7). None of the deletion strains were defective in biofilm formation (Table S2). However, the conservation of expression of the four gene families in C. albicans and C. parapsilosis suggest that they may be suitable targets for subsequent study of core biofilm components. Somewhat surprisingly, processes associated with adhesion and those classified as being involved in biofilm formation by the Candida Genome Database [109] are enriched in C. albicans biofilms and not in C. parapsilosis biofilms (Figure 6A). These may be because biofilm development in C. albicans is strongly correlated with the switch from yeast to hyphal growth [29]. One of the major differences between C. parapsilosis and C. albicans is that C. parapsilosis does not make true hyphae. Biofilms produced by C. albicans are composed of a mixture of yeast, pseudohyphal and hyphal cells, with the yeast cells forming the basal layer and an upper layer of hyphal cells [29]. In contrast C. parapsilosis biofilm contains only yeast and pseudohyphal cells in compact layers [28]. We used RNA-seq to identify the targets of the major C. parapsilosis transcriptional regulators during biofilm growth. The Bcr1 and Cph2 regulons are similar (Figure 7) and they have a high degree of overlap with genes enriched in the biofilm transcriptome (Figure 8). This suggests that Cph2 and Bcr1 are major biofilm regulators in C. parapsilosis. Both transcription factors regulate expression of genes involved in transport of carboxylic and other organic acids, and in glycolysis and monosaccharide metabolism. We observed that carboxylic acid transport or metabolism gene sets are enriched in the genes differentially regulated in many of the deletion strains (Figure 7). Even Efg1, which otherwise has little overlap with the C. parapsilosis biofilm network, regulates carboxylic acid transport (Figure 8C). In addition, genes that are differentially regulated in both C. albicans and C. parapsilosis biofilms are enriched in processes associated with carboxylic acid metabolism (Figure 6A). This supports our hypothesis that regulation of metabolism is important for biofilm production in both species. The shared role of Bcr1 as a biofilm regulator in the two species may therefore be related to regulation of carboxylic acid metabolism. In C. parapsilosis biofilms, Cph2 regulates expression of transporters and oxidoreductases, including the highly expressed ATO (Ammonia Transporter Outward) family. The cph2 GSEA network is enriched for gene sets associated with C. albicans biofilms (Figure 8B), suggesting that CpCPH2 may perform some of the role of the C. albicans biofilm regulators. We exploited the RNA-seq data to look for evidence that the transcriptional regulators we identified work together in a network (Figure 9). We also included ChIP-seq data for Efg1 from planktonic cells [18]. We placed Efg1 at the center of the network to simplify the model, because it has many connections with the other regulators. Efg1 is also known to have many roles in both species [18], [95], [110], [111], [112], [113]. Several of the transcription factors regulate expression of other members of the putative network; some are activators (for example deleting EFG1 reduces expression of CZF1) and some as repressors (deleting CZF1 increases expression of EFG1). Some are linked to the putative network by only one connection (ACE2 and GZF3). The role of ACE2 in biofilm development is likely to be associated with its function in regulating cell separation [68], [97], and it may be more important for adherence rather than for mature biofilm growth [40]. It is unlikely that we have identified all the biofilm regulators in C. parapsilosis. For example, our transcription factor deletion collection is not complete. In addition, we have not obtained transcriptional profiling data for the GZF3 deletion, and we have not identified the direct targets of the transcription factors (for example using ChIP-seq). We were also unable to determine the role of NDT80, which our GSEA analysis suggests may be involved (Figure 5). Overall, our analysis suggests that there is some degree of conservation of biofilm networks in C. albicans and C. parapsilosis. This is not surprising, considering the two species are closely related [2]. However, there are also significant differences, with four transcription factors specific to C. parapsilosis, and three unique to C. albicans. In future work, careful characterization the C. parapsilosis biofilm regulon will help us to elucidate the evolution of the networks in these species.
Candida parapsilosis strains (Table S1) were grown in YPD medium (1% yeast extract, 2% peptone, 2% glucose) at 30°C. For colony selection 2% agar was added. To select for transformants, nourseothricin (Werner Bioagents Jena, Germany) was added to YPD agar at a final concentration of 200 µg ml−1. Transformants containing the LEU2 and HIS1 markers were selected on synthetic complete (SC; 0. 19% yeast nitrogen base without amino acids and ammonium sulphate, 0. 5% ammonium sulphate, 2% dextrose, 0. 075% mixture of amino acids, 2% agar) media without leucine or histidine. For biofilm formation, C. parapsilosis was grown in synthetic defined (SD) medium (0. 67% yeast nitrogen base) containing 50 mM glucose. C. albicans was grown in Spider media (1% nutrient broth, 1% mannitol, 0. 2% potassium phosphate). The media used for phenotype screening is shown in Table S2. All deletion strains were grown in 96 well plates in YPD media at 30°C overnight. The cultures were then diluted 1∶100 into a new 96 well plate containing fresh YPD media. The strains were then pinned onto agar plates using a 48 pin bolt replicator. Plates were incubated at 30°C and photographed after 2 and 3 days of growth. Each knockout was scored on growth in comparison to the control strains (CLIB214 and CPRI) on the same media, where −4 indicates a severe growth defect, −3, −2, and −1 indicate strong, moderate and marginal growth defects, 0 is similar to the control strains, and +1 is stronger growth than control strains. Scores were assigned only where the two independent replicates had the same behavior. Screens were repeated at least twice. Growth on different chemical concentrations (e. g. CuCl2, ketoconazole) were combined to give a single score (see Table S2). Deletion strains with interesting phenotypes were further validated by plating exact numbers of cells in decreasing concentration on test media. Control strains were included in each plate. Scores were converted to a Heatmap using Bioconductor [114]. To delete LEU2 (CPAR2_805510), approximately 500 bp upstream and downstream of the open reading frame was amplified using the primer pairs CpLEU2KO1/CpLEU2KO2 and CpLEU2KO3/CpLEU2KO4 respectively (Table S6). Primers KO1 and KO2 contain recognition sites for KpnI and ApaI respectively and KO3 and KO4 for SacII and SacI. The PCR products were purified using a Qiagen PCR purification kit and ligated at either end of a SAT1 flipper cassette in pCD8 [115], generating plasmid pCpLEU2. The entire cassette plus flanking regions were excised by digestion with KpnI and SacI, gel purified and transformed into C. parapsilosis CLIB214. Integration of the cassette at LEU2 was confirmed by PCR using a primer 5′ to LEU2 (CpLEU2KO5) and a primer from inside the cassette (BUT237). Intact LEU2 alleles were identified by PCR with CpLEU2KO5 and CpLEU2KO6. Primers CpLEU2KO1 and CpLEU2KO4 were used to verify recycling of the cassette and deletion of the LEU2 gene. The same cassette was used to delete both LEU2 alleles (Figure S1). A similar method was used to delete the HIS1 gene (CPAR2_100200) using primers CpHIS1KO1 and CpHIS1KO2 to amplify an upstream region and primers CpHIS1KO3 and CpHIS1KO4 to amplify a downstream region. Integration of the cassette was confirmed by PCR using a primer 5′ to the gene CpHIS1KO5 and BUT237. To confirm the presence of an intact allele the primer pair CpHIS1KO5 and CpHIS1KO6 were used and primer pair CpHIS1KO1 and CpHIS1KO4 were used to verify recycling of the cassette and deletion of the HIS1 (Figure S1). The leu2−/his1− strain (CPL2H1) was used as a background for all other deletion constructs. Target genes were deleted in strain CPL2H1 using a fusion PCR method described in Noble et al [45]. Target genes are listed in Tables S1 and S2, and the primers used are listed in Table S6. Approximately 500 bp upstream and downstream of the target gene was amplified using Phusion Taq (New England BioLabs) with primer pairs 1/3 and primers 4/6. The annotated CZF1 open reading frame was corrected (and elongated) by re-sequencing. The selectable markers, C. dubliniensis HIS1 and C. maltosa LEU2 genes were amplified using primers 2 (universal primer) and primer 5 from the plasmids pSN52 and pSN40 respectively. All PCR products were purified using a Qiagen PCR purification kit. The 5′ tails of primers 2 and 3 have complementary regions, as do primers 4 and 5. The 5′ PCR product, the 3′ PCR product and one of the selectable marker products were fused by PCR using primers 1 and 6. The resulting disruption cassette was transformed into the background strain (Figure 1, Figure S1). The first allele was always deleted using the CmLEU2 selectable marker and the second allele using the CdHIS1 gene. Correct integration of the marker gene at the target locus was confirmed by PCR of both ends of the deletion construct; the 5′ region was confirmed using primers 5′check and either LEUcheck1/HIScheck1, and the 3′ region using primers 3′ check and either LEUcheck2/HIScheck2. Loss of the open reading frames was also confirmed using the appropriate ORF primers (Table S6). To create the control reintegration strain, CPRI, upstream and downstream regions of the deleted HIS1 gene were amplified using primers CpHIS1KO1/CpHIS-3 and CpHIS-4/CpHISKO7. The Candida dulbiniensis HIS1 and Candida maltosa LEU2 genes were amplified from the appropriate plasmids using primers 2 (universal primer) and CpHIS-5. The resulting PCR product were fused using primers CpHIS1KO1 and CpHIS1KO7 and transformed into the background strain and in turn re-introducing the C. dulbiniensis HIS1 and C. maltosa LEU2 genes at the original site of the C. parapsilosis HIS1 alleles. Correct integration of the marker genes were confirmed by PCR using the primer CpHIS1KO5 and either LEUcheck1 or HIScheck1. The cph2, czf1, gzf3, ace2, mkc1 and ume6 mutant strains were complemented by introducing one copy of the gene at its original site on the genome [87]. Briefly the entire ORF and promoter were amplified using the primers listed in Table S6. The resulting PCR products were digested using the enzyme sites incorporated into each primer and cloned into pSFS2a. The plasmid was then linearized within the promoter region using the relevant enzyme (MKC1 and UME6 (HpaI), CPH2 and CZF1 (MluI), GZF3 (BsmBI), ACE2 (BsiWI) ) and transformed into the appropriate background mutant strain. Transformants were selected on nourseothricin plates and correct integration of the plasmid was determined by PCR and using one primer upstream of the promoter and one inside the ORF (Table S6). Strains were transformed by electroporation as described previously with some modifications [115]. After electroporation, 950 µl of fresh YPD was added immediately and the mixture incubated at 30°C for 3–4 h. Following incubation, cells were pelleted, washed once in 1 ml of water and resuspended in 300 µl of water. 100 µl was plated onto YPD plates supplemented with nourseothricin at a concentration of 200 µg ml−1. Transformants were obtained following 48 h of incubation at 30°C. The SAT1 cassette was recycled by growing overnight in YPM (1% yeast extract, 2% peptone and 2% maltose). 100 cells were plated onto YPD plates containing 10 µg ml−1 of nourseothricin and incubated overnight at 30°C. Following incubation a mixture of large and small colonies were visible on the plate. Small colonies were restreaked onto fresh YPD agar plates and checked for nourseothricin sensitivity. The second allele was deleted using the same protocol. Most deletion strains were constructed using chemical transformation. An overnight culture was diluted to an A600 of 0. 2 in 30 ml of YPD broth. This was grown at 30°C to an A600 of 1. The culture was centrifuged at 4000 g for 5 min and the pellet resuspended in 3 ml of ice-cold water. The re-suspended pellet was centrifuged again as above and the pellet re-suspended in 200 µl of ice-cold TE-LiOAC (0. 1M lithium acetate, 10 mM Tris and 1 mM EDTA). A transformation mix was set up that contained 10 µl of boiled and cooled salmon sperm DNA (10 mg ml−1), 20–30 µl of fusion PCR product and 100 µl of competent cells. This was incubated at 30°C for 30 min followed by addition of 700 µl of PLATE (0. 1M lithium acetate, 10 mM Tris, 1 mM EDTA and 40% PEG 3350). Samples were incubated overnight at 30°C. Cells were heat shocked at 44°C for 15 min, centrifuged, and washed with 1 ml of YPD. The cells were centrifuged again and finally re-suspended in 100 µl of YPD followed by incubation at 30°C for 2 h. Cultures were then spread on the appropriate drop-out agar plates and incubated for 2–3 days at 30°C. Strains were tested for biofilm development crystal violet staining and direct observation using Nunclon Delta 24-well polystyrene plates as follows. C. parapsilosis trains were grown overnight at 30°C and washed twice in phosphate buffered saline (PBS), diluted to an A600 of 1 in SD media with 50 mM glucose, and 1 ml was added to each well. The cultures were incubated for 2 h at 37°C at 50 rpm. Wells were then washed once with 1 ml of PBS to remove non-adherent cells. 1 ml of fresh SD 50 mM glucose media was then added to each well. Plates were then incubated for 48 h at 37°C at 50 rpm. The supernatants were removed, and each well was washed twice with 1 ml of PBS. Plates were allowed to dry overnight at room temperature. Biofilms were stained with 500 µl 0. 4% crystal violet for 10 min. The dye was removed and wells were washed with PBS. Plates were allowed to dry and biofilm was photographed. For C. albicans biofilms, 6 well plates were pre-treated with 10% fetal bovine serum overnight which was removed by washing with 1 ml of PBS. C. albicans strains from overnight cultures were washed twice with PBS, diluted to a starting A600 of 0. 5 in Spider media and 5 ml was added to the pre-treated wells. The plates were incubated for 90 min at 37°C at 100 rpm. Each well was then washed once with PBS and fresh spider media was added followed by incubation for 48 h at 37°C at 100 rpm. The supernatants were removed and each well was washed twice with 1 ml of PBS. For dry mass measurements, C. parapsilosis biofilms were formed in Nunclon Delta 6 well plates. Assays were set up as for the 24-well plates, except that 5 ml of media was used in each well. After the final wash, 1 ml of PBS was added to each well and adherent biofilms were scrapped from the bottom of the wells. The contents of two wells were vacuum filtered over a pre-weighed 0. 8 µm nitrocellulose filter (Millipore). The filters were dried overnight in a warm room and weighed the following day. The average total biomass of each strain was calculated for 3 independent samples by subtracting the initial weight of the filter from the final weight. Statistical significance was calculated using the students two tailed paired t-test, and only those with a P-value <0. 005 were retained. C. parapsilosis biofilms were grown on Thermanox (Nunc) slides in in 6 well plates. Briefly, overnight cultures were washed twice in PBS and diluted to an A600 of 1 in SD media with 50 mM glucose. 5 ml was added to each well, and incubated for 2 h at 37°C at 50 rpm. The slides were gently removed and placed into a fresh well containing 5 ml of PBS, and gently placed into a well containing fresh SD/50 mM glucose media and incubated for 48 h at 37°C at 50 rpm. The slides were removed and washed as above, and then stained with 25 µg ml−1 concanavalin A (conA) -Alexa Fluor 594 conjugate (C-11253; Bio-science) for 45 min at 37°C. The liquid was removed from each well, and the Thermonox slides were flipped and placed on a 35-mm-diameter glass-bottomed petri dish (MatTek Corp. , Ashland, MA). The biofilms were observed with a Zeiss LSM510 confocal scanning microscope with a ×40-magnification oil objective. A HeNe1 laser was used to excite at a 543-nm wavelength. All images were captured and analyzed using a Zeiss LSM Image Browser and Fiji. Biofilms for RNA-seq analysis were grown in SD media with 50 mM glucose at 37°C in Nunc 6 well plates. The biofilm was scraped from the bottom of each well and combined for RNA isolation. The number of wells combined depended on the amount of biofilm produced and the amount of biomass needed to obtain sufficient RNA for library preparation (6 wells for CLIB214 and 6–10 wells for the deletion strains). RNA was isolated using an Ambion Ribopure-Yeast RNA kit. To isolate RNA for planktonic cells, strains were grown to an A600 of 1 in SD media with 50 mM glucose at 37°C. Strand specific RNA-seq library preparation and sequencing was carried out by BGI (www. genomics. cn, Hong-Kong). Paired-end reads (Illumina HiSeq 2000,2×90 bp, 2 GB clean data) were obtained from three biological replicates from wild type (Candida parapsilosis CLIB214) in planktonic and biofilm conditions, and from ace2, cph2, efg1, czf1, ume6 and bcr1 deletion strains in biofilms. Two replicates were obtained form one deletion construct, and the third from the independent replicate strain. Samples were aligned to the genome [61] using TopHat2 [116]. HTSeq [27] was used to count mapped reads per gene. Differentially expressed genes were identified using DESeq2 [117] with an adjusted p-value threshold of 0. 001 and a log2 fold change threshold of −1. 5 and 1. 5. Default parameters in DESeq2 were used, except that Cook distance filtering was turned off. Significantly differentially expressed genes were clustered using hierarchical clustering in R [114]. C. albicans orthologs were obtained from the Candida Gene Order Browser (CGOB) [47]. The GO term finder from Candida Genome Database [109] was used to carry out Gene Ontology analyses. All RNA-seq data is available from GEO accession number GSE57451. Gene families in C. parapsilosis and C. albicans were identified using Blast similarity searches [118] and the MCL algorithm [74]. In both species an inflation factor of 2. 1 was used with MCL to robustly identify similar clusters. The average log2 fold change of significantly differentially expressed genes in each cluster was calculated, and ranked from high to low (Figure 8, Table S3). For C. albicans biofilms, gene expression levels were reported from a mixture of RNA-seq and microarray analysis [39]. We used significant RNA-seq values were available, and values from microarrays if the equivalent RNA-seq data was not significant. P-values for microarrays were not available. Significant genes were defined as log2 fold change greater than 1. 5 or less than −1. 5, and with an adjusted p-value less than 0. 001 (for RNA-seq only). Plots were generated in R and analyzed manually. The pre-ranked Gene Set Enrichment Analysis tool (GSEAPreRanked, see http: //www. broadinstitute. org/gsea/) was used to determine gene sets that are enriched in differentially expressed genes from the RNA-seq data. Only C. parapsilosis genes with orthologs in C. albicans were included, with gene sets provided by Dr. Andre Nantel (http: //www. candidagenome. org/download/community/GSEA_Nantel_2012/) [70]. Genes were ranked from highest to lowest by log2 fold change. GSEAPreRanked was used with the default options, excluding gene sets with less than 5 or more than 1000 genes, resulting in the analysis of 5249 gene sets in total. Significant results were defined as p-value lower than 0. 005 and an FDR Q-value lower than 0. 1. GSEAPreRanked results were further analyzed using Cytoscape (version 2. 8. 2) and the EnrichmentMap plugin (version 1. 2, [71]) using an overlap coefficient cutoff of 0. 5 and default settings. The size of the nodes correlates to the number of genes in the gene sets from Sellam et al [70], which are based on C. albicans. We use a background of C. parapsilosis orthologs for the GSEA, but the networks generated in Cytoscape reflect the original gene sets from Sellam et al [70]. | Candida species are among the most common causes of fungal infection worldwide. Infections can be both community-based and hospital-acquired, and are particularly associated with immunocompromised individuals. Candida albicans is the most commonly isolated species and is the best studied. However, other species are becoming of increasing concern. Candida parapsilosis causes outbreaks of infection in neonatal wards, and is one of the few Candida species that is transferred from the hands of healthcare workers. C. parapsilosis, like C. albicans, grows as biofilms (cell communities) on the surfaces of indwelling medical devices like feeding tubes. We describe here the construction of a set of tools that allow us to characterize the virulence properties of C. parapsilosis, and in particular its ability to grow as biofilms. We find that some of the regulatory mechanisms are shared with C. albicans, but others are unique to each species. Our tools, based on selectively deleting regulatory genes, will provide a major resource to the fungal research community. | Abstract
Introduction
Results
Discussion
Materials and Methods | genetics
molecular biology
biology and life sciences
mycology | 2014 | Comparative Phenotypic Analysis of the Major Fungal Pathogens Candida parapsilosis and Candida albicans | 16,113 | 245 |
Gene expression is a heritable cellular phenotype that defines the function of a cell and can lead to diseases in case of misregulation. In order to detect genetic variations affecting gene expression, we performed association analysis of single nucleotide polymorphisms (SNPs) and copy number variants (CNVs) with gene expression measured in 869 lymphoblastoid cell lines of the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort in cis and in trans. We discovered that 3,534 genes (false discovery rate (FDR) = 5%) are affected by an expression quantitative trait locus (eQTL) in cis and 48 genes are affected in trans. We observed that CNVs are more likely to be eQTLs than SNPs. In addition, we found that variants associated to complex traits and diseases are enriched for trans-eQTLs and that trans-eQTLs are enriched for cis-eQTLs. As a variant affecting both a gene in cis and in trans suggests that the cis gene is functionally linked to the trans gene expression, we looked specifically for trans effects of cis-eQTLs. We discovered that 26 cis-eQTLs are associated to 92 genes in trans with the cis-eQTLs of the transcriptions factors BATF3 and HMX2 affecting the most genes. We then explored if the variation of the level of expression of the cis genes were causally affecting the level of expression of the trans genes and discovered several causal relationships between variation in the level of expression of the cis gene and variation of the level of expression of the trans gene. This analysis shows that a large sample size allows the discovery of secondary effects of human variations on gene expression that can be used to construct short directed gene regulatory networks.
Genome-wide association studies (GWAS) have discovered a large number of loci implicated in many complex traits and diseases [1]. The vast majority of variants discovered are found in non-coding regions (88%), which challenges the interpretation of their functional effect [1]. One way to overcome this challenge is to look for associations between variants and an intermediate cellular phenotype, such as gene expression. Expression quantitative trait loci (eQTL) analysis have been successful in mapping variants to gene expression in several cell types providing a better understanding of the genetics of gene expression, and revealing functional impacts of variants associated with complex traits and diseases [2]–[10]. Most studies so far were conducted on relatively small sample sizes [11], limiting the power to detect variants affecting gene expression in cis and to a greater extent in trans, as trans-eQTLs typically have weaker effect sizes than cis-eQTLs [12]. Detecting eQTLs with small effect sizes is indeed important, as a variant can have a weak effect in the tissue sampled but a strong effect in the tissue relevant for a specific disease. Furthermore, small effects in cis can be important if the associated gene plays a substantial role in a cellular process. A large sample size also allows the discovery of variants that are both cis and trans-eQTLs, suggestive of a regulatory relationship between the cis regulated gene and the trans regulated gene. Here we measured gene expression in Lymphoblastoid cell lines (LCLs) from 869 genotyped individuals of the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort in order to map single nucleotide polymorphisms (SNPs) and copy number variants (CNV) with minor allele frequency >1% to gene expression in cis and in trans. We then investigated trans effects of cis-eQTLs and used causal models to investigate the mechanism by which a variant can affect the expression of a gene in cis and in trans.
In order to better understand how genetic variations affect gene expression in LCLs, we associated 2' 290' 057 imputed single-nucleotide polymorphisms (SNPs) from the HapMap2 reference set and 3329 copy number variants (CNVs) to gene expression in 869 individuals from the ALSPAC cohort. We first looked for cis-eQTLs, defined as variants associated with gene expression in a 2 MB window surrounding the transcriptional start site (TSS). We used spearman rank correlation to test for association between 20' 745 probes on autosomes, measuring the expression of 14' 835 genes, and variants present in the windows. A gene-based significance threshold was determined by permuting all expression phenotypes 1000 times (Methods). We discovered that 3534 genes had a cis-eQTL at a false discovery rate (FDR) of 5% (3498 due to SNPs, 36 due to CNVs) (Table S1, Table S2). As shown previously [5], [9], we found that cis-eQTLs cluster around the TSS (Figure S1). CNVs were more often cis-eQTLs than expected by chance (i. e: if SNPs and CNVs had the same probability to be a cis-eQTL, we would expect the same fraction of CNVs and SNPs in our cis-eQTL results than in the whole data set) (odd ratio: 7. 1, Fisher' s exact test pvalue<2. 2e-16), suggesting that CNVs are more likely to affect gene expression than SNPs. We observed a strong correlation (rho = 0. 98, pvalue = 1e-13) between sample size and the number of genes with a cis-eQTL discovered (Figure S2). Although, it appears that the rate of discovery is decreasing for large sample sizes, it is likely that further increases in sample size would allow the discovery of more genes with at least one cis-eQTL. In order to detect genes affected by more than one cis-eQTL, we performed conditional regression by removing the effect of the cis-eQTL (s) on gene expression and repeating the association analysis on the residuals until no significant associations could be detected (Methods) [13], [14]. We found that 694 genes (19. 6% of the genes with a cis-eQTL) have at least two independent cis-eQTLs (Figure S3). For 374 genes (53. 9%), independent cis-eQTLs were located in the same recombination interval, showing that variants in linkage disequilibrium can have different functional effects on gene expression. In order to evaluate the importance of independent cis-eQTLs on the heritability of gene expression, we obtained heritability estimates from the MUTHER cohort in LCLs [2]. On average, the heritability of genes with multiple cis-eQTLs was greater than for genes with only one cis-eQTL detected (mean heritability = 0. 24 for genes with one cis-eQTL, mean heritability = 0. 38 for genes with multiple eQTLs, Mann-Whitney U test pvalue<2. 2e-16). For each gene with multiple eQTLs, we compared the heritability explained by the best eQTL to the heritability explained by all independent eQTLs using a linear model where the standard normal expression of the gene is explained by the best eQTL or all independent eQTLs (Methods). We observed that the best eQTL explains on average 46% of the heritability of the traits while all independent eQTLs explain on average 57% of the heritability of gene expression (Figure 1). These results show that independent cis-eQTLs are detected preferentially in genes with a relatively high genetic component of their expression and that independent cis-eQTLs explain 11% more of the heritability of the gene expression on average than using only the best cis-eQTL. Several studies have shown that the effect of variants on gene expression is tissue dependent [5], [9], [10], [15]. Indeed, some eQTLs can have different effect sizes in cells of different developmental origin [9] or can be detected only in specific tissues [5], [9], [10], [15]. However, the estimated tissue sharing of eQTLs has steadily increased in function of the cohort sample sizes, ranging from 20%–31% [5] with a small cohort to 56–83% in a larger cohort [2], questioning the relevance of interrogating different tissues. In order to address this question, we took advantage of the large sample size of the ALSPAC cohort to investigate the effect of sample size on tissue sharing. We obtained cis-eQTLs detected in LCLs, skin and adipose tissues from the MUTHER cohort [2], one of the largest studies investigating eQTL tissue specificity. We assessed tissue sharing as a function of sample size in a continuous manner by matching cis-eQTLs detected by the MUTHER study (1%FDR) with the pvalues detected for the same associations in different subsets of individuals of the ALSPAC cohort and computed the π1 statistic (estimate of the proportion of true positives in a pvalue distribution) for each sample size [16]. We observed little tissue sharing for small sample sizes (30. 6% with adipose tissue, 34. 8% with skin, and replicated 46. 9% in LCLs for 50 individuals) (Figure 2). In contrast, using the entire ALSPAC cohort, we replicated 79. 2% of the eQTLs in LCLs, detected by the MUTHER project, and estimated tissue sharing to be 61. 6% for adipose tissue and 61. 7% for skin cells. We did not observe an increase in the π1s for LCL and skin cells after 600 individuals, while the sharing of MUTHER adipose eQTLs with ALSPAC LCLs continued to increase slightly for larger sample sizes (Figure 2). We found stronger concordance in the directionality of eQTLs replicating within LCLs (1. 8% with opposite directionality at 5% FDR) compared to eQTLs shared across tissues (10. 4% with opposite directionality in skin and 10. 5% in adipose tissue at 5%FDR). These results indicate that a relatively large proportion of cis-eQTLs detected in one tissue cannot be detected in other tissues and support the idea that one should perform eQTL analysis in different tissues in order to map all regulatory variants in the genome. We next investigated whether we could detect variants affecting gene expression in trans. We defined trans-eQTLs as variants affecting gene expression at a distance greater than 5 MB from the TSS or on another chromosome. We used spearman rank correlation to test for association between 21' 634 probes, measuring the expression of 14' 441 genes on autosomes and chromosome X, and all variants further than 5 MB from their TSS. A genome-wide significance threshold was determined by permuting a subset of the expression phenotypes 1000 times (Methods). We discarded trans associations of CNVs when the gene associated was on the same chromosome or on a chromosome with SNPs correlated with the CNV (r2>0. 1) as some CNVs appeared to be mismapped (Text S1). In addition, we discarded all significant associations of CNVs with an imbalance in copy number between males and females, as this resulted in the false trans associations of the CNVs with genes differentially expressed between males and females. After filtering, we discovered trans-eQTLs for 48 genes (FDR = 5%) (45 due to SNPs and 3 to CNVs) (Table 1, Table S2, Table S3). We assessed the replication of the trans-eQTLs using the MUTHER cohort [2], a twin cohort, which we separated in two sets of unrelated individuals (group 1: 340 individuals, group 2: 338 individuals). We replicated 22 trans-eQTLs (of 40 tested) with a pvalue<0. 05 in the first group and 19 in the second group (union = 23). In order to validate the array-based trans-eQTLs with an independent technology, we used the GEUVADIS [17] cohort (373 individuals with RNA-seq in LCLs) and replicated 11 trans-eQTLs (of 32 tested). As for cis-eQTLs, CNVs were more often trans-eQTLs than expected by chance (i. e: if SNPs and CNVs had the same probability to be a trans-eQTL, we would expect the same fraction of CNVs and SNPs in our trans-eQTL results than in the whole data set) (odd ratio: 45. 8, Fisher' s exact test pvalue = 5e-5). Two variants, rs1156058 and rs705170, were associated with a total of 14 and 7 genes in trans respectively (Figure S4). We also found that rs4781011, located on chromosome 16 within 5 kb of the TSS of the gene CIITA (class II, major histocompatibility complex transactivator), a gene known to activate in trans the HLA locus on chromosome 6, was a trans-eQTLs of CD74 on chromosome 5, a protein that regulates antigen presentation. This analysis shows that it is much more difficult to detect trans-eQTLs than cis-eQTLs at the same false discovery rate. Although our replication cohorts had a sample size representing only roughly 40% of the discovery cohort, we replicated approximately 50% of the trans-eQTLs attempted. This encouraging result suggests that more trans-eQTLs could be replicated with a bigger replication cohort and that our trans-eQTLs detection methodology is efficient. Genome-wide association studies (GWAS) found many SNPs associated with diverse phenotypes but the mechanistic link between the GWAS-SNP and the phenotype remains unclear for the vast majority of the associated SNPs. One possibility is that a GWAS-SNP affects gene expression, which then leads to the phenotype. It was previously shown that trait associated SNPs were more likely to be cis-eQTLs [18]. However, since the publication of this result, many more GWAS were performed, increasing dramatically the number of variants associated with complex traits and a much larger number of eQTLs were discovered in this study. In order to confirm that GWAS identified variants are more likely to be cis-eQTLs and to investigate if a similar relationship exists for trans-eQTLs, we accessed the catalog of published genome-wide association studies (http: //www. genome. gov/gwastudies/) on 19 March 2012. 5381 SNPs reported in the catalog at that date were genotyped in our study. We looked for GWAS-SNPs overlapping eQTLs and found that 850/3 (15. 8%/0. 06% of the GWAS-SNPs) GWAS-SNP co-localized with variants significantly associated in cis/trans (Table S5) (Table S6). This is significantly more than using SNPs matched to the GWAS-SNPs for distance to closest gene and minor allele frequency (for cis-eQTLs, median = 585, pvalue<0. 001) (for trans-eQTLs, median = 0, pvalue<0. 01) (Figure S5). This confirms that many GWAS-SNPs are probably playing a role on disease susceptibility by affecting gene expression in cis and that trait associated SNPs are also more likely to be trans-eQTLs [19]. We next sought to determine whether trans-eQTLs were also cis-eQTLs, as this may indicate that the genes regulated in cis play a role in the regulation of the trans genes. We examined the overlap between trans-eQTLs and cis-eQTLs and found that 5 (18. 5%) of the unique trans-eQTLs were also associated with gene expression in cis. This overlap is significantly greater than the overlap obtained using variants matched to the trans-eQTLs for distance to closest gene and minor allele frequency (1000 permutations, median = 0, pvalue<0. 001) (Figure S6). We find that the cis-eQTLs of two transcription factors, BATF3 and HMX2, are associated to the most genes in trans. The cis-eQTL of BATF3, a gene involved in the differentiation of CD8α+ dendritic cells and IL17-producing T helper cells [20], [21], is a trans-eQTL of 14 genes, distributed on 8 chromosomes. The cis-eQTL of HMX2 is a trans-eQTL of 7 genes distributed on 4 chromosomes. HMX2 is a transcription factor directing development of inner ear and hypothalamus in mice [22] and deletion of the chromosomal region containing HMX2 in human is associated to inner ear malformations, vestibular dysfunction and hearing loss [23]. Other genes with a cis-eQTL that is also a trans-eQTL are: GNA15, a G protein, S1PR4, a G protein coupled receptor, PIDD, an effector of p53 apoptosis in mice and CRIPAK, an inhibitor of the PAK1 transcription factor. These results show that we can detect potential new functional targets of important genes in LCLs by combining cis-eQTLs and trans-eQTLs. In order to detect more possible functional relationships between genes regulated in cis and in trans by the same variants, we looked for the trans effects of the subset of variants that were found to be cis-eQTLs. Since we discovered 3475 variants associated to 3534 genes in cis, a trans-analysis of this subset of variants has the advantage of reducing the number of tests performed and therefore allows us to discover more trans effects of cis-eQTLs. Before investigating which cis-eQTLs are affecting which genes in trans, we first aimed to assess how many trans effects of cis-eQTLs could be detected if we had a much larger sample size. We used spearman rank correlation to test for associations between 23' 935 probes, measuring the expression of 16' 505 genes and all unique cis-eQTLs further than 5 MB from the TSS. We obtained approximately 23' 935 trans association pvalues per cis-eQTL and computed the π1 statistics (estimate of the proportion of true positives in a pvalue distribution) on each set of pvalues, resulting in 3475 π1 estimates [16]. These estimates represent the proportion of probes that are affected in trans by the 3475 variants that are cis-eQTLs, without being able to pinpoint all individually significant effects. We observed that a large number of cis-eQTLs are affecting a large number of probes in trans (52% of the cis-eQTLs have a π1 >0) ranging from a few probes affected to up to 37. 2% of the probes (median = 0. 004603 corresponding to 110 probes) (Figure 3A). Interestingly, the variant with the most trans effects (37. 2% of the probes), rs482519, is the cis-eQTL of WHSC1 (Wolf-Hirschhorn Syndrome candidate1), a histone methyltransferase. A potential explanation for this result it that variation of the level of expression of this histone methyltransferase could affect the expression of many genes by modifying chromatin accessibility. The second variant with the most trans effects (33. 5% of the probes), rs2978387, is the cis-eQTL of ZNF16 (Zinc Finger Protein 16), a protein that may act as a transcription factor. The third variant with the most trans effects (32. 1% of the probes), rs12196956, is the cis-eQTL of TBC1D22B (TBC1 domain family member 22B), a protein that may act as a GTPase-activating protein for Rab family protein. Furthermore, we observed a negative correlation between the strength of the cis-eQTLs and the number of probes affected in trans (spearman rho = −0. 1, pvalue = 4. 2e-10), suggesting that strong cis-eQTLs may be selected against in the population for genes modulating the expression of many genes. These results show that cis-eQTLs can have trans effects on many genes, which have direct consequences on regulatory network perturbations. Although we estimated that a large number of cis-eQTLs are affecting many genes in trans, we would need a very large sample size to detect all of them at a reasonable false discovery rate. In order to assess which cis-eQTL is affecting which genes in trans, a genome-wide significance threshold was determined by permuting all expression phenotypes 1000 times (Methods). 92 genes had significant trans-effects due to cis-eQTLs (FDR = 5%) (Table S4). We replicated 31 associations (of 79 tested) in the first set of twins of the MUTHER cohort, 22 in the second set (union = 34) and 27 in GEUVADIS (of 75 tested). We discovered substantially more trans-effects of the cis-eQTLs of BATF3 and HMX2 with 39 and 18 genes regulated in trans respectively (Figure 3B). Other examples of cis-eQTLs with several significant trans associations include the cis-eQTL of PSMG1 (proteasome assembly chaperone 1) affecting 3 genes in trans and the cis-eQTL of BRWD1 (bromodomain and WD repeat-containing protein 1), which may be a transcriptional activator [24], also affecting 3 genes in trans. We did not find significant effects of the cis-eQTL of WHSC1, indicating that the large number of effects on gene expression have too small effect sizes to be discovered individually given our sample size. In total we found that 26 variants are cis-eQTLs of 27 genes and trans-eQTLs of 92 genes. 4 genes associated in cis to a cis/trans-eQTLs also had independent cis-eQTLs. We regressed out the effect of the main eQTLs on the trans genes expression and found that in 95% of the cases the independent eQTLs had the same allelic effect as the main eQTLs, i. e. the high expressing allele of the main eQTL has the same effect - high or low – in the trans gene as the high expressing allele of the second independent eQTL in 95% of the cases. This concordance further highlights the biological relevance of these trans eQTLs since their downstream biological effects, mediated by the modulation of the cis genes, are replicated by independent variants. 1 independents cis-eQTL (associated to HMX2) was also significantly associated to 1 gene in trans (5% FDR) and had the same allelic effect as the main eQTL. The strong concordance in allelic effects between main cis-eQTLs and independent cis-eQTLs indicate that for those 4 genes, most of the trans effects are due to variations in the level of expression of the cis gene. We then explored whether the trans associations of the cis-eQTLs were causally due to the variation in the expression level of the cis genes. We assessed the likelihood of three different models using two methods: Bayesian networks and a causal inference test (CIT) (Methods) [25]. The first model (SCT) states that the variant is affecting the expression level of the cis gene, which then leads to variation in the level of expression of the trans gene. The second model (INDEP) states that the trans effect and the cis effect are independent and the third (STC) unlikely model states that the variant is affecting the level of expression of the trans gene, which then affects the level of expression of the cis gene. We observed that 100%/100%/94% of the SCT/STC/INDEP models detected by the CIT method are also detected as the best model by Bayesian networks. Conversely, 86%/33%/100% of the SCT/STC/INDEP models called by Bayesian networks were also detected as the best model by CIT. By taking the overlap of the two methods, we obtain 19 SCT, 2 STC and 49 INDEP relationships. We found causal effects (SCT) of CRIPAK on AVP, CCL5 on NPSR1, BATF3 on three genes and HMX2 on 14 genes (Table 2). The large representation of INDEP relationships is due to several factors. First, false positives will be called INDEP because their association is not due to the cis gene expression. As we found 92 trans associations of cis-eQTLs at a 5% FDR, we expect ∼5 INDEP relationships due to false positives. In addition, we expect 1 INDEP relationship because one cis-eQTLs is associated to two genes in cis and 2 genes in trans in total. It is unlikely that both of the cis associated genes would have causal effect on one trans associated gene leading to INDEP calls for 1 relationship. Finally, we observed that 34 INDEP associations are due to the cis-eQTL of BATF3. The INDEP relationships show that the trans gene associations are not due to the cis gene expression. However, they could be due to change in the structure of the protein if other functional variants, such as non-synonymous SNPs or splice variants, are in linkage disequilibrium with the cis-eQTLs. Alternatively, the cis-eQTLs could be affecting the expression of non-coding RNA in the vicinity of the cis genes that could drive the trans associations. As we hypothesised that the trans-effects of some cis-eQTLs could be due to changes in the protein structure, we investigated the trans effects of 11564 non-synonymous SNPs discovered by the 1000 genome project and genotyped in the ALSPAC cohort. We used spearman rank correlation to test for associations between the 23' 935 probes, measuring the expression of 16' 505 genes and all non-synonymous SNPs further than 5 MB from the TSS. We first looked at large effects that could be detected given our sample size and found that 9 genes were affected in trans by non-synonymous SNPs (5% FDR) (Table S7). We replicated 4 associations (of 6 tested) in the first set of twin of the MUTHER cohort, 4 in the second set (union = 4) and 1 (of 7 tested) in GEUVADIS. We then looked at the global effects of non-synonymous SNPs on gene expression in trans by looking at the proportion of true positive in the distribution of the trans association pvalues for each variant using the π1 statistic [16]. We found that non-synonymous SNPs have significantly less trans effects on gene expression than cis-eQTLs (Mann-Whitney U test, pvalue = 1. 8e-9) (Figure S7), as observed previously [4]. This result is compatible with the observation that common regulatory SNPs have more effects on complex traits and common diseases than common non-synonymous SNPs. Finally, we explored the trans effects of the 5381 SNPs associated with complex traits and diseases in order to detect potential effects of these variants on gene expression and discovered that 66 of them are significantly associated to 10 genes in trans (5% FDR) (Table S8). We replicated 6 associations (of 7 tested) in the first set of twin of the MUTHER cohort, 6 in the second set (union = 6) and none in GEUVADIS (of 3 tested). For example, we found that rs11171739, which is associated to type 1 diabetes, is a trans-eQTL of DCAF16, as previously shown in monocytes [26] and is also a trans-eQTL of BEND4. We also found that rs4781011, which is associated to ulcerative colitis, is a trans-eQTL of CD74, a protein involved in immune response. rs2227139, which is associated to hematological parameters was associated in trans to ERG, which regulates hematopoiesis and the function of adult hematopoietic stem cells [27]. These results show that we can detect downstream effects of disease-associated variants, an important step to understand the relevant biological pathways in common diseases.
The large sample size of the ALSPAC cohort allowed us to discover that 3534 genes are affected by genetic variants in cis and 48 in trans. We found that CNVs are enriched in the best associations per gene in cis and to an even greater extent in trans. This enrichment is not surprising as CNVs are more likely to disrupt regulatory elements than SNPs due to their size [8]. This result indicates that CNVs are more likely to be causal than SNPs in genetic diseases resulting from the misregulation of gene expression. Several examples of genetic disorders, such as aniridia, sex-reversal and holo-prosencephaly are already known to be caused by duplications or deletions of CNVs located in non-coding regions of developmental genes [28]. We found that SNPs associated with complex traits and common diseases are more likely to be cis and trans-eQTLs than matched variants. Although some of these overlaps might be coincidental [3], these results further confirm that a significant fraction of trait associated SNPs are acting at the gene expression level. We observed that many eQTLs detected in skin and adipose tissues could not be detected in LCLs irrespectively of the sample size, showing that a significant fraction of eQTLs is tissue specific. Therefore, eQTL studies in many different tissues are needed in order to map all regulatory variants in the human genome and understand their precise tissue specific effect, a necessary step to understand why a specific tissue becomes the “disease” tissue and not other tissues. We estimated that 52% of the cis-eQTLs have trans-effects on gene expression ranging from a few probes to up to 37. 2% of the probes. The large number of trans-effects of cis-eQTLs is concordant with the fact that on average 65% of the heritability of gene expression is trans to the gene in LCLs [2]. As we can detect only a minority of these effects at a reasonable false discovery rate with our relatively large sample size, it indicates that most of the trans-effects of the cis-eQTLs are of small effect sizes. If complex traits and common diseases have the same underlying architecture as gene expression, a substantial part of the missing heritability will then be due to many common variants of very small effect sizes. Using Bayesian networks and causal inference tests [25], we could detect 19 cases where a variant affects the expression of a gene in cis that is causally affecting the expression of a gene in trans (14 due to the cis-eQTL of HMX2). For example, the level of expression of CRIPAK, a protein implicated in cytoskeleton remodelling and influencing PAK1 mediated estrogen transactivation activity [29], is causally affecting the level of expression of AVP (arginine vasopressin), a hormone with anti-diuretic effects on the kidney and affecting social behaviour [30]. Taken together, these results show that population based strategies allow to detect important relationships between genes. Ultimately, this type of approach performed with larger sample sizes will allow us to uncover the cascade of events that lead a disease associated variants to the disease phenotype.
Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. ALSPAC is a prospective birth cohort which recruited pregnant women with expected delivery dates between April 1991 and December 1992 from Bristol UK. 14,541 pregnant women were initially enrolled with 14,062 children born. Detailed information on health and development of children and their parents were collected from regular clinic visits and completion of questionnaires. A detailed description of the cohort is available on our website (http: //www. bristol. ac. uk/alspac/researchers/) and has been published previously [31]. Please note that the study website contains details of all the data that is available through a fully searchable data dictionary (http: //www. bris. ac. uk/alspac/researchers/data-access/data-dictionary/). DNA has been extracted as described previously from blood samples collected from cord blood at research clinics [32]. Lymphoblastoid cell lines were established by transforming lymphocytes from blood samples taken when the study participants were 9 years old, with Epstein Barr Virus. ALSPAC individuals were genotyped using the Illumina HumanHap550 quad genome-wide SNP genotyping platform by 23andMe subcontracting the Wellcome Trust Sanger Institute, Cambridge, UK and the Laboratory Corporation of America, Burlington, NC, USA. Markers with <1% MAF, >5% missing genotypes or which failed an exact test of Hardy-Weinberg equilibrium (P<5×10−7) were excluded from further analysis. Any individuals who did not cluster with the CEU individuals in multidimensional scaling analysis, who had >3% missing data, minimal or excessive heterozygosity (>33% or <31%), evidence of cryptic relatedness (>10% IBD) or incorrect gender assignments were excluded from further analysis. After data cleaning 315,807 SNPs were left. Imputation was carried out using MACH 1. 0. 16, Markov Chain Haplotyping [33], [34], using CEPH individuals from phase 2 of the HapMap project as a reference set. Imputed markers with imputation quality r2<0. 8, with MAF<1% or which failed an exact test of Hardy-Weinberg equilibrium (P<5×10−7) were excluded resulting in a total of 2' 290' 057 high quality SNPs. The CNVs were genotyped using a targeted Agilent 105K CGH array. The design of the array and the methodology for analyzing the array data was previously described in details [35]. LCL' s from unrelated individuals were grown under identical conditions and cells frozen in RNAlater. RNA was extracted using an RNeasy extraction kit (Qiagen) and was amplified using the Illumina TotalPrep-96 RNA Amplification kit (Ambion). Expression profiling of the samples, each with two technical replicates, were performed using the Illumina Human HT-12 V3 BeadChips (Illumina Inc) including 48,804 probes where 200 ng of total RNA was processed according to the protocol supplied by Illumina. Raw data was imported to the Illumina Beadstudio software and probes with less than three beads present were excluded. Log2 - transformed expression signals were then normalized with quantile normalization of the replicates of each individual followed by quantile normalization across all individuals. We restricted our analysis to 23' 935 probes tagging genes annotated in Ensembl. Principal component analysis was performed on 931 individuals. 62 individuals with principal component 1 or 2 greater than one standard deviation of the population were excluded from further analysis. Raw expression data are available upon request at http: //www. bristol. ac. uk/alspac/researchers/data-access/policy/. All eQTL analysis were performed at the single variant level and assumed an additive model. We used spearman rank correlation to test for association between probe expression and genotype. For the cis-analysis, we limited the variants tested to variants present in a 2 MB window surrounding the transcription start site of the gene and we filtered out probes containing SNPs with minor allele frequency >1% according to the 1000 genomes project dataset [36]. To assess significance, we permuted all expression probes 1000 times and kept the best pvalue per gene after each permutation. For each gene, we ranked the permutation pvalues and assessed whether a variant in the non-permuted data had a lower association pvalues than the permutation threshold considered. We then computed the false discovery rate associated with the permutation threshold and subsequently selected the permutation threshold that provides a 5% false discovery rate. For the trans analysis, we tested all variants except variants present in a 5 MB window surrounding the transcription start site. In order to remove false positives, we excluded probes mapping to multiple locations according to ReMOAT [37]. To assess significance, we permuted 1000 times 288 random probes, each corresponding to one gene. As each probe is tested by approximately the same number of SNPs and as we used spearman rank correlation, which is robust to outliers, we treated our permutations as if we had permuted one probe 288' 000 times. We combined all pvalues obtained from the permutations (288*1000), ranked them and increased the genome-wide pvalue threshold until we reached a 5% false discovery rate (corresponding to a pvalue of 9. 5e-11). For the trans analysis of cis-eQTLs, we tested all unique cis-eQTLs except variants present in a 5 MB window surrounding the TSS. In order to remove false positives, we excluded probes mapping to multiple locations according to ReMOAT [37]. To assess significance, we permuted all expression probes 1000 times. As for the trans analysis of all variants, we combined all pvalues obtained, ranked them and increased the genome-wide pvalue threshold until we reached a 5% false discovery rate (corresponding to a pvalue of 7. 6e-8). For the trans analysis of non-synonymous SNPs and SNPs associated to complex traits and diseases, we tested all SNPs except variants present in a 5 MB window surrounding the TSS. In order to remove false positives, we excluded probes mapping to multiple locations according to ReMOAT [37]. To assess significance, we permuted 1000 random probes, corresponding to 1000 genes, 10000 times. As for the other trans analysis, we combined all pvalues obtained, ranked them and increased the genome-wide pvalue threshold until we reached a 5% false discovery rate (corresponding to a pvalue of 2. 0e-9 for non-synonymous SNPs and 5. 4e-10 for SNPs associated to complex traits and diseases). For each gene with an eQTL, we performed linear regression of the best variant on the standard normalized probe expression and kept the residuals. We repeated the association analysis on the residuals using spearman rank correlation and kept any SNPs passing the gene-based permutation threshold obtained during the initial association analysis. We repeated this procedure regressing out all previous best associations until no variants were significant. For each gene with cis-eQTL (s), we computed the variance explained (r2) by the best cis-eQTLs or all independent cis-eQTLs on the standard normalized probe expression using the lm () function in R. We then obtained the heritability explained by dividing the heritability of the probe with the variance explained by the cis-eQTL (s). If the variance explained by the cis-eQTL (s) was greater than the heritability estimate of the probe, the heritability explained was set to 1. We matched each significant variant (cis-eQTLs, trans-eQTLs or GWAS SNPs) with a variant with the same minor allele frequency in our data set (±1%) and distance to the closest gene (±2 kb). Bayesian networks (BN) are directed acyclic graphs where nodes represent random variables and edges represent the conditional dependences among nodes. The direction of the edges between two nodes can be interpreted as causal relationships and allowed to infer causality in genetics studies previously [38]–[40]. Likelihood methods are commonly used to compare different BN and estimate the most likely—that is, the set of causal relationships among the different variables that better agrees with the data. In a BN, every node is associated with a probability distribution and, together with the conditional dependencies represented by the edges, forms the join probability distribution of the network. BN satisfy the local Markov property—that is, each variable is conditionally independent of its non-descendants given its parent variables. The Markov property allows the decomposition of the joint probability distribution of the network into a set of local distributions, which allows to easily calculate the likelihood of a given BN. We used the R package bnlearn [41] to calculate the maximum likelihood of three different networks that we defined using eQTLs as anchors. In the first network (SCT), we fixed the first node as the eQTL genotype with a forward directional edge to the second node (standard normalized cis gene expression) and a second forward directional edge starting from the second node to the third node (standard normalized trans gene expression). For the second network (STC), we fixed the first node as the eQTL genotype with a forward directional edge to the second node (standard normalized trans gene expression) and a second forward directional edge starting from the second node to the third node (standard normalized cis gene expression). For the third network (INDEP), we fixed the first node as the eQTL genotype with a forward directional edge to a node representing the standard normalized cis gene expression and a second forward directional edge starting from the first node to a node representing the standard normalized trans gene expression. Different networks often have different complexities and it is common to use a score that takes into account the network complexity instead of the raw likelihood to compare different networks. We used the Akaike Information Criterion (AIC) score (AIC = 2k-2ln (L), where k is the number of parameters (5 for all models in our case) and L is the maximum likelihood) to compare our networks. To compare how good is a network compared to another, we used the relative likelihood of one network against the other. If we have two networks, N1 and N2 and AIC (N1) ≤AIC (N2), then the relative likelihood of N2 with respect to N1 is defined as: exp ( (AIC (N1) –AIC (N2) ) /2). We kept only networks where the best model was at least ten times more likely than the second best model. In order to have high confidences in our calls, we required that the Causal Inference Test (CIT), described previously [25], also calls the same model as the most likely. The CIT is a semi-parametric method that tests a series of conditions and then provides p-values for the SCT and STC models. If none of them has a pvalue<0. 05, it makes a call for the INDEP model, and if both of them are significant it makes no call. In order to take into account multiple testing with the CIT method and to reduce the number of networks resulting in a “no call” by the CIT, we used Bonferroni corrected pvalues for model calling instead of the nominal pvalue of 0. 05. | Humans differ in their genetic sequences at millions of positions but only a subset of these differences have a functional effect. In order to detect functional genetic differences, we assessed the impact of common genetic variants on gene expression in 869 individuals and discovered that the expression of many genes is affected by common variants in cis or in trans. We show that the effect of some variants on gene expression cannot be detected in other tissues, highlighting the tissue specificity of gene regulation. In addition, we show that variants associated to common diseases are more likely to affect gene expression in cis and in trans. Finally, we show that variants affecting gene expression in cis often affect gene expression in trans, which suggests that the trans effects are due to the cis genes expression. We tested this hypothesis and discovered several cases of genes regulated in trans by a cis regulated gene in a causal manner. This shows that a population-based strategy with a large number of individuals has the potential to detect secondary effects of common variants that can be used to construct short directed regulatory networks. | Abstract
Introduction
Results
Discussion
Methods | genome-wide association studies
genomics
functional genomics
gene regulatory networks
genome analysis
gene expression
genetics
biology and life sciences
computational biology
human genetics | 2014 | Cis and Trans Effects of Human Genomic Variants on Gene Expression | 10,323 | 235 |
Fatal human respiratory disease associated with the 1918 pandemic influenza virus and potentially pandemic H5N1 viruses is characterized by severe lung pathology, including pulmonary edema and extensive inflammatory infiltrate. Here, we quantified the cellular immune response to infection in the mouse lung by flow cytometry and demonstrate that mice infected with highly pathogenic (HP) H1N1 and H5N1 influenza viruses exhibit significantly high numbers of macrophages and neutrophils in the lungs compared to mice infected with low pathogenic (LP) viruses. Mice infected with the 1918 pandemic virus and a recent H5N1 human isolate show considerable similarities in overall lung cellularity, lung immune cell sub-population composition and cellular immune temporal dynamics. Interestingly, while these similarities were observed, the HP H5N1 virus consistently elicited significantly higher levels of pro-inflammatory cytokines in whole lungs and primary human macrophages, revealing a potentially critical difference in the pathogenesis of H5N1 infections. These results together show that infection with HP influenza viruses such as H5N1 and the 1918 pandemic virus leads to a rapid cell recruitment of macrophages and neutrophils into the lungs, suggesting that these cells play a role in acute lung inflammation associated with HP influenza virus infection. In addition, primary macrophages and dendritic cells were also susceptible to 1918 and H5N1 influenza virus infection in vitro and in infected mouse lung tissue.
The influenza pandemic from 1918 to 1919 was the most devastating infectious disease pandemic ever documented in such a short period of time, killing nearly 50 million people worldwide [1]. Unlike the epidemiological profiles of most influenza infections, young adults aged 18–35 yrs old had the highest mortality rate, so much so that the average life expectancy during those years was lowered by 10 years [2]. In 1918, severe destruction of lung tissue observed by pathologists at autopsy was unlike that typically seen in cases of pneumonia [3] and histopathological analysis of lung tissue showed severe tissue consolidation with unique destruction of the lung architecture [3], [4]. Human infections with highly pathogenic avian influenza (HPAI) strains of subtype H5N1 since the first outbreak in 1997 have also been particularly severe for children and young adults [5]–[7]. Assessing pulmonary infiltrates in response to influenza H5N1 virus infection has been difficult due to the lack of autopsy material. The basis for the high morbidity and mortality associated with the 1918 virus and recent H5N1 viruses remains inconclusive based on viral genetic analysis alone and accounts of patient lung pathology provide only qualitative information about the host factors contributing to disease [4], [8], [9]. Great concern about a pandemic caused by a novel avian H5 subtype virus warrants comparative studies to better understand the cellular pathology caused by a pandemic virus and potentially pandemic viruses. Identification and quantification of the inflammatory cell types associated with highly pathogenic respiratory infections represent prospective targets for modulation of host innate immune responses. Recent studies using animal models to investigate the mechanism (s) of severe influenza virulence have implicated the innate immune system in complicating lung tissue recovery [10]–[13]. Mouse models of highly pathogenic (HP) H5N1 [14]–[18] and 1918 [19], [20] influenza virus infection confirm histological observations of severe lung pathology in human patients, however, the types of immune cells present during the peak of lung pathology have not been fully elucidated. Excessive immune cell infiltration during an acute lung injury may impair tissue restoration directly by interfering with gas exchange, or indirectly through the release of soluble immune mediators. In the present study, we determined key immune cellular components in the murine lung following infection with matched H5N1 and H1N1 virus pairs that represent high and low virulence infections of each influenza subtype as previously determined in the mouse model [18], [21]. The two H5N1 viruses used in this study (A/Thailand/16/2004 and A/Thailand/SP/83/2004) were isolated in 2004 from fatal human cases in Thailand but have a differential pathogenic outcome in mice, specifically a low and high mouse lethal does 50 (LD50 = 1. 7 and 5. 6 log10 PFU respectively) [18]. For relevant comparison, we also used a contemporary (non lethal) seasonal H1N1 human isolate from 1991 (A/TX/36/91) and the reconstructed 1918 pandemic virus [21]. A detailed flow cytometry evaluation of lung cells demonstrated that macrophages and neutrophils are the prominent cell types associated with and potentially mediating the severe lung pathology following infection with the highly virulent H5N1 and 1918 viruses. Moreover, inoculation of macrophages and dendritic cells with the HP viruses in vitro or ex vivo reveals that some innate immune cells can themselves serve as targets of viral infection.
Female BALB/c mice were infected intranasally with either highly pathogenic (HP) or low-pathogenic (LP) influenza viruses (Table 1, Methods) based on known LD50' s and phenotypes of disease in mouse [21] and ferret [18], [22] models. As shown in Figure 1, the 1918 (H1N1) and A/Thailand/16/2004 (Thai/16, H5N1) viruses replicated to high titers that were at least 5-fold higher than the respective LP viruses, A/Texas/36/91 (TX/91, H1N1) and A/Thailand/SP/83/2004 (SP/83, H5N1) as early as 1 day post-infection (p. i.) and sustained higher levels of replication than the LP viruses throughout the course of infection. 1918 and TX/91 virus infected lungs reached peak titers on day 3 p. i. and remained elevated for the duration of the study. The HP Thai/16 virus reached peak titers later than the 1918 virus at day 5 pi. while the less mouse-virulent SP/83 virus reached it' s highest titers on day 7 p. i. . Lung virus titers were significantly higher (*p<0. 05) in the 1918 and Thai/16 virus groups compared to TX/91 and SP/83 infected mice at all time points measured. Whole lungs collected (without perfusion, therefore including the bronchoalveolar lavage contents (BAL) ) from both 1918 and Thai/16 virus-infected mice showed an increase in overall tissue cellularity as early as 3 days p. i. (Figure 2A). By day 3 p. i. , lungs infected with the HP influenza viruses (containing between 4. 3–5. 1×107 cells) had nearly twice as many total lung cells than were measured in the HP infection groups just 24 hours previously at day 2 p. i. (2. 1–2. 4×107 cells). Significant differences (*p<0. 05) were observed between HP and LP infection groups in total lung cell number at every time after day 2 p. i. Total lung cell numbers doubled in HP infection groups between days 5 and 7 p. i. and by day 7 p. i. , when viral titers were high for all four viruses, total cell numbers in the lungs of mice infected with either HP H1N1 or H5N1 viruses were 6-fold higher than those in PBS-inoculated mice, and at least 3-fold higher than those found in LP virus-infected lungs (Figures 1 and 2A). On day 7 p. i. , there were as many as 1. 3×108 cells in HP-infected lungs compared with 4. 0–8. 0×107 cells in LP-infected and 1. 8×107 cells in PBS- inoculated lungs. To quantify the immune cell sub-populations responding to viral infection, we next determined the total cell numbers of specific inflammatory cell populations in the infected lungs using flow cytometry (Figure 2B). Compared with PBS- inoculated animals, mice infected with each of the viruses exhibited an increase in the numbers of macrophages (CD11b+, CD11c−, Ly6G/c−, CD4−, CD8−) beginning 3 days p. i. and continuing through day 9 p. i. Strikingly, there were significantly (*p<0. 05) more macrophages in HP virus-infected lungs than in LP virus-infected lungs from days 2 through 9 p. i. (Figure 2B, a). As early as 2 days p. i, there was 1–2 million more macrophages in the lungs of HP-infected lungs compared to LP infected lungs. 1918 and Thai/16 virus–infected lungs had twice as many and nearly 4 times as many macrophages compared to LP viruses at days 3 and 5, respectively. Numbers of lung macrophages peaked in 1918 and Thai/16 infected mice (1. 5 and 1. 2×107 cells, respectively) at day 7 p. i. before waning as demonstrated at day 9 p. i. An increase of at least twice as many neutrophils (CD11b+, CD11c−, Ly6G/c+, CD4−, CD8-) was observed as early as 1 day p. i. in all infection groups, compared with neutrophils numbers in PBS-inoculated mice. On day 2 p. i. there was on average (n = 3 mice) over four hundred thousand more neutrophils in the lungs of 1918 virus inoculated mice than TX/91 inoculated mice. Significant differences in neutrophil populations (*p<0. 05) between the HP and the LP virus groups emerged on day 3 p. i. and were sustained at each subsequent time point measured (Figure 2B, b). On day 3 p. i. , more than twice as many neutrophils were found in HP compared to LP infected lungs. Between days 3 and 7 p. i. , lungs infected with HP viruses displayed a three fold increase in neutrophil numbers. At the peak of neutrophil infiltration on day 7 p. i, there was 4–8 million more neutrophils in the lungs of HP-infected animals compared to LP virus-infected mice. Although numbers of dendritic cells (CD11b−, CD11c+, Ly6G/c−, CD4−, CD8−) (Figure 2B, c) and CD4 + T (CD11b−, CD11c+, Ly6G/c−, CD8−) and CD8+ T (CD11b−, CD11c+, Ly6G/c−, CD4−) cells (Figure 2B, d and e) in all groups of infected mice increased slightly compared with numbers in PBS-inoculated animals during the course of infection, no significant differences were found between HP and LP infection groups, suggesting that these cell populations are not major contributors to the histopathological consolidation observed in lungs during HP H5N1 virus infections in mice [16]. We next determined specific immune cell sub-populations given by percent of the total lung leukocyte population in order to reveal differential population dynamics amongst the immune cell populations measured in these studies (Table 2). Strikingly, macrophage populations by day 3 pi, represented 24% and 24. 4% (nearly one-quarter) of the total gated lung leukocytes of mice infected with 1918 and Thai/16 (HP) viruses, respectively, which was significantly higher than the frequency of macrophages detected in TX/91 and SP/83 virus-infected mice from days 2 though 9 p. i. (ˆ p<0. 005). Neutrophil populations were elevated in all infection groups compared to mock levels beginning 1 day p. i (Figure 2, panel B) however percentages of neutrophils in HP virus infected mice were significantly higher (* p<0. 05) than those in LP virus-infected mice beginning day 2 p. i. and levels remained elevated at each subsequent time point measured (Table 2). In contrast, dendritic cell populations as percent of the total leukocyte population declined over the course of infection with HP viruses while percentages increased in LP infections, peaking at days 3 and 5 p. i. (Table 2). Significant differences in percent dendritic cells between HP and LP infection groups were observed day 3 p. i. and at all other subsequent time points (* p<0. 05). Percentages of CD4+ T cells in the lungs decreased in all infection groups but no significant differences were observed in CD4+ or CD8+ T cell population dynamics between HP and LP infection groups. Together, these results indicate that macrophages and neutrophils are responsible for the majority increase in total lung cell numbers following infection with the 1918 and HP H5N1 influenza viruses. To better understand potential influences on immune cell population dynamics in the lung following HP influenza virus infection as demonstrated in Figure 2B and Table 2, we analyzed 17 chemokines and cytokines in the lungs of mice on days 1 and 4 p. i and report data here on 6 of the analytes that revealed significant differences among the viruses tested (Figure 3). These earlier time points were chosen in an effort to understand the temporal relationship with lung immune cell infiltration in 1918 and Thai/16 infections (Figure 2) and day 4 p. i. is a time when significant differences were observed in lung virus titers between HP and LP infection groups (Figure 1). As shown in Figure 3, on day 1 p. i. TX/91 infected lungs exhibited higher titers of IL-1α, IFN-γ, and KC compared to 1918 virus-infected lungs though levels of MIP-1α, MCP-1, and IL-6 were higher in 1918 virus infected lungs. Cytokine levels were similar between the H5N1 viruses at day 1 p. i. . In contrast, lung tissue cytokine and chemokine levels at day 4 p. i. were higher among 1918 and Thai/16 virus-infected mice compared to those infected with the subtype-matched LP (TX/91 and SP/83) counterpart viruses for all cytokines and chemokines measured. Protein levels of the potent monocyte chemoattractant MCP-1 were 10-fold higher in 1918 and H5N1 virus-infected lungs than TX/91 virus-infected lungs, whereas levels of the MIP-1α chemokine were notably elevated in Thai/16 virus-infected lungs. KC (the mouse equivalent of human IL-8) [20], [23] levels were 10-fold higher in Thai/16 and 3-fold higher in 1918 virus infected lungs compared to TX/91 virus-infected lungs with SP/83 levels similar to 1918 infected lungs. The HP viruses were also potent inducers of IL-1α and IFNγ. IL-6, (a generally pro-inflammatory cytokine) was also elevated in 1918 and H5N1 virus-infected lungs compared to TX/91 levels. As shown in Figure 3, significant differences (ˆ p<0. 05) were measured between 1918 and TX/91 virus infected lungs on day 4 p. i. for each cytokine and chemokine, accentuating important immunological differences in the immune response to the H1N1 pandemic virus versus a contemporary H1N1 virus. Due to the increased presence of macrophages in the lungs following HP influenza virus infection, replication of paired H1N1 and H5N1 viruses (Table 1) was assayed over time in primary human PBMC-derived macrophages and mouse lung macrophages to address whether these cells are specific targets of viral infection and can productively replicate HP viruses. Mouse lung macrophages were harvested from naive BALB/c mice and infected in vitro (MOI = 0. 1, Figure 4A) as described. While the HP 1918 and Thai/16 viruses exhibited a slight increase in log titer very early after inoculation (13–24 hrs p. i.), overall these viruses did not replicate efficiently compared to the growth kinetics of these viruses observed in lung epithelial cells [22], [24]. At 13 hrs p. i. there was 18 fold higher 1918 and Thai/16 virus released from infected cells than TX/91 and SP/83 virus-inoculated cultures and nearly 54 times more 1918 and Thai/16 virus than TX/91 or SP/83 virus at 24 hrs p. i. While virus titers in the 1918, TX/91 and SP/83 infection groups declined after the 48 hr time point, the Thai/16 infected macrophages maintained virus titers and these differences were statistically significant (ˆ p<0. 05) at 72 hrs p. i. . The lack of prolific replication of these four viruses in primary mouse lung macrophages was confirmed further when a higher MOI (1. 0) was utilized (data not shown). Human macrophages also supported replication of all four viruses (Figure 4B, MOI = 0. 1). At 48 hrs p. i. 1918 and Thai/16 infected macrophages exhibited 180 times higher virus titers than TX/91 and SP/83 infected cultures. Interestingly, 1918 virus-infected macrophages exhibited a higher baseline titer soon following infection that was found to be statistically significant when compared to the other infection groups († p<0. 05). In summary, while human macrophages are a target of viral replication and support replication well, mouse lung macrophages support low levels of 1918 and Thai/16 virus production early following infection. Additional experimentation with primary human macrophages revealed that levels of pro-inflammatory cytokines were higher for H5N1-infected cells than either 1918 or TX/91 virus infected cells (48 hrs p. i. , MOI = 0. 1, Figure 5). Significant differences in cytokine levels were observed between H5N1 and H1N1 virus infected macrophages in all cytokines measured (*p<0. 05) except IL-8 where all four viruses elicited similar levels of this chemokine. Interestingly, the 1918 virus elicited similar cytokine responses as to TX/91 inoculated cells in every cytokine measured even though 1918 virus titers at this time point post-infection were 2. 5 logs greater than TX/91 infected macrophages (Figure 4B). Thai/16 infected cultures elicited at least a 2 fold greater cytokine response than the SP/83 virus infected cultures in every cytokine measured except IL-8 and MCP-1 chemokines where Thai/16 levels were only slightly higher than SP/83 levels. Because of their important role in antigen presentation to T cells, we also assessed the ability of primary dendritic cells to productively replicate pandemic H1N1 and human H5N1 viruses (Figure 6A and B). While the HP Thai/16 H5N1 virus replicated slowly over time in cultured mouse lung dendritic cells (Figure 6A), these cells failed to productively replicate 1918, TX/91 and SP/83 viruses. Significant differences in virus titers were measured between HP (1918 and Tha/16) infected cells and TX/91 and SP/83 infected cells at 24 hrs p. i. (Figure 6A, ˆ p<0. 05). Significant differences in virus titers was also observed between Thai/16 inoculated cultures and the other infection groups at 48 and 72 hrs p. i. (* p<0. 05). Similar to infected human macrophages, these trends were conserved when primary human dendritic cells were the target of infection with these viruses (Figure 6B). Significant differences in virus titers was observed between Thai/16 inoculated cultures and the 1918, TX/91, and SP/83 infection groups at 24,48 and 72 hrs p. i. (* p<0. 05). To determine if innate immune cells are being productively infected in vivo, macrophages and dendritic cells were purified from lungs of infected mice and cultured for infectious virus. Lungs from infected (3 days p. i.) mice were harvested and ex vivo cultures containing either lung macrophages or dendritic cells were sampled for infectious virus over a 65 hour time period. While the seasonal influenza isolate, TX/91 virus was not produced from either macrophages (CD11b+) or dendritic (CD11c+) cells, the 1918 pandemic virus as well as the two human H5N1 isolates were released into the culture supernatant (Figure 7), indicating that these cells are being productively infected in the mouse lung. In CD11b+ (macrophages) cell cultures (Figure 7A), the 1918 and Thai/16 virus infected cells released more infectious virus over time than the LP SP/83 virus infected culture. In CD11c+ (dendritic) cell cultures, the Thai/16 virus infected cultures released more infectious virus than the 1918 and SP/83 virus infected cultures (Figure 7B). These data further demonstrate that mouse lung macrophages and dendritic cells are susceptible to highly pathogenic influenza virus infection in the lung tissue.
Lung consolidation has been described as a pathological feature of severe influenza virus infection caused by the 1918 pandemic virus and H5N1 viruses in humans [9], [25], [26] as well as in animal models [11], [12], [18]. Using a detailed flow cytometry evaluation, the current study set out to characterize differences in the cellular innate immune response in the mouse lung following highly pathogenic (HP) or low pathogenicity (LP) influenza virus infections. Lungs from mice infected with the HP 1918 H1N1 virus and a recent H5N1 human isolate (A/Thailand/16/04 (Thai/16) ) exhibited a significant increase in cellularity in comparison to the LP seasonal H1N1 isolate, A/Texas/36/91 (TX/91). Significant differences in titer between HP and LP virus-infected mice were observed as early as day 1 post-inoculation (p. i.), likely forecasting the dramatic increase in immune cell infiltration in the lungs of these virus-infected mice. Interestingly at day 7 p. i. when peak lung cellularity was observed in HP virus infection groups, differences in virus titers between paired subtype viruses (1918 compared to TX/91 and Thai/16 compared to SP/83) were minimal and were limited to a maximum difference of 1 log (Figure 1), indicating a failure by the immune system to clear the viral infection. Lung cellularity was further investigated by characterizing the comprising immune cell sub-populations. We observed a significant increase in macrophages and neutrophils early following infection with the 1918 and Thai/16 viruses, and their sustained presence in the lung tissue mark a distinction between HP and LP influenza virus infection. These data show that virus replication in the lungs of HP influenza infections are sustained at high levels the first week of infection regardless of the high numbers of immune cells present in the tissue. Although the current study did not determine whether these cells are playing an antiviral role against the HP viruses, it has been previously shown that neutrophils and macrophages assist in the clearance of influenza virus early during infection; these cells appear to be capable of only partially reducing the virus load in the lung despite their presence at high numbers [20]. We also observed a decrease in the percentage of lung–associated dendritic cells and T cell lymphocytes during HP influenza virus infection. A decrease in the number of circulating lymphocyte populations has also been previously observed in the peripheral blood of humans and mice infected with H5N1 viruses [5], [9], [10], [15], [20], [27]. The precise mechanism of leukocyte depletion during H5N1 infections is not well understood, but evidence of apoptosis in the spleen and lungs of HP H5N1-infected mice detected in situ suggests a mechanism for cell loss [16]. Influenza virus growth in the respiratory epithelium and the subsequent release of chemotactic proteins from those cells may encourage the increased presence of macrophages and neutrophils [28], [29]. Macrophages and neutrophils can secrete chemokines and cytokines that can act in an autocrine fashion which in turn can promote the increased migration of those cells and other leukocytes into the lung tissue [30]. Elevated levels of certain chemokines and cytokines have been associated with high viral load and severe disease in H5N1 virus infected patients [26], [31]. These studies show that infection with HP 1918 and Thai/16 H5N1 viruses result in elevated amounts of pro-inflammatory chemokines and cytokines in the lungs of mice day 4 post-infection compared with TX/91 and SP/83 infected mice, a time point that correlates with rising but significantly different lung virus titers between infection groups. Elevated levels of the chemokines MCP-1 [32] and MIP1-α were observed among H5N1 and 1918 virus infected mouse lungs. Although, MIP-1α does not appear to be critical for virus replication and spread in the mouse model [10], this chemokine exhibits a variety of pro-inflammatory activities including macrophage and neutrophil recruitment and has been associated with fatal outcomes in human H5N1 virus infections [31]. Lungs infected with the 1918 pandemic and Thai/16 H5N1 viruses also exhibited significantly higher levels of IFN-γ on day 4 p. i compared to their subtype-paired LP virus counterparts. IFN-γ is known to mediate the increased production of nitric oxide [33] which can subsequently result in the recruitment of more neutrophils and macrophages. Higher levels of IL-6 were measured in 1918 and Thai/16 virus infected lungs, supporting observations obtained with the 1997 H5N1 viruses [10] and has been correlated with systemic illness symptoms and fever in experimental human TX/91 infections [34]. By directly measuring cytokine protein levels, these data provide confirming evidence of a heightened lung cytokine response to 1918 and H5N1 infection in mice [13]. Interestingly, while we reveal marked similarities between the HP 1918 and Thai/16 viruses in overall lung cellularity, virus growth and patterns of immune cell sub-population dynamics over time, Thai/16 virus infection consistently resulted in higher levels of chemokines and cytokines both in mouse lungs and human macrophages. Although it has been shown recently by two independent research groups that the lack of key cytokines (through the use of single cytokine gene knockout mice) had no effect on the overall disease outcome or virus replication among H5N1 virus-inoculated mice [10], [35], the present results along with data from others continue to indicate that pro-inflammatory cytokines correlate with disease outcome [36]. It is thought that these immune mediators do not act singularly in vivo and it will be critical to reveal these concerted interactions (both locally in the lung and systemically) to further our understanding of the pathogenesis of H5N1 infection in animal models and in human patients. Macrophages and dendritic cells play a fundamental role in the lung at all stages of influenza virus infection [20], [37], [38]. We have provided evidence regarding the higher replication efficiency of HP influenza viruses in primary human macrophages and dendritic cells, a property that has also been demonstrated previously in other primary cells [24], [39]. Mouse macrophages were also susceptible to virus infection in vitro, however they did not support productive replication to the level observed in primary human monocyte-derived macrophages (Figure 4B) or lung epithelial cells [21], [24]. However, the cytopathic effects (estimated by visual examination of monolayers) among HP virus-infected mouse and human macrophages was observed in a shorter period of time compared to LP virus infected cells. Interestingly, the 1918 virus exhibited higher baseline titers in human and mouse macrophages as early as 2 hrs p. i. compared to the TX/91 H1N1 virus or the H5N1 viruses, indicating a curious property of this pandemic virus. While the importance of the binding properties of the HA molecule has been demonstrated elsewhere extensively [22], [40], [41], further resolution of this interesting finding and its immunological importance deserves further experimentation. The role of other cell surface molecules on innate immune cells such as the mannose receptor in HP influenza infection should be investigated [42], [43]. The higher viral replication of HP viruses in dendritic cells also correlated with the severe pulmonary disease observed in mice. We also demonstrated definitively that these cells are targets of infection ex vivo by highly pathogenic influenza viruses like the 1918 pandemic virus and recent H5N1 isolates (Figure 7). Thus, it appears that macrophages and dendritic cells may contribute to the pathogenesis of HP virus infection due to their susceptibility to influenza virus infection. An inability to mount an adaptive immune response due to direct infection of important innate immune cells such as macrophages and dendritic cells may be a critical difference in host outcome during influenza virus infection [44]. This coupled with the phenomenon of T cell depletion in infected mice [17] and human patients [9], [31] may allow for uncontrolled viral replication. While reducing viral load through anti-viral intervention remains the best treatment option for H5N1 patients, therapies that moderate immunopathology may help to reduce the high case fatality rate currently associated with virus infection [45].
All in vitro and in vivo experiments were performed under the guidance of the U. S. National Select Agent Program in negative pressure HEPA-filtered biosafety level laboratory (BSL-3+) enhanced laboratories and with the use of a battery powered Racal HEPA-filter respirator and according to Biomedical Microbiological and Biomedical Laboratory procedures. Influenza viruses used in these experiments included the reconstructed A/South Carolina/1/18 virus (H1N1) [21], A/Texas/36/91 (H1N1), A/Thailand/16/2004 (H5N1), and A/Thailand/SP/83/2004 (H5N1). These viruses were chosen as pairs within subtypes due to their respective fifty percent lethal dose (LD50) titers in mice and ferrets (Table 1). Low pathogenicity in this manuscript refers to the non-lethal phenotype of the seasonal H1N1 TX/91 virus and the low virulence SP/83 H5N1 isolate [18], [21]. The Thai/16 and SP/83 viruses differ from each other in 13 amino acids in 7 proteins and this sequence comparison has been published previously [18]. The A/Texas/36/91 and H5N1 viruses were grown in 10 day old embryonated hen' s eggs and the 1918 viruses grown in MDCK cells. All virus stocks were titered by plaque assay on MDCK cells prior to mouse infections. Human peripheral blood monocytes (PBMC' s) were obtained by Histopaque (Sigma-Aldrich, St. Louis, MO) density gradient centrifugation of whole blood donated by healthy donors aged 20–40 yrs old without history of influenza vaccination in the past year. Whole blood was obtained through an approved protocol by both the Emory University Institutional Review Board (IRB) and CDC IRB (Emory University Hospital Blood Bank is an FDA-accredited Blood Bank). Human monocytes were obtained by negative selection column enrichment (Miltenyi Biotech, Auburn, CA) yielding approximately 90% CD14+ purity as determined by FACS analysis. For development of macrophages, monocytes were cultured at 37°C in 6 well plates in Macrophage SFM media (Gibco, Grand Island, NY) with 20% heat inactivated autologous serum for 7 days in the presence of GM-CSF (1000 U) before infection [46]. Human macrophages cultured in this manner typically displayed classical morphology with the phenotype: CD11blow, CD11clow, HLA-DRlow, CD14low, CD40low, CD80low, CD83low, CD86- (Figure 4C and D). For the development of dendritic cells (DC' s), monocytes were grown in RPMI (Gibco) with 20% heat inactivated autologous serum for 10 days in the presence of IL-4 (1000U) as previously described [46]. Dendritic cells developed in this manner typically displayed a classical morphology with the presence of dendritic processes with the phenotype: CD11blow, CD11chigh, HLA-DRhigh, CD14low, CD40high, CD80high, CD83high, CD86high (Figure 6B). To obtain primary mouse lung macrophages and dendritic cells, lungs from naïve mice were removed and tissue disrupted as described above through the use of collagenase digestion and cell suspensions prepared. Macrophages (CD11b+) and dendritic cells (CD11c+) were extracted from contaminating cells by selection on magnetic columns (Miltenyi Biotech, Auburn, CA). Cells were washed twice with media containing 20% FCS (Macrophage SFM for macrophages (Gibco) or RPMI (Gibco) for dendritic cells) and cultured for 24 hours before in vitro infection. Primary mouse lung macrophages typically displayed the phenotype: CD11bhigh, CD11c-, MHCIIhigh, CD40high, CD80high, CD83high (Figure 4A and B). Primary mouse lung dendritic cells typically displayed typical morphology with dendritic extensions and the phenotype: CD11b-, CD11chigh, MHCIIhigh, CD40high, CD80high, CD83high (Figure 6A). Primary human and mouse cells were washed 3× with serum free growth media and infected for 1 hour with viruses (Table 1) at a multiplicity of infection (MOI) of 0. 1. Following infection, cells were washed 3× with serum free growth media and 1 ml SFM media, containing 1 µg/ml of TPCK-treated trypsin (Sigma-Aldrich), was placed into the wells. Virus growth was measured over time in triplicate wells for each experiment and titered in duplicate by standard plaque assay on MDCK cells. All macrophage and dendritic cell data reflects at least three independent experiments (Figures 4 and 6). Cytokine levels produced from infected human macrophages (MOI = 0. 1) were quantitated 48 hrs p. i. by BioPlex assay (Figure 5). Escherichia coli lipopolysaccharide (LPS, 100 ng, Sigma-Aldrich) and Poly I/C (100 ng, Sigma-Aldrich) were used as positive control stimulants. All animal research was conducted under the guidance of CDC' s Institutional Animal Care and Use Committee and in an Association for Assessment and Accreditation of Laboratory Animal Care International- accredited facility. 8–10 week old female BALB/c mice (Harlan, Indianapolis, IN) were anesthetized with Avertin [20] (Sigma-Aldrich) and infected intranasally (i. n.) with 50 µl of 102 PFU of influenza viruses prepared in phosphate buffered saline (PBS). Avertin was chosen as the anesthetic because it provides consistent mouse infections with the viruses used in these studies. Using the sublethal (102 PFU) inoculum, 1918 and Thai/16 virus infected mice survive a prolonged disease course allowing for the measurement of the influx of inflammatory cells into the lung tissue during a full course (∼7–9 days) of influenza virus infection. At indicated times post-infection (n = 3 mice per virus group) mice were euthanatized and exsanguinated. Lungs were removed from individual mice without PBS perfusion and included total lung cell counts included cells located in the bronchoalveolar airways. Perfusion was not possible in many cases of HP influenza virus infection due to the presence of microvascular hemorrhage. We obtained similar results when performing these lung cell quantitation assays with or without lung perfusion in LP virus infected mice and therefore did not introduce this variable in our high containment laboratory. Whole lung cell suspensions were prepared in Dulbecco' s minimal essential media (DMEM) with 20% fetal calf serum following collagenase-DNase treatment and manual disruption [47]. Red blood cells were removed by lysis buffer treatment (Sigma-Aldrich). Total viable lung cell number was determined for each mouse by trypan blue exclusion on a hemocytometer. For the ex vivo experiment, macrophages and dendritic cells were isolated from the lungs of infected BALB/c mice. Two or three mice were infected i. n. with 102 PFU of each of the four viruses described in this study (Table 1). Three days post-inoculation, lungs were removed without perfusion and cell suspensions were prepared as described above. Lungs were pooled from mice in each virus infection group. Macrophages and dendritic cells were isolated by positive selection on CD11b+ or CD11c+ MACS columns. Columns containing bound immune cells were washed extensively (5x) and CD11b+ or CD11c+ cells were eluted off the magnetic columns and cultured in 6-well plates in 5 ml of RPMI containing 5% BSA. Supernatants were collected at the indicated times and virus content was determined in a standard plaque assay on MDCK cells. Lung cell suspensions were incubated with anti-Fc block (anti-mouse CD16/CD32) to reduce non-specific antibody binding for 10 min. prior to staining for 1 hr with fluorophore-conjugated antibodies (BD Biosciences, San Diego, CA) specific for immune cell populations according to standard protocols [48] and included: CD11b-PE (pan-macrophage), CD11c-APC (pan-dendritic cell), Ly6G/C-FITC (neutrophil), CD4-PE and CD8-APC T cell markers (Table 2, Figure 2). Cells were washed twice with PBS and fixed overnight at 4°C with 2% paraformaldehyde. Samples were safety tested for infectious virus and removed from the BSL3+ laboratory. Flow cytometry was performed on a FACSAria flow cytometer (BD Biosciences). To further characterize primary mouse and human macrophage and dendritic cells we utilized the following fluorescently conjugated antibodies for flow cytometric analysis: MHC II (I-A/I-E) -PE, HLA-DR-PE, CD14-FITC, CD40-FITC, CD80-FITC, CD83-APC, and CD86-APC (BD Biosciences) (Figures 4 and 6). At various times post-infection (n = 3 mice per virus group) lungs were removed and stored at −70°C until virus and cytokine levels could be quantified. Lungs were homogenized individually in 1 ml PBS. Virus was titered from clarified lung homogenates by standard plaque assay on MDCK cells in duplicate and titers are reported as plaque forming units per ml PBS (PFU/ml, Figure 1). Cytokine protein levels were measured (day 4 post-infection (p. i.) ) by the Bioplex Protein Array system [49] (Bio-Rad, Hercules, CA) using beads specific for mouse G-CSF, IL-1α, IL-1β, IL-3, IL-6, IL-9, IL-12 (p40), IL-12 (p70), IL-13, Eotaxin, TNFα, RANTES, KC, MIP1-α, MIP-1β, MCP-1, and IFN-γ. Cytokine protein levels were measured according to the manufacturers instructions by fluorescently conjugated monoclonal antibodies in duplicate against a standard curve (Figures 3 and 5). Statistical significance of differences between experimental groups was determined through the use of the unpaired, non-parametric Student' s t test. Values of p<0. 05 were considered significant. | Patients who succumbed to influenza during the 1918 pandemic had severe lung pathology marked by extensive inflammatory infiltrate, indicating a robust immune response in the lung. Similar findings have been reported from H5N1-infected patients, raising the question as to why people expire in the presence of a strong immune response. We addressed this question by characterizing the immune cell populations in the mouse lung following infection with the 1918 pandemic virus and two H5N1 viruses isolated from fatal cases. Our data shows excessive immune cell infiltration in the lungs contributing to severe consolidation and tissue architecture destruction in mice infected with highly pathogenic (HP) influenza viruses, supporting the histopathological observations of lung tissue from 1918 and H5N1 fatalities. We found that certain cells of the innate immune system, specifically macrophages and neutrophils, increase significantly into the mouse lung shortly following HP virus infection. Interestingly, lung macrophages and dendritic cells were shown to be susceptible to 1918 and H5N1 virus infection in vitro or ex vivo, suggesting a possible mechanism of immunopathogenesis. Identification of the precise inflammatory cells associated with lung inflammation will be important for the development of treatments that could potentially enhance or modulate host innate immune responses. | Abstract
Introduction
Results
Discussion
Materials and Methods | virology/animal models of infection
virology | 2008 | H5N1 and 1918 Pandemic Influenza Virus Infection Results in Early and Excessive Infiltration of Macrophages and Neutrophils in the Lungs of Mice | 9,641 | 280 |
Enterococcus faecium has become a nosocomial pathogen of major importance, causing infections that are difficult to treat owing to its multi-drug resistance. In particular, resistance to the β-lactam antibiotic ampicillin has become ubiquitous among clinical isolates. Mutations in the low-affinity penicillin binding protein PBP5 have previously been shown to be important for ampicillin resistance in E. faecium, but the existence of additional resistance determinants has been suggested. Here, we constructed a high-density transposon mutant library in E. faecium and developed a transposon mutant tracking approach termed Microarray-based Transposon Mapping (M-TraM), leading to the identification of a compendium of E. faecium genes that contribute to ampicillin resistance. These genes are part of the core genome of E. faecium, indicating a high potential for E. faecium to evolve towards β-lactam resistance. To validate the M-TraM results, we adapted a Cre-lox recombination system to construct targeted, markerless mutants in E. faecium. We confirmed the role of four genes in ampicillin resistance by the generation of targeted mutants and further characterized these mutants regarding their resistance to lysozyme. The results revealed that ddcP, a gene predicted to encode a low-molecular-weight penicillin binding protein with D-alanyl-D-alanine carboxypeptidase activity, was essential for high-level ampicillin resistance. Furthermore, deletion of ddcP sensitized E. faecium to lysozyme and abolished membrane-associated D, D-carboxypeptidase activity. This study has led to the development of a broadly applicable platform for functional genomic-based studies in E. faecium, and it provides a new perspective on the genetic basis of ampicillin resistance in this organism.
Enterococci rank third overall as causative agents of healthcare-associated infections [1], [2]. Up to the late 1980s, Enterococcus faecalis was responsible for practically all enterococcal infections, but starting from the 1990s nosocomial infections with E. faecium became more frequent. Currently E. faecium causes approximately 40% of all enterococcal infections that are acquired during hospital stay [2]–[4]. Clinical isolates of E. faecium have rapidly accumulated antibiotic resistance genes, including those for clinically important antibiotics such as ampicillin and vancomycin, which leads to treatment failure and increased mortality rates [2], [5]–[7]. In the USA, nosocomial infections caused by ampicillin-resistant E. faecium (ARE) were first detected in the 1980s and the resistance rates were steadily increasing up to 80% of E. faecium isolates in the 1990s [8], [9]. Vancomycin-resistant E. faecium (VRE) also emerged in the late 1980s and increased rapidly during the 1990s [9], [10]. Currently, VRE is widespread among clinical E. faecium strains in North America, but less common in hospital-acquired infections in Europe [11]. Ampicillin resistance has spread much further and it is currently being reported in over 80% of clinical E. faecium isolates from all over the world [1], [2] (European Antimicrobial Resistance Surveillance Network: http: //www. ecdc. europa. eu/en/activities/surveillance/EARS-Net/Pages/index. aspx). In addition to ARE and VRE, the emergence of E. faecium strains that are resistant to new classes of antibiotics is challenging the few remaining therapeutic options [12]–[14]. Thus, the development of new anti-enterococcal agents may become critical for the successful treatment of infections caused by this multi-drug resistant organism in the future. The intrinsic resistance to β-lactam antibiotics of enterococci was reported 60 years ago, soon after the introduction of penicillin in the early 1940s, when enterococci were found to be considerably less susceptible to β-lactams than streptococci [15]. Mutations in the high-molecular weight class B penicillin-binding protein 5 (PBP5) have been considered the main cause for the resistance to β-lactams in E. faecium. Upregulated expression of pbp5 and/or mutations in the 3′ end of the gene lead to a further reduced susceptibility to ampicillin [16]–[18]. However, several studies have suggested that the high minimum inhibitory concentration (MIC) of ampicillin against E. faecium is not exclusively due to the presence of low-affinity PBP5 but also to other genes or mechanisms that remain to be identified [19], [20]. Recently, Mainardi et al. [21], [22] showed in a spontaneous mutant that was obtained in the laboratory by selection on agar media containing ampicillin, that the D, D-transpeptidase activity of the PBPs could be bypassed by a β-lactam resistant L, D-transpeptidase (Ldtfm) that catalyses the formation of 3→3 cross-links between peptidoglycan side chains instead of the classical 4→3 cross-links. A D, D-carboxypeptidase, termed DdcY, is an important component in the L, D-transpeptidase mediated pathway of peptidoglycan cross-linking. However, the ddcY gene is only present in a small proportion of E. faecium isolates [23], again suggesting that additional ampicillin resistance determinants in E. faecium remained to be identified and characterized. Genome-wide studies of clinical E. faecium isolates have long been hampered by a lack of appropriate genetic tools. In this study, we describe the construction of a high density mariner transposon mutant library and the development of a powerful tool for functional genomics, termed Microarray-based Transposon Mapping (M-TraM), in E. faecium. By comparing the mutant library following growth in the presence or absence of ampicillin, we identified a compendium of genes affecting the sensitivity to ampicillin. Targeted mutants of the identified genes with predicted roles in cell wall synthesis were generated for further characterization, which resulted in the identification of several intrinsic ampicillin resistance determinants in E. faecium. These ampicillin resistance determinants may serve as targets for the development of novel antimicrobial therapeutics.
To attempt genome-wide transposon mutagenesis of the E. faecium genome, we constructed the transposon delivery plasmid pZXL5. As shown in Figure S1, this plasmid was composed of a Gram-positive thermo-sensitive replicon, a gentamicin resistant mariner transposon with two outward-facing T7 promoters, a nisin-inducible mariner transposase, a ColE1 replicon and a cat gene. The sequence of pZXL5 was determined by Sanger-sequencing of both DNA strands (Baseclear; Leiden, The Netherlands) and was deposited in GenBank (GenBank Accession Number: JQ088279). Using pZXL5, we have produced a transposon mutant library in E. faecium strain E1162, an ampicillin-resistant clinical isolate from a bloodstream infection, for which a draft genome sequence has previously been determined [24]. The randomness of the transposon insertions and the absence of multiple transposon insertion events were determined by randomly selecting 17 mutants from the library and carrying out Southern blot hybridizations, using a fragment of the transposon as a probe (Figure S2A), as well as inverse PCR and sequence analysis to determine the location of the transposon insertion point (Figure S2B). The results showed that each mutant carries a single transposon inserted in the genome, and that the transposon was distributed in different loci in the 17 mutants. PCR footprinting was performed to estimate the genome-wide coverage of transposon insertions in the mutant library (Figure 1). An outward-facing primer was designed based on the mariner transposon sequence. The other PCR primer was designed for three target genes, ddl (which encodes a D-alanine∶D-alanine ligase that is essential for bacterial cell wall biosynthesis [25]), esp (which is non-essential and encodes a large surface protein involved in biofilm formation and infection [26]–[28]) and nox (which is a non-essential gene encoding a predicted NADH oxidase). Genomic DNA isolated from the pooled mutant library was used as a template. A range of products can be amplified by these primers, each corresponding to a transposon insertion mutant in the library. If a gene is essential for survival, its transposon insertion mutants should not be present in the library after overnight growth, and consequently no PCR products should be amplified in the corresponding size range. As expected, no PCR product was detected within the ddl gene (Figure 1) while many PCR bands were found in esp and nox at intervals of less than 100 bp, indicating the transposon insertions covered the nonessential genes of the genome at high density. Furthermore, mapping the transposon insertion sites to the complete genome sequence of E. faecium Aus0004 [29] revealed that transposon insertions were randomly distributed over the genome of this strain and not confined to a specific chromosomal region (data not shown). To establish whether pZXL5 has broad applicability in E. faecium we attempted to transform four other clinical E. faecium isolates from different geographic origins (Table S1) with pZXL5 using our optimized electroporation protocol. All four strains were efficiently transformed with transformation efficiencies ranging between 110 and 105 transformants per µg DNA (Figure S2C). We then continued to generate transposon mutant libraries in two of these strains (Figure S2D). These observations show that the transposon mutagenesis approach that we initially developed for strain E1162, can also be used for functional genomics in other clinical E. faecium isolates. As described in the Materials and Methods and in Figure 2, we developed a technique to track the presence of all mutants in the library by simultaneously mapping the transposon insertion sites using microarray hybridization. We termed this technique M-TraM for Microarray-based Transposon Mapping. To validate the reproducibility of M-TraM, we independently grew two aliquots containing approximately 107 cells from the library in 20 ml of BHI broth. After 20 hours of culturing at 37°C, genomic DNA was isolated from the two replicate cultures and used for the generation of cDNA. The cDNA samples were labeled with Cy3 and Cy5 respectively and hybridized to a microarray that was designed using the E. faecium E1162 genome sequence. The result showed that M-TraM is highly reproducible, with a correlation coefficient of 0. 94 between the two independent experiments (Figure 2C). To identify genes required for ampicillin resistance, we grew the pool of mutants in the presence or absence of a subinhibitory concentration (20 µg ml−1) of ampicillin, and used M-TraM to determine which mutants were selectively lost during culturing in the presence of ampicillin. Eleven genes belonging to a variety of functional categories were identified to be involved in ampicillin resistance (Table 1). Four genes involved in cell wall biogenesis were identified and we decided to further focus on these genes. The EfmE1162_0447 (ddcP) and EfmE1162_1886 (ldtfm) genes were predicted to encode a D-alanyl-D-alanine carboxypeptidase (D, D-carboxypeptidase, DdcP) and beta-lactam-insensitive peptidoglycan transpeptidase (L, D-transpeptidase, Ldtfm), respectively. Previous studies have shown that a different D, D-carboxypeptidase (DdcY) and Ldtfm were able to bypass the D, D-transpeptidase activity of the PBPs by forming 3→3 cross-links instead of the classical 4→3 cross-links thereby conferring resistance to β-lactams [21], [23]. DdcP and DdcY only share 13. 7% amino acid identity and have a completely different protein domain architecture as DdcY is a β-lactam-insensitive VanY-type carboxypeptidase [23], [30], while DdcP belongs to the family of low-molecular-weight (LMW) PBPs [30] (Figure S3). The ddcY gene is absent from 24 (including E1162) of the 29 E. faecium genomes available (on 9 January 2012) at NCBI Genomes. The ddcP gene is conserved in all 29 E. faecium genomes. EfmE1162_0975 (pgt) was predicted to encode a glycosyl transferase group 2 family protein which is 63% identical to GltB, a protein that was proposed to be involved in glycosylation of cell wall teichoic acid in serotype 4b Listeria monocytogenes [31]. EfmE1162_2487 (lytG) was predicted to encode an exo-glucosaminidase that could be acting as a peptidoglycan hydrolase involved in cell wall lysis, remodeling and cell division [32]. An overview of the protein domain architecture and predicted cellular localization of DdcP, Ldtfm, Pgt and LytG is provided in Figure S3. Like the ddcP gene, the ldtfm, pgt and lytG genes are present in all the 29 E. faecium genomes as well, suggesting these genes are part of the E. faecium core genome. Notably the E. faecium ampicillin resistance determinants ddcP, ldtfm and pgt do not have homologs (defined here by proteins with >30% amino acid identity) in E. faecalis. It should be noted that the L, D-transpeptidase from E. faecalis that was biochemically characterized by Magnet et al. , [33] has not been experimentally linked to β-lactam resistance in E. faecalis and is only remotely related (26% amino acid identity) to ldtfm. To validate the results of the M-TraM screen and to further characterize the role of the identified genes in ampicillin resistance, we constructed targeted mutants in the ddcP, ldtfm and pgt genes, which were identified with the most significant P-values (Table 1) and have predicted functions in cell wall biogenesis. We also generated a targeted mutant in the lytG gene of which the inactivation could confer hyper-resistance to ampicillin, as suggested by the M-TraM data. Targeted deletion mutants of ddcP, pgt, and lytG were generated using a novel Cre-lox-based system for the generation of markerless mutants in E. faecium that we developed as part of this study (Figure S4). For ldtfm no double cross-over mutant could be constructed and instead a single cross-over mutant was constructed using the pWS3 vector [34]. Mutants of ddcP, ldtfm, and lytG were also complemented in trans and the ampicillin resistance of E1162 (wild-type), the mutants and the complemented strains was determined. The pgt mutant could not be complemented as constructs containing the pgt gene could not be transformed to either Escherichia coli or E. faecium, presumably due to toxicity of the gene product. In the absence of ampicillin we did not detect significant differences in growth speed or cell density upon entry into stationary phase between wild-type and the mutant strains (Figure S5). When these strains were grown in BHI with 20 µg ml−1 ampicillin, the ΔddcP mutant was dramatically affected in its growth (Figure 3A). Growth of the ldtfm: : pWS3 and Δpgt mutants was also poorer than wild-type (Figure 3B and 3C). The in trans complemented strains of the ΔddcP and ldtfm: : pWS3 strains could fully or partially restore the ampicillin resistance to wild-type levels (Figure 3A–3C). The ΔlytG mutant had a growth rate that was similar to the parental strain' s (0. 998±0. 007 h−1 for ΔlytG vs. 0. 989±0. 018 h−1 for E1162) but could grow to slightly higher optical densities (Figure 3D). The in trans complemented lytG mutant exhibited a significantly lower growth rate (0. 859±0. 017 h−1) in exponential phase (Figure 3D). The empty vector had no effect on the growth of the mutants in BHI supplemented with 20 µg ml−1 ampicillin (data not shown). MICs of ampicillin against the wild-type E1162 and ΔddcP, ldtfm: : pWS3, Δpgt and ΔlytG strains were determined by microdilution in cation-adjusted Muller-Hinton broth as 43,8, 16,27 and 43 µg ml−1, respectively, which is in accordance with the growth performance of the mutants in BHI with 20 µg ml−1 ampicillin (Figure 3). The seemingly contradictory observation that the ldtfm: : pWS3 mutant can grow in BHI medium supplemented with ampicillin at a concentration above the MIC can be explained by the differences in growth media used and the approximately 4-fold larger inoculum size used in the growth experiments in BHI compared to the inoculum size used in the MIC determinations. The MIC of vancomycin was determined to be 0. 5 µg ml−1 for all strains. We used microarray-based transcriptome analysis on exponentially growing (OD660 = 0. 3) E. faecium E1162 cultures in BHI medium with or without 20 µg ml−1 ampicillin to identify genes that are regulated by the exposure to sub-MIC levels of ampicillin. Compared to the untreated control, only sixteen genes were identified to be differentially regulated between the two conditions and none of these genes were upregulated more than 2. 1-fold in the presence of ampicillin (Table S2), indicating that E. faecium does not require major transcriptional rearrangements to cope with the presence of sub-inhibitory levels of ampicillin. None of the genes that were identified by M-TraM were identified to be differentially regulated by the presence of ampicillin, which indicates that the identified ampicillin resistance determinants are constitutively expressed, even in the absence of ampicillin. After the identification of ddcP, ldtfm and pgt as ampicillin resistance determinants, we studied the susceptibility of the wild-type strain E1162 and its mutants to another compound targeting the cell wall, i. e. lysozyme, which is one of the most important antimicrobial enzymes of the host innate immune system. Lysozyme kills Gram-positive bacteria by enzymatic lysis of the bacterial cell wall [35]. The results demonstrated that deletion of pgt had no significant effect on lysozyme resistance, while ΔddcP and ldtfm: : pWS3 mutants were significantly more sensitive to lysozyme challenge than the wild-type strain (Figure 4). The in trans complemented strains of ΔddcP and ldtfm: : pWS3 could restore the resistance to lysozyme. Hence, ddcP and ldtfm contribute not only to β-lactam resistance but also to the resistance against the peptidoglycan-hydrolyzing enzyme lysozyme. Of all novel ampicillin resistance determinants identified in this study, the ddcP gene contributes most to ampicillin resistance in E. faecium E1162. The DdcP protein was annotated as a D-alanyl-D-alanine carboxypeptidase but a functional study confirming this activity has not been performed. To confirm its predicted function, we determined the D-alanyl-D-alanine carboxypeptidase activity in cellular extracts of E1162, the ΔddcP mutant and the in trans complemented strain ΔddcP+ddcP. As shown in Figure 5, the D-alanyl-D-alanine-carboxypeptidase activity of ΔddcP membrane extracts was completely abolished. In the complemented ΔddcP+ddcP strain enzymatic activity was restored in the membrane fraction, revealing that DdcP is responsible for D-alanyl-D-alanine carboxypeptidase activity in E. faecium E1162. The D, D-carboxypeptidase activity was approximately 5 fold lower in the cytoplasmic fractions than in the membrane fractions (data not shown), strongly suggesting that the DdcP protein is associated with the membrane. When E. faecium E1162 was grown in the presence of 20 µg ml−1 ampicillin, all D-alanyl-D-alanine carboxypeptidase activity was undetectable, which is in accordance with the designation of DdcP as a LMW-PBP. Notably, pbp5 was not identified in this M-TraM screen even though this gene has been implicated in high-level ampicillin resistance in E. faecium [16]–[18]. Our failure to identify pbp5 by M-TraM may partially be explained by the presence of an AluI restriction site on one of the two microarray probes for pbp5. The other microarray probe maps to an internal region of the pbp5 gene and is located on a small AluI restriction fragment with a window of 158 nucleotides for the transposon insertion. Previous genome-wide studies using transposon mutagenesis have shown that transposon insertions might fail to fully inactivate the target genes. The failure is commonly found with insertions situated near either end of a gene, but has also been observed with internal insertions [36], [37], after which genes might be capable of intracistronic complementation. We constructed targeted insertional mutants of pbp5, ddcP and ldtfm by the insertion of the plasmid pWS3 5′ to the gene probes and close to the central region of the gene. Consistent with the M-TraM results and the phenotypes of the markerless mutants, the insertional mutants of ddcP (data not shown) and ldtfm (Figure 3B) were sensitized to ampicillin. Surprisingly the insertional mutation in pbp5 did not significantly affect the sensitivity to ampicillin (Figure 6), implying that insertional mutation (by either a plasmid or a transposon) could not fully disrupt pbp5. Therefore a markerless deletion mutant for pbp5 was generated to completely abolish the function of this gene using the Cre-lox system described above. To our knowledge, a targeted mutant of the pbp5 gene has not been generated previously in E. faecium. The pbp5 mutant was unable to grow in cultures containing 20 µg ml−1 ampicillin and the ampicillin resistant phenotype could be partially be restored by in trans complementation of Δpbp5 with the pbp5 gene (Figure 6). The MIC of ampicillin for Δpbp5 was determined by broth microdilution to be only 0. 2 µg/ml. Our results revealed that the relative contribution of pbp5, ddcP, ldtfm and pgt to ampicillin resistance in E. faecium E1162 can be summarized as pbp5≫ddcp>ldtfm>pgt. The pbp5 mutant was also assayed for its survival in the presence of lysozyme, but no significant difference was found with the parental strain E1162 (data not shown).
Ampicillin resistance in E. faecium has emerged in the late 1970s and has spread rapidly since [38]. Practically all clinical isolates are currently resistant to ampicillin. The resistance to β-lactams of E. faecium complicates the treatment of infections with this organism, particularly when resistance to other antibiotics has also been acquired. The goal of the research described here was to identify genes involved in ampicillin resistance in E. faecium in a high-throughput fashion. We developed a system for the generation of a large random transposon mutant library in E. faecium, coupled to a microarray-based screening approach (termed M-TraM) to simultaneously monitor the relative fitness of individual mutants undergoing selection by growth in the presence of ampicillin. The lack of appropriate genetic tools has long been a bottleneck for the studies of E. faecium. In this study, we constructed a random high-density transposon mutant library in E. faecium, developed a powerful screening technique to track transposon mutants and adapted the Cre-lox [39] recombination system to construct targeted, markerless mutants in E. faecium, which enabled us to perform high-throughput genome-wide analysis and specific targeted investigations in a clinical E. faecium isolate. When E. faecium is exposed to ampicillin the D, D-transpeptidase activity of PBPs is inhibited, with the exception of the low-affinity PBP5, which can catalyze the last cross-linking step of the D, D-transpeptidation pathway of cell wall assembly [17]. This implies that any D, D-transpeptidase activity conferred by genes other than pbp5, is not essential for ampicillin resistance. Consequently, no genes encoding D, D-transpeptidases were identified in this study. However, in our transposon mutant screen, we identified a novel D-alanyl-D-alanine carboxypeptidase (DdcP) that plays an important role in resistance to ampicillin. DdcP is the only gene that is responsible for D-alanyl-D-alanine carboxypeptidase activity during exponential growth of strain E1162. The observation that the enzymatic activity of DdcP is abolished in the presence of ampicillin is in accordance with the prediction that DdcP is a LMW-PBP. The in vivo functional roles of LMW-PBPs are relatively poorly understood but they are generally not essential for survival and are thought to contribute to peptidoglycan-remodelling in both Gram-positive and Gram-negative bacteria [40]. Deletion of the gene encoding a LMW-PBP can lead to an increased sensitivity towards β-lactam antibiotics [41]–[42]. Although a mechanistic understanding for this sensitive phenotype is currently lacking, it has been proposed that these LMW-PBPs may enzymatically inactivate β-lactams or, alternatively, save other PBPs from inactivation by sequestration of the β-lactams to the LMW-PBPs [41]. Further functional characterization of DdcP in E. faecium, also involving strains that have higher and lower resistance towards ampicillin than strain E1162, is needed to identify the exact mechanism by which DdcP contributes to ampicillin resistance in E. faecium. Previous biochemical studies have indicated that Ldtfm is a crucial component of the β-lactam-insensitive L, D-transpeptidation pathway, which catalyzes the cross-links of tetrapeptides [21]–[23]. However, genetic evidence for the role of Ldtfm in ampicillin resistance was so far lacking. Here, we have constructed a mutant, confirming the role of this pathway in ampicillin resistance in E. faecium. The identification of pgt as an ampicillin resistance determinant in E. faecium suggests that wall teichoic acid is involved in β-lactam resistance which is in line with a similar observation in methicillin-resistant Staphylococcus aureus (MRSA) strains [43]. None of the mutations had an effect on vancomycin resistance of E. faecium, presumably because tetrapeptide precursors for peptidoglycan crosslinks are present at levels that are too low to confer resistance to this antibiotic in wild-type and the mutant strains. We did not generate mutants or performed functional analyses to confirm the function of the other genes that were identified in our M-TraM screening (Table 1). However, it seems likely that at least some of these genes also contribute to ampicillin resistance in E. faecium, in particular EfmE1162_2490 (predicted to function as an NAD- or NADP-dependent oxidoreductase) and EfmE1162_0256 (possibly acting as a zinc-dependent β-lactamase), because the transposon mutants in these genes appear to have a similar loss in fitness in the presence of ampicillin as the transposon mutants in ldtfm and pgt (Table 1). The lytG gene was the only gene that was identified in our screen which, upon inactivation by a transposon, appeared to result in a hyper-resistant phenotype. However, this phenotype could not be confirmed in a markerless deletion mutant of lytG, which had the same MIC for ampicillin as the wild-type strain. The observed slightly higher optical density reached by ΔlytG as compared to wild-type and the slower growth rate of the in trans complemented lytG mutant may suggest that there is a subtle role for lytG in ampicillin resistance in E. faecium but further experiments would be required to exactly determine its role. The recent emergence of E. faecium as a major nosocomial pathogen can be explained by the acquisition of genes that contribute to colonization or infection [26], [44] and by the acquisition of resistance to antibiotics, particularly to ampicillin and vancomycin. Interestingly, resistance against ampicillin and vancomycin emerged predominantly in E. faecium while both resistances are virtually absent in E. faecalis [8]. This study provides insights into the genetic basis of intrinsic β-lactam resistance in E. faecium and identified the ampicillin resistance determinants DdcP, Ldtfm and Pgt in this organism. All three genes are conserved in E. faecium but absent from E. faecalis, indicating that E. faecium possesses more innate β-lactam resistance determinants than E. faecalis. This observation supports the concept that E. faecium has a higher potential to develop high-level β-lactam resistance than E. faecalis, thereby explaining the faster emergence of ampicillin resistance in E. faecium than in E. faecalis [1], [2], [45]. We have identified several novel mechanisms, besides the low-affinity penicillin-binding protein PBP5, that contribute to ampicillin resistance in E. faecium. These proteins could serve as targets for the development of novel therapeutics against this multi-resistant organism. Our study showed that DdcP and Ldtfm also contribute to resistance to lysozyme of E. faecium. An inhibitor of these proteins may thus provide the dual benefit of compromising resistance to the innate immune system as well as enhancing antibiotic susceptibility. In the course of this study we have also developed a number of novel genetic tools for E. faecium allowing for genome-wide analysis of this bacterium. Further functional genomic-based studies to understand the mechanisms involved in colonization, infection and antibiotic resistance of this important nosocomial pathogen are now a realistic opportunity for future research.
E. faecium and E. coli strains used in this study are listed in Table S1. The ampicillin-resistant E. faecium strain E1162 was used throughout this study. This strain was isolated from a bloodstream infection in France in 1996 and its genome has recently been sequenced [24]. Unless otherwise mentioned, E. faecium was grown in brain heart infusion broth (BHI; Oxoid) at 37°C. The E. coli strains DH5α (Invitrogen) and EC1000 [46] were grown in Luria-Bertani (LB) medium. Where necessary, antibiotics were used at the following concentrations: chloramphenicol 4 µg ml−1 for E. faecium and 10 µg ml−1 for E. coli, gentamicin 300 µg ml−1 for E. faecium and 25 µg ml−1 for E. coli, spectinomycin 300 µg ml−1 for E. faecium and 100 µg ml−1 for E. coli, and erythromycin 50 µg ml−1 for E. faecium (with added lincomycin at 50 µg ml−1) and 150 µg ml−1 for E. coli. All antibiotics were obtained from Sigma-Aldrich (Saint Louis, MO). Growth was determined by measuring the optical density at 660 nm (OD660). The transposon delivery plasmid, pZXL5, was constructed in several steps. The Gram-positive lacZ gene of pCJK72 [47] was PCR (Accuprime DNA polymerase, Invitrogen) amplified using the primers pCJK72_PstI_lacZ_F and pCJK72_KpnI_lacZ_R (primer sequences are listed in Table S3) and cloned into pTEX5500ts [48] between PstI and KpnI sites to create pZXL1. A fragment containing the chloramphenicol resistance (Chlr) cassette, the lacZ gene and a gram-positive thermosensitive origin of replication (repAts, functional at 30°C, but not at 37°C) from pZXL1 was PCR amplified using the pZXL1_EcoRI _cm_ori_F and pZXL1_EcoRI _cm_ori_R. Meanwhile, another fragment containing a nisin inducible mariner transposase C9 and the nisRK genes (encoding a two-component system required for the transcriptional activation of the transposase gene in the presence of nisin), and the ColE1 origin of replication was PCR amplified from pCJK55 [47] using the primers pCJK55_EcoRI_tps_F and pCJK55_EcoRI_tps_R. These two fragments were digested with EcoRI and then ligated together to generate pZXL2. To construct a mariner transposon [49], the 5′ and 3′ ITRs of Himar1-mariner were amplified from pMMOrf [50] using the primer pMMOrf_SacII_ITR that resulted in two SacII recognition sites at both ends of the amplified DNA. The amplified fragment was cloned into pGEM-T Easy (Promega) forming pGEM-ITR. A gentamicin-resistance cassette was PCR amplified from pAT392 [51] using primers pAT392_EheI_T7_genta_F and pAT392_EheI_T7_genta_R, resulting in a gentamicin-resistance cassette with outward-facing T7 promoters on both ends of the cassette, which allows for the generation of RNA products corresponding to the regions flanking the site of transposon integration in genomic DNA. This fragment was digested with EheI and cloned into a SmaI site present between the 5′ and 3′ ITRs in pGEM-ITR, thereby forming the transposon cassette. This transposon cassette was then cut out with SacII and subcloned into the SacII site of pZXL2 producing pZXL3. This vector was electroporated to E. faecium E1162 but pZXL3 was found to be able to replicate at 37°C in E. faecium E1162 and blue/white screening using lacZ proved to be ineffective (data not shown). We therefore replaced the repAts and lacZ gene of pZXL3 by the repAts from pAW068 [52]. To this aim, a fragment of pZXL3 containing the nisin inducible mariner transposase and the transposon cassette was PCR amplified by primers pZXL3_BfrI_tn_F and pZXL3_BfrI_tn_R. Another fragment carrying the repAts and Chlr cassette was amplified from pAW068 by PCR using primers pAW068_BfrI_cm_ori_F and pAW068_BfrI_cm_ori_R. After digestion with BfrI these two PCR products was ligated together, resulting in the generation of pZXL5. All restriction enzymes were obtained from New England Biolabs. Electrotransformation of the different E. faecium strains (Table S1) with the plasmid pZXL5 was performed according to previously described methodologies [26], [34] with optimizations in preparing electrocompetent cells and the cell-plasmid mixture. To obtain the electrocompetent cells, overnight cell culture from BHI was diluted 1000 fold in 25 ml BHI supplemented with 1% of glycine and 200 mM sucrose and again grown overnight at 37°C. Cells were then resuspended in same volume of pre-warmed BHI supplemented with 1% glycine and 200 mM sucrose and incubated at 37°C for 1 hour. Cells were washed three times with ice-cold wash buffer (500 mM of sucrose and 10% glycerol), and resuspended with 1. 25 ml ice-cold wash buffer. A 100 µl aliquot of the cell suspension was mixed with 0. 1–1 µg of plasmid and transferred into an ice-cooled electroporation cuvette (2-mm gap) and kept on ice for 20 minutes before electroporation. Gentamicin-resistant transformants were grown overnight in BHI broth supplemented with 300 µg ml−1 gentamicin and 10 µg ml−1 chloramphenicol at the permissive temperature of 28°C, after which 100 µl (approximately 108 viable cells) of this overnight culture were inoculated in 200 ml of pre-warmed BHI broth supplemented with gentamicin and 25 ng ml−1 nisin and grown overnight at the non-permissive temperature of 37°C with shaking at 150 rpm. Subsequently, 100 µl of this culture was transferred to 200 ml of fresh pre-warmed BHI broth and similarly grown overnight without nisin. Cultures were then stored at −80°C in BHI broth containing 50% (v/v) glycerol in 1 ml aliquots as mutant library stocks. To evaluate the randomness and coverage of transposition, we performed Southern blot analysis, identified the sites of transposon insertion and used PCR footprinting. Southern blot analysis was performed as described previously [34]. Genomic DNA of 17 arbitrarily picked gentamicin-resistant colonies from the library was isolated using the Wizard Genomic DNA Purification kit (Promega), digested with HaeIII and BamHI. The probe consisting of a 414 bp fragment within the gentamicin-resistance gene was amplified from pZXL5 by PCR, using the primer pair genta_probe_F/R. To map the sites of transposon insertion, genomic DNA of 17 mutants from the library was digested with HaeIII and then self-ligated, forming circular DNA. Loci in which the transposon had inserted were amplified using the transposon-specific primer pair IPCR_HaeIII_R/F with AccuPrime DNA polymerases (Invitrogen) with the following conditions: 94°C for 1 min; 32 cycles of 94°C for 18 sec, 53°C for 30 sec, 68°C for 10 min; and 68°C for 7 min. Sequencing of the PCR product was performed using the primer IPCR_HaeIII_R and/or IPCR_HaeIII_F. PCR footprinting was conducted on genomic DNA of mutant library as described elsewhere [53] with a transposon-specific primer, ftp_tn, and gene-specific primers, ftp_ddl, ftp_nox or ftp_esp, respectively. Transposon insertion mapping was based on the previously published method of Genomic Array Footprinting (GAF) [54]. Because we observed that T7 polymerase will transcribe E. faecium genomic DNA aspecifically (data not shown), it was required to modify GAF to specifically enrich for the junction sites of the transposon and the flanking E. faecium DNA. Genomic DNA from mutant libraries was isolated using the Wizard Genomic DNA Purification kit (Promega), digested with AluI (New England Biolabs) and then purified on a Qiagen QIAquick PCR Purification column (Qiagen). 200 ng of the digested DNA was self-ligated by the Quick Ligation Kit (New England Biolabs) in a reaction volume of 20 µl. This ligation reaction was directly used as template for PCR amplification of the transposon–chromosome junctions with primer pair IPCR_AluI_F and IPCR_AluI_R in a reaction volume of 200 µl, using AccuPrime Taq DNA polymerases High Fidelity (Invitrogen) with the following conditions: 94°C for 1 min; 26 cycles of 94°C for 18 sec, 56. 5°C for 30 sec, 68°C for 50 sec; and 68°C for 7 min. PCR products were purified using Qiagen QIAquick PCR Purification Kit. After purification, 200–500 ng DNA was redigested by AluI and used for in vitro transcription (IVT) in a volume of 20 µl using the T7 MEGAshortscript kit (Ambion) at 37°C for 6 hours. The RNA was first treated with DNase (Ambion) and then purified with the MEGAclear Kit (Ambion). 5–10 µg of the purified RNA was used for generating labeled cDNA using the FairPlay III Microarray Labeling Kit (Agilent Technologies) as described in the manufacturer' s protocol. Samples of both conditions (grown in BHI and BHI with 20 µg ml−1 ampicillin) were labeled with Cy3 or Cy5. Dyes were swapped between samples to minimize the effect of dye bias. Microarray hybridizations were carried out using the Gene Expression Hybridization Kit (Agilent) following the manufacturer' s instructions, using 60 ng of labeled cDNA. The experiment was performed with four biologically independent replicates. The microarrays used in this study were custom-made E. faecium E1162 microarrays using Agilent' s 8×15 K platform. Probes were designed by Agilent' s eArray server. As probes 60-mer oligonucleotides were designed on coding sequences (CDS) only. A total of 2650 CDS are covered by 2 probes (98. 4% of the total number of CDS in the E. faecium E1162 genome sequence; NCBI accession number NZ_ABQJ00000000), which were spotted in duplicate. A total of 11 CDS are covered by a single probe (0. 4% of the total number of CDS) and these probes were spotted in quadruplicate. For 33 CDS no probes could be designed. Microarray data were extracted and normalized using Agilent Feature Extraction Software Version 10. 7. 1. 1 (FE 10. 7. 1). Statistical differences in hybridization signals between the conditions were analyzed using Cyber-T [55] (http: //cybert. microarray. ics. uci. edu/). Probes exhibiting Bayesian P-value<0. 001 were deemed statistically significant. A gene with two identical probes or all four probes meeting this criterion were classified as significantly selected during exposure to ampicillin. To carry out identification of genes required for ampicillin resistance, aliquots containing approximately 107 CFU from the mutant pool stored at −80°C were diluted 1 to 1000 in 20 ml of BHI broth or BHI broth with 20 µg ml−1 ampicillin. Cells were grown at 37°C for 8 hours, after which 1 ml of the bacteria cultures were spun down and used for the extraction of genomic DNA, which was then further processed as described above. The same protocol (except that the cells were grown for 20 hours instead of 8 hours) was used to determine the reproducibility of the M-TraM screening in which two independent mutant libraries cultures were mapped using the approach described above. For this study, we developed a new method to construct markerless mutants in E. faecium based on the Cre-lox recombination system [39]. The 5′ and 3′ flanking regions (approximately 500 bp each) of the target genes were PCR amplified with the primers in Table S3. The two flanking regions were then fused together by fusion PCR (generating an EcoRI site between both fragments) and cloned into pWS3. Then a gentamicin-resistant cassette was PCR amplified from pAT392 using primers pAT392_EcoRI_lox66_genta_F and pAT392_EcoRI_lox71_genta_R, resulting in a gentamicin-resistant cassette flanked by lox66 and lox71, which allows for the deletion of the gentamicin-resistant cassette in the presence of Cre recombinase. This fragment was digested with EcoRI and cloned into the EcoRI site that was generated between the 5′ and 3′ flanking regions in the pWS3 construct and then electrotransformed into E. faecium as previously described [26], [48]. A transformant containing the plasmid was grown overnight in BHI broth at 30°C supplemented with gentamicin. The cell culture was then diluted 10,000-fold in prewarmed BHI broth and grown at 37°C overnight without antibiotics. The cells were then plated on BHI agar plates with gentamicin and incubated at 37°C. Colonies were then restreaked on BHI agar plates with spectinomycin and BHI agar plates with gentamicin, respectively. The gentamicin-resistant but spectinomycin-susceptible colonies were supposed to be marked deletion mutants and checked by PCR (Table S3). To remove the marker and obtain the markerless mutants a Cre cassette was cut from pRAB1 [56] by digestion with PstI and SacI, blunted by Quick Ligation Kit (New England Biolabs) and cloned into the EcoRV site of pWS3 producing pWS3-Cre, which was subsequently electrotransformed into the marked mutants. Spectinomycin-resistant transformants containing pWS3-Cre were then grown overnight in BHI broth at 30°C supplemented with spectinomycin and then diluted 10000 fold in pre-warmed BHI broth and grown at 37°C overnight without antibiotics. These cultures were plated on BHI agar plates and incubated at 37°C for 18–24 h. Single colonies were then restreaked on BHI agar with spectinomycin, BHI agar with gentamicin and BHI agar without antibiotics. The colonies that were susceptible to both gentamicin and spectinomycin resulted from a recombination event catalyzed by Cre and subsequent loss of the thermosensitive plasmid, resulting in a markerless deletion mutant of the gene of interest. This was verified by PCR and sequencing. Insertional mutagenesis was performed as previously described [34]. Internal DNA fragments of target genes were PCR amplified using primers listed in Table S3, cloned to a Gram-positive thermosensitive plasmid and electrotransformed into E. faecium as previously described [26], [48]. After electrotransformation, the cells were recovered for 2 hours at 30°C, after which the cells were plated on BHI plates supplemented with 300 µg ml−1 spectinomycin at 30°C to select for transformants. Spectinomycin-resistant colonies were picked and grown overnight in 200 ml of BHI broth at an elevated temperature (37°C) to cure the plasmid. The cells were then plated on BHI agar plates with spectinomycin at 37°C. Single-cross-over integrations into the target genes were verified by PCR with a pWS3-specific primer, check_pWS3, and a gene-specific primer (Table S3). Plasmids for the in trans complementation of the ddcP, ldtfm, lytG and pbp5 mutants were produced by PCR amplification of the genes using the primers listed in Table S3. PCR products were ligated into the downstream region of PnisA promoter of pMSP3535 [57]. The resulting plasmids were introduced into the appropriate host strains by electroporation as described above. A BioScreen C instrument (Oy Growth Curves AB, Helsinki, Finland) was used to monitor effects of ampicillin on bacterial growth. Wild-type E. faecium, mutants and in trans complemented strains were grown overnight in BHI and BHI containing appropriate antibiotics. Cells were inoculated at an initial OD660 of 0. 0025 into 300 µl BHI and BHI with ampicillin 20 µg ml−1 and 1 µg ml−1. The cultures were incubated in the Bioscreen C system at 37°C with continuous shaking, and absorbance of 600 nm (A600) was recorded every 15 min for 9 hours. Each experiment was performed in triplicate. MIC of ampicillin of the wild-type and mutants were determined in triplicate by broth microdilution in cation-adjusted Muller-Hinton broth as previously described [58]. E. faecium E1162 was incubated in BHI broth and BHI broth supplemented with 20 µg ml−1 ampicillin for 18 hours. Cultures were then diluted to OD660 0. 025 in 20 ml of prewarmed BHI broth and BHI broth containing 20 µg ml−1 ampicillin respectively, and grown until OD660 0. 3. Cells were centrifuged for 12 seconds at 169000 g at room temperature, and pellets were flash frozen in liquid N2 prior to RNA extraction. RNA was isolated using TRI Reagent (Ambion) according to the manufacturer' s protocol. RNA quantity and quality was determined by spectrophotometry (Nanodrop 1000, Thermo Scientific, Wilmington DE, USA) and by Bioanalyzer 2100 analysis (Agilent). Labeling of 5 µg of total RNA, hybridization and data analysis were performed as described above. Genes for which all four probes exhibited a Bayesian P<0. 001 in Cyber-T [55] were deemed differentially expressed. To compare the lysozyme sensitivity of the parental strain E1162, the mutant strains and in trans complemented strains, overnight cell cultures were diluted 100 fold in fresh BHI and grown to OD660 0. 5. Two ml of the cell cultures were harvested by centrifugation. The pellets were resuspended in 1 ml phosphate buffered saline (PBS; NaCl 137 mM; 2. 7 mM KCl; 10 mM Na2HPO4; 2 mM KH2PO4; pH 7. 4) as negative control and in 1 ml PBS containing 0. 5 mg ml−1 lysozyme. After a 30-minute incubation at 37°C, cells were washed with PBS and resuspended in 1 ml of PBS. Survival of the strains was determined following serial dilution and plating on BHI agar plates. The experiment was performed in triplicate and statistical analysis of the data was performed using a two-tailed Student' s t-test. The enzymatic activities in the enterococcal extracts of wild-type, ΔddcP and ΔddcP+ddcP were assayed as described previously with slight modifications [59], [60]. In short, strains were grown until an OD600 of 0. 7. Bacteria were then harvested by centrifugation and lysed by treatment with lysozyme at 37°C for 1 hour followed by sonication. The membrane fraction was then pelleted by ultracentrifugation (100,000 g, 45 min). The supernatant (cytoplasmic fractions) was collected and the pellet (membrane fractions) was resuspended in 0. 1 M phosphate buffer (pH 7. 0) and both fractions were assayed for D, D-carboxypeptidase activity [59], [60]. The amounts of D-Ala released from the pentapeptide (Ala-D-γ-Glu-Lys-D-Ala-D-Ala, Sigma-Aldrich) by D, D-carboxypeptidases were determined by using D-amino acid oxidase and horseradish peroxidase in a colorimetric assay. The microarray data generated in this study have been deposited in the ArrayExpress database (http: //www. ebi. ac. uk/arrayexpress) under accession numbers E-MEXP-3501 for the M-TraM screening for ampicillin resistance determinants, E-MEXP-3502 for the assay of the reproducibility of the M-TraM procedure and E-MEXP-3564 for the transcriptome analysis data. | Enterococcus faecium has emerged as an important nosocomial pathogen around the world. Clinical E. faecium isolates are often resistant to multiple antibiotics, thereby complicating therapeutic interventions. However, the molecular mechanisms that contribute to the recent emergence of E. faecium as a nosocomial pathogen of major importance are only poorly understood, which is, at least partially, due to the lack of appropriate genetic tools for the study of this organism. Here, we developed a systematic genome-wide strategy, based on transposon mutagenesis and microarray-based screening, to identify E. faecium genes that contribute to ampicillin resistance. We also adapted the Cre-lox recombination system to construct targeted, markerless mutants in E. faecium. These tools enabled us to perform both high-throughput genome-wide analysis and specific targeted investigations in a clinical E. faecium isolate. We comprehensively identified, confirmed, and characterized a compendium of genes affecting the sensitivity to ampicillin in E. faecium. The identified intrinsic ampicillin resistance determinants are highly conserved among E. faecium, indicating that this organism has a high potential to evolve towards ampicillin resistance. These ampicillin-resistance determinants may serve as targets for the development of novel antimicrobial therapeutics. | Abstract
Introduction
Results
Discussion
Materials and Methods | medicine
functional genomics
genetic mutation
microbiology
gene function
bacterial diseases
genome analysis tools
enterococcus infection
bacterial pathogens
infectious diseases
microbial pathogens
biology
gram positive
mutagenesis
genetic screens
genetics
genomics
genetics and genomics | 2012 | Genome-Wide Identification of Ampicillin Resistance Determinants in Enterococcus faecium | 13,584 | 331 |
Health institutions may choose to screen newly admitted patients for the presence of disease in order to reduce disease prevalence within the institution. Screening is costly, and institutions must judiciously choose which patients they wish to screen based on the dynamics of disease transmission. Since potentially infected patients move between different health institutions, the screening and treatment decisions of one institution will affect the optimal decisions of others; an institution might choose to “free-ride” off the screening and treatment decisions of neighboring institutions. We develop a theoretical model of the strategic decision problem facing a health care institution choosing to screen newly admitted patients. The model incorporates an SIS compartmental model of disease transmission into a game theoretic model of strategic decision-making. Using this setup, we are able to analyze how optimal screening is influenced by disease parameters, such as the efficacy of treatment, the disease recovery rate and the movement of patients. We find that the optimal screening level is lower for diseases that have more effective treatments. Our model also allows us to analyze how the optimal screening level varies with the number of decision makers involved in the screening process. We show that when institutions are more autonomous in selecting whom to screen, they will choose to screen at a lower rate than when screening decisions are more centralized. Results also suggest that centralized screening decisions have a greater impact on disease prevalence when the availability or efficacy of treatment is low. Our model provides insight into the factors one should consider when choosing whether to set a mandated screening policy. We find that screening mandates set at a centralized level (i. e. state or national) will have a greater impact on the control of infectious disease.
Hospital associated infections represent a major cause of morbidity and mortality [1]. Disease screening can facilitate efforts to reduce disease prevalence through the treatment of infected patients or reducing transmission within a healthcare setting. Because screening and treatment are expensive, institutions must choose a level of screening to balance the associated costs with the benefits of lower disease prevalence. However, in a setting where patients flow back and forth between different institutions, the prevalence in one institution will directly affect others that it shares patients with. This implies that the screening decisions of one institution can impact the prevalence and, consequently, the optimal screening levels of other institutions with a shared patient population. One can imagine a situation in which a single institution could “free-ride” off of a low prevalence rate induced by neighboring institutions that rigorously screen and treat patients. In such a strategic environment, it is important to determine wether or not individual health institutions will choose a screening level that is socially optimal. The economic literature is full of examples where strategic settings induce suboptimal decision making; in some cases, only a social planner can induce the socially optimal decision. In the case of screening, this might imply setting a screening policy at a national, state or regional level. Indeed, a handful of countries outside of the US have already implemented nationally mandated screening policies for diseases such as MRSA [2] [3], and policy makers and health professionals in the US have also called for universal or mandatory screening of MRSA [4] [5]. Such screening mandates arise because policy makers believe that individuals or healthcare organizations may not make optimal screening decisions on their own. The goal of this paper is to provide insight into the strategic disease-screening process along with the factors that should be considered when choosing an optimal screening policy. To do so, we develop a theoretical model of disease screening and analyze behavior across a wide range of disease and screening parameters. Properly studying this strategic screening problem requires a game theoretic model that can incorporate the underlying dynamics of disease transmission. In order to capture both disease dynamics and strategic decision making, we draw on an existing body of literature that has sought to combine epidemiologic compartmental models of disease transmission with economic models of decision making [6]–[12]. Specifically, this study develops a theoretical model by incorporating a Susceptible-Infected-Susceptible (SIS) compartmental model into a game-theoretic setting where institutions must choose the number of patients to screen, given the screening decisions of other institutions that they share patients with. Using this framework, we can analyze how incentives are influenced by changing the number of decision makers, i. e. making institutions more or less autonomous in setting their own screening policy. In the concept of a game, changing the level of autonomy is effectively equivalent to changing the number of players in the game while the underlying population remains fixed. Our model extends previous literature, which has analyzed how the number of decision makers influences optimal decision making in disease settings [9], by allowing for (1) a variable number of decision makers in a fixed patient population, (2) movement of patients between health care institutions and (3) variability of treatment efficacy. Thus, we develop a general model intended to be broadly applicable to the disease-screening process.
To model the spread of disease within a population composed of different healthcare institutions and a non-institutional community, we build an SIS compartmental model of disease. In this model, individuals within each subpopulation are assumed to exist in one of two states: (1) susceptible, the disease is not present, and (2) infected, the disease is present. Susceptible individuals may receive the disease and become infected, whereas infected individuals may transmit the disease or recover and return to the susceptible population. We can represent the proportion of a subpopulation k that is infected by, where and are the number of infected and susceptible individuals, respectively. We are interested in two types of subpopulations, the various healthcare institutions that screen patients and the community. We assume there are N different institutions and a single community. Let and represent the disease prevalence in a particular health institution i and the community, respectively. There are two ways in which screening may influence disease prevalence, namely treatment and transmission. Let be the percentage of the patient population that institution i decides to screen. If treatment for a disease exists, screening may identify infected individuals that can be treated and returned to the susceptible state. We denote the daily treatment recovery rate within an institution as a function of its screening level by. This function will depend on the efficacy of treatment and screening. The only restrictions we place on its functional form are that it is assumed to be nondecreasing in the screening level, , and must not be greater than the rate of screening, . This first assumption implies that greater screening does not lead to less recovery from treatment, and the second implies that patients are screened before they are treated. We must also account for the fact that individuals may recover from the disease naturally without treatment. Let λ denote the natural recovery rate of the disease. These two parameters imply that the recovery from the disease in a health institution i is given by. In addition to treatment, screening may also facilitate reduced transmissibility. For example, cases that are identified may be isolated or additional hygienic measures may be taken in areas around such cases. Let denote the institutional transmission rate as a function of the screening rate. The change in disease prevalence at health institution i from individuals becoming infected via transmission is given by. We assume the transmission rate is a non-increasing function of the unit' s screening level, . In addition, transmission may occur in the community as well. Let denote the transmission rate in the community. Both the transmission and treatment functions are developed in further detail and given specific functional forms below. Finally, disease prevalence within each institution as well as the community will be affected by the movement of infected individuals among the various populations. This movement is described by two types of flow parameters: (1) the rate at which individuals move out of a particular subpopulation, and (2) the direction of movement to the various subpopulations. First, let denote the rate of patient turnover at institution i (i. e. is the average patient length of stay in institution i); we assume this rate is constant among the different institutions. Similarly, let denote the turnover rate in the community (i. e. rate of admission to health institutions). Second, let denote the proportion of patients transferred from subpopulation h to subpopulation k, where h and k include both health institutions and the community (the order of subscripts reflects the direction of patient movement). Figure 1A provides a summary of this patient movement in a setting with four different institutions. Since institutions are assumed to be homogenous, the proportion of movement between the various health institutions must be equivalent, i. e. where i and j are different institutions, and. Similarly, the homogeneity assumption implies that the proportion of movement between each institution and the community must be identical, i. e. for all. In equilibrium these flow parameters must capture all patient movement, i. e. where k includes all other institutions and the community. Therefore, these flow rates will be fixed by the number of institutions and the proportion of movement from the community; flow parameters can then be rewritten as. Given the parameters described above along with each institution' s chosen screening level, the dynamics of disease prevalence within each institution, , and the community can be represented by the following system of differential equations: (1) where and represent the derivatives of and z with respect to time. We can solve this system to obtain disease prevalence in a single institution i as a function of i' s screening level, the screening level of all other institutions and time. Thus, disease prevalence in institution i can be expressed by the function, where represents the set of screening policies set by all other institutions. It can be shown that this function is decreasing in i' s screening level and the screening level of all other institutions. The economic decision problem facing each institution is to choose the level of screening that minimizes the net present value of all future costs. Institutions must tradeoff the costs associated with screening versus the cost associated with increased disease prevalence. Screening costs in institution i, at a single point in time, are given by the following: (2) where is the cost of screening per patient screened, is the per patient cost of treatment, and is the screening true positive rate. Although this functional form specifically implies that only cases that correctly test positive receive treatment, the values for the parameters and or can be adjusted to capture the possibility that false positives also receive treatment. Moreover, the value for treatment costs can also be designed to include costs associated with secondary or confirmatory screening (see [13]). Although screening is costly, there are also costs associated with disease prevalence. For example, increased prevalence within an institution may lead to greater transmission, make treating unrelated diseases more costly, require additional hospital personnel and, if conditions become severe enough, diminish the institution' s reputation. Let the costs associated with disease prevalence be given by γ, so the total costs associated with prevalence in institution i are given by. Given these two cost sources, an institution must choose a single level of screening that minimizes its net present value of all future costs. Because disease dynamics and the associated prevalence evolve over time, as described by system (1), total costs are derived by integrating over changes in disease prevalence across time. Thus, the institution' s decision problem can be stated as the following: (3) where ρ is the economic discount rate. We can see from this equation that an institution' s optimal screening level will be dependent on the screening levels set in all other institutions, . This implies that choosing an optimal screening level is a game; an institution' s optimal screening level will depend on the screening strategy taken by its opponents (i. e. the other institutions). We can define an institution' s best response function, given the screening level set by its opponents, by the following: (4) Because all institutions are homogeneous, we are primarily interested in finding solutions that constitute a symmetric pure strategy Nash equilibrium, i. e. a fixed screening level that is mutually optimal when all institutions respond identically. We obtain the symmetric equilibrium by finding a fixed point of this best response function, i. e. a value such that. The purpose of this study is not to determine a specific institution' s optimal screening level given a set of parameters, but rather to describe what happens to the optimal level as institutions are made more or less autonomous in setting their own screening policy. We can simulate the effects of varying the level of autonomy by clustering sets of individual institutions into groups where screening policy in each group is then set by a single decision maker. We will refer to these groups of institutions as Decision Units (DUs) and assume there are a total of DUs covering all institutions. If screening policy were to be set at a national level we would have DU, whereas if screening were set at a state level we would have DUs. Thus, when units are perfectly autonomous. Notice that we can find the socially optimal screening level, the single screening level that minimizes total societal costs across all institutions, by setting. For simplicity, we assume that these DUs are homogenous and each contains the same number of institutions, i. e. the number of institutions in each DU is. We also assume that a single decision maker managing each DU will consider the costs over all the institutions within the DU when choosing a screening level. Thus, the problem facing a decision maker managing a set of institutions is simply to choose a screening level that minimizes (3) summed across all institutions contained in m. Because we are only looking for symmetric pure strategy equilibria, the above problem can be reduced to considering only two DUs, namely a focal DU and a second DU composed of all other DUs. Since all DUs are homogenous, solving the problem from the perspective of the focal unit must then solve the problem for all units. We begin by reducing system (1) to that of a focal unit A, a cluster of all other DUs B and a community z. This system governing disease dynamics then becomes the following: (5) Where and denote transfers made between institutions within the same DU, and denote transfers between institutions in different DUs, and denotes the proportion of transfers to a single institution that are from the community. This system is stated in terms of only two DUs in order to solve for the symmetric equilibrium, but the flow parameters (, , etc.) account for the fact that patients flow between institutions both within the same DU and among DUs. Figure 1B depicts an example of such patient flow in a setting with four institutions and two DUs. Comparing the flow parameters between Figures 1A and 1B demonstrates how the patient flow between individual institutions becomes internalized when patient movement is modeled in terms of a focal DU and all other DUs. In equilibrium these flow parameters between DUs must satisfy the following: (6) Because all institutions and DUs are assumed to be homogenous, the only flow parameter that must be specified in this system is the proportion of admitted patients coming from the community. Given the above system of disease dynamics, we can solve for disease prevalence in the focal DU as a function of its own screening level, the screening level of its opponent, the total number of DUs M and time t, i. e. . This allows us to rewrite the best response function from (4) for the focal DU by the following: (7) Finally, we obtain the equilibrium of interest by finding a fixed point of this best response function, i. e. a value such that where is given by (7). Note: from this point forward we will refer to M as a measure of institutional autonomy, where the level of institutional autonomy can be thought to be increasing with the number of DUs given by M. This SIS compartmental model lacks a closed form solution; thus, we solve the model numerically while exploring a realistic space of parameter values. Table 1 contains a summary of the model parameters along with the range of values that was explored. To solve the model numerically, it is first necessary to specify functional forms for both the transmission and the treatment functions. We use the following form to estimate the transmission function: (8) where, , and are all positive, constant, parameter values that can be calibrated using disease transmissibility data. This functional form has a number of useful properties. First, it exhibits diminishing marginal returns from screening. We assume that increasing the screening level when the initial screening level is very low has a greater impact on transmissibility than when the initial level is very high. For example, initial screening may identify areas of the hospital or specific patient populations to focus on. This also allows us to capture the fact that patients may be screened on their probability of infection. Second, this function can be calibrated to converge to a minimum level of transmissibility with perfect screening, given by the value. We assume that even when all patients are screened and treated, there may still exist a positive probability of transmission to susceptible patients. For example, individuals may contract disease from contact with surfaces in a hospital setting. Finally, the parameters and can be used to calibrate this function to the rate at which transmission declines with increased screening and treatment. Because our objective here was to provide a general understanding of screening decision process applicable to a wide range of diseases, we explored a range of values for, , and and analyzed the effects of varying each of these values individually and simultaneously. This analysis is described in further detail in the sensitivity analysis section below. In addition, we analyzed the impact of using a linear transmission function in place of (8); a description of this analysis can be found in Appendix S1. Finally, we use the following function to model the treatment recovery rate: where τ is the treatment efficacy rate and, as before, is the true positive rate of screening. With this function we assume that anyone who is identified as having the disease gets treated and recovers at a rate proportional to the efficacy of treatment.
To ensure that the results presented above were not tied to a specific set of parameter values we performed a sensitivity analysis by exploring a wide range of potential parameter values. Because we wanted our results to be generalizable to both a variety of possible diseases and a range of screening settings, we chose to explore a wide range of values for each parameter rather than fit the parameters to specific empirical estimates. This parameter space was intended to cover all empirically plausible values a parameter could represent. In Table 1, each of the specific parameters in our model is described along with the range of values that were explored. For each parameter, a univariate analysis was performed by varying the parameter over this space. Additionally, a number of multivariate sensitivity analyses were performed on sets of related variables by simultaneously varying multiple parameters over this space. A detailed description of this sensitivity analysis can be found in Appendix S1. Throughout this sensitivity analysis we tested the strength of our major findings by comparing the changing shape of the best response curve as we varied the level of treatment efficacy τ and number of DUs M. The results of our sensitivity analysis confirmed that our two major findings hold across the entire parameter space explored. First, Figures S1, S2, S3, S4, S5, S6, S7, S8, S9, S10 show that along the entire parameter space, a decrease in treatment efficacy was associated with a transformation of the best response curve in such a way that its downward trend diminished. For most of the parameter values explored, this decrease in treatment efficacy resulted in a transition of the best response curve from downward sloping to upward sloping. However, while the downward trend of the best response curve did diminish in all cases we explored, for some parameter values the decline in treatment efficacy is not enough to cause the curve to become upward sloping (see Figures S5 or S10). Second, Figures S11, S12, S13, S14, S15, S16, S17, S18, S19, S20 depict the effect of changing the number of DUs across the parameter space and for different levels of treatment efficacy. These figures show that across the entire range of parameters analyzed, an increase in the number of decision makers was associated with a decline in the equilibrium screening level. This sensitivity analysis demonstrates that while the disease dynamics of the underlying compartmental model may be affected by the specific parameter values, the basic outcomes of the screening game appear to be stable across a wide range of parameter values. This finding suggests that our results are applicable to a wide range of diseases and settings.
This study provides insights into the optimal level of disease screening in a strategic multi-institutional setting where potentially infected patients move back and forth between different screening institutions. This model allows us to analyze how the optimal level of screening is influenced by the number of decision makers, i. e. the level of institutional autonomy in choosing a screening level, as well as various disease parameters. The overarching message of this study is that having more decision makers in the screening process leads to less screening and increased disease prevalence. This means that mandated screening levels, either at a state or national level, may be more effective at controlling the spread of disease than simply allowing individual institutions to set their own screening level. Having a greater number of decision makers increases the opportunity to free-ride, and, in a setting where infected patients spread disease while moving between institutions, the actions of a single decision maker will be less effective than those of a coordinated group. In such multi-institutional settings, coordination of actions, which can only occur under a limited number of decision makers, essentially makes screening cheaper and more effective. This basic result, that more decision makers may lead to worse outcomes, is fairly intuitive and is consistent with previous research [8]. In fact, an understanding of this result has likely been the motivation for efforts to implement screening mandates for certain diseases. However, this study adds two significant refinements to the current literature, which have additional policy implications. First, the marginal benefit that is provided by decreasing the number of decision makers declines as the number of decision makers increases. Moving from 50 to 25 decision makers was shown to have a much smaller relative benefit than moving from five decision makers to two. Consequently, efforts that only slightly reduce autonomy (e. g. setting policy across affiliated institutions) may have little discernible impact on screening and disease prevalence. This fact might encourage policy makers to adopt a “go big or go home” type of strategy, by only choosing to implement screening mandates that move to a single or very limited number of decision makers. The second major refinement that this study provides is to highlight the fact that the added benefit of moving to a smaller number of decision makers is largely dependent on the availability and efficacy of treatment. The marginal benefit from reducing the number of decision makers is much greater for diseases with no, or less effective, treatment. In some cases, if a disease is less treatable and institutions are given too much autonomy, they may end up choosing not to screen at all. This occurs because with diseases that are less treatable, strategies aimed at reducing transmission become the primary method for reducing prevalence, and the effectiveness of such strategies is largely dependent on the level of coordination between institutions. If treatment is not an option, a single institution will have a much harder time trying to prevent transmission when neighboring institutions are not doing the same and will, consequently, give up. Therefore, mandated screening policy becomes much more important for diseases that are less treatable. For diseases such as Klebsiella pneumonia with New Delhi metal-beta-lactamase-1 (NDM-1) [14] or Acinetobacter baumannii [15], it may be much more important for screening policy to be mandated by a central planner. Thus, the emergence of multi-drug resistant organisms where there are no, or very few, antibiotics available for treatment leads to a situation where a central screening policy is more effective. One limitation of this study that deserves mention is the assumption of a homogeneous setting. In our model, institutions and DUs were assumed to be identical, and patients were assumed to be homogeneously mixed in order to guarantee the existence of a pure strategy equilibrium. In reality, institutions are not perfectly homogeneous, and patient transfers often vary greatly between institutions. This is especially true when comparing different types of wards within a hospital or between different types of institutions. In such a heterogeneous setting, it is likely that different institutions will have different screening incentives and, consequently, choose different optimal screening levels. For example, a large hospital with a high patient flow may find it preferable to screen at a greater rate than a smaller hospital. However, amending the model to include a heterogeneous institutional population would very likely eliminate the existence of pure strategy equilibria for many parameter values. A heterogeneous screening model is also likely to lead to the existence of multiple mixed strategy equilibria. Because our goal in this paper was to provide a general understanding of how the level of institutional autonomy can influence the screening decision process and because such mixed strategy equilibria are often difficult to interpret intuitively, we chose to model institutions homogeneously. Therefore, the results presented here should be thought of as providing an intuition for how the number of decision makers and the level of treatment efficacy impact screening incentives in general, rather than providing an empirical estimate of how specific institutions will behave. Although screening results have been presented in terms of economic optimality, it is worth noting that such optimality specifically refers to the theoretical optimization problem present in our model; in reality, the empirically optimal screening level may be dependent on factors that our model did not account for. One example where this distinction may be important is when interpreting the result that an increased number of decision makers may lead to less socially optimal screening outcomes. Our model did not take into account the costs associated with establishing and enforcing mandated screening levels. Mandating screening across institutions at a national or even state level is likely to entail a number of political, administrative and enforcement costs that were not considered in this model. Policy makers will need to consider such costs along with the implications presented here when choosing the best policy. However, one result from this study that may shed light on this policy making process is the fact that the marginal benefit of implementing a mandatory screening policy will be greater when it involves a smaller number of individual decision makers. This result directly implies that if the costs of imposing a screening mandate are roughly the same for establishing either a local or federal mandate, it will be a much more cost effective decision to establish a national rather than local mandate. Moreover, the costs of establishing a local mandate will need to be significantly lower than those associated with a national mandate in order for the local mandate to be a preferable option. | Healthcare associated infections are a major cause of morbidity and mortality. Screening patients on admission to the hospital may reduce prevalence by identifying infected individuals; infected individuals can then be treated or isolated to prevent further spread. Because screening is costly, institutions must weigh the benefits of reduced prevalence against the costs of screening. However, patients move between institutions carrying disease with them; consequently, when choosing who to screen, institutions must also consider the rates at which neighboring institutions screen patients as well. We develop a theoretical model that describes this strategic decision process. Using this model we are able to analyze the screening decision problem along three dimensions: (1) how disease specific parameters, such as the effectiveness of treatment, influence the optimal screening level, (2) how the degree of centralization in screening policy (e. g. local, state or federal) influences the optimal screening level, and (3) how these two sets of factors combine to influence the optimal screening level. Our model highlights factors to consider when choosing to implement screening policy, and results are of use to policy makers wishing to reduce the prevalence of infectious disease. | Abstract
Introduction
Methods
Results
Discussion | medicine
infectious diseases
social and behavioral sciences
health economics
infectious disease epidemiology
epidemiology
epidemiological methods
economics | 2014 | Optimal Screening Strategies for Healthcare Associated Infections in a Multi-Institutional Setting | 5,892 | 230 |
In response to DNA damage during S phase, cells slow DNA replication. This slowing is orchestrated by the intra-S checkpoint and involves inhibition of origin firing and reduction of replication fork speed. Slowing of replication allows for tolerance of DNA damage and suppresses genomic instability. Although the mechanisms of origin inhibition by the intra-S checkpoint are understood, major questions remain about how the checkpoint regulates replication forks: Does the checkpoint regulate the rate of fork progression? Does the checkpoint affect all forks, or only those encountering damage? Does the checkpoint facilitate the replication of polymerase-blocking lesions? To address these questions, we have analyzed the checkpoint in the fission yeast Schizosaccharomyces pombe using a single-molecule DNA combing assay, which allows us to unambiguously separate the contribution of origin and fork regulation towards replication slowing, and allows us to investigate the behavior of individual forks. Moreover, we have interrogated the role of forks interacting with individual sites of damage by using three damaging agents—MMS, 4NQO and bleomycin—that cause similar levels of replication slowing with very different frequency of DNA lesions. We find that the checkpoint slows replication by inhibiting origin firing, but not by decreasing fork rates. However, the checkpoint appears to facilitate replication of damaged templates, allowing forks to more quickly pass lesions. Finally, using a novel analytic approach, we rigorously identify fork stalling events in our combing data and show that they play a previously unappreciated role in shaping replication kinetics in response to DNA damage.
In response to DNA damage during the G1 or G2 phase of the cell cycle, DNA damage checkpoints block cell cycle progression, giving cells time to repair damage before proceeding to the next phase of the cell cycle [1,2]. The response to DNA damage during S phase is more complicated, because repair has to be coordinated with ongoing DNA replication [3]. DNA damage during S phase activates the intra-S DNA damage checkpoint, which does not completely block S-phase progression, but rather slows DNA replication, presumably allowing for replication-coupled repair [4]. Lack of the intra-S DNA damage checkpoint predisposes cells to genomic instability [5]. Nonetheless, the mechanisms by which replication is slowed, and the roles of checkpoint-dependent and -independent regulation in the S-phase DNA damage response, are not well understood. The slowing of S phase in response to damage involves both inhibition of origin firing and reduction in fork rate [6–12]. The effect of the checkpoint on origin firing has been characterized in budding yeast and mammalian cells. The checkpoint prevents activation of late origins by targeting initiation factors required for origin firing. In mammals, checkpoint kinase 1 (Chk1) inhibits origin firing by targeting the replication kinases, cyclin-dependent kinase (CDK) and Dbf4-dependent kinase (DDK) [13–15]. In budding yeast, Rad53 targets Sld3, an origin initiation factor, and Dbf4, the regulatory subunit of DDK [16,17]. In addition to Sld3 and Dbf4, Mcm4 plays a critical role in suppressing late origin firing in response to replication stress [18]. Although checkpoint inhibition of origin firing is conserved from yeast to mammals, the contribution of origin regulation to damage tolerance is not clear. For instance, budding yeast mutants such as mec1-100, SLD3-m25 and dbf4-m25, which cannot block origin firing in response to damage, are not sensitive to damaging agents such as methyl methanesulfonate (MMS) [16,17,19,20] suggesting that checkpoint regulation of origin firing is not as critical as the checkpoint’s contribution to damage tolerance via fork regulation. The effect of checkpoint activation on replication forks is less well understood. Because many DNA damage lesions block the replicative polymerases, for forks to pass leading-strand lesions, they must have some way to reestablish leading-strand synthesis downstream of the lesion. Recruitment of trans-lesion polymerases, template switching and leading-strand repriming have all been proposed to be involved in the fork by-pass of lesions, but which is actually used in vivo and how the checkpoint may affect that choice, is unclear [21–26]. A consistent observation is that forks slow in response to DNA damage. Whether slowing of forks in the presence of damage is checkpoint-dependent or simply due to the physical presence of lesions is not clear [27–29]. Initial work in budding yeast showed that replication forks in checkpoint mutant and wild-type cells were slowed to the same extent in the presence of damage, suggesting that slowing is checkpoint-independent [27]. However, subsequent work showed that checkpoint activation inhibited replication of damaged DNA, suggesting an active role in the slowing of replication forks [28]. Furthermore, work in mammalian cells showed a role for checkpoint signaling in DNA-damage-dependent slowing of replication forks [29]. Checkpoint regulation clearly affects fork stability [30–33], but that conclusion is largely based on the response to hydroxyurea (HU) -induced replication stress, which blocks fork progression due to deoxynucleoside triphosphate (dNTP) depletion, and thus does not directly address the role of checkpoint regulation in the replication of damaged templates. The question of how replication forks are regulated in response to DNA damage is complicated by the difficulty of measuring fork progression rates. To directly address this question requires single-molecule resolution. Bulk assays of replication kinetics, such as the quantitation of radioactive thymidine incorporation or flow cytometry, provide only an average profile of replication kinetics, convolving the effects of origin firing and fork progression and obscuring any heterogeneity in fork slowing. The effects of DNA damage on specific origins and on forks replicating specific loci can be analyzed by gel- or sequence-based methods [9,12,27,28,34], but these techniques still only reveal the average response to DNA damage. Such approaches lack the single-molecule resolution necessary to identify heterogeneity in response to damage and to distinguish, for instance, if all forks pause briefly at all lesions or if only a fraction of forks stall, but for a longer time. Therefore, to investigate the effect of polymerase blocking lesions on individual origins and forks on a global scale, we have used a single-molecule DNA combing assay. DNA combing is a single-fiber visualization technique that allows mapping of thymidine-analog incorporation patches on uniformly-stretched, megabase-length DNA fibers [35–41]. The analysis of replication kinetics by DNA combing involves isolating DNA from cells pulse-labeled in vivo with thymidine analogs. Labeled DNA is stretched on a coverslip and DNA replicated during the pulse is identified by immunofluorescence. Sequential labeling with two different analogs allows us to determine the direction and speed of replication of labeled tracks on a fiber [37,38] (Fig 1). Because combing reveals the behavior of individual replication forks, it allows us to unambiguously study the effect of checkpoint on origin firing and fork rates. In particular, it allows us to measure the heterogeneity of fork rates and determine if all forks respond the same way to lesions. The importance of measuring the heterogeneity of fork rates can be illustrated in a case where some forks pass lesions without slowing, but others stop at the lesion and do not resume replication for the duration of the experiment, a condition we refer to as fork stalling. The fork stalling is particularly difficult to analyze because stalled forks create ambiguous patterns of nucleotide incorporation which cannot be definitively interpreted in isolation [42]. However, such stalling events can, in principle, be unambiguously identified on fibers that are long enough to contain multiple replicons, allowing potentially ambiguous incorporation patterns to be definitively interpreted from the context of surrounding forks (Fig 2). We have, for the first time, developed an analysis strategy that rigorously incorporates fork stalling and used it to unambiguously quantitate fork stalling rates and the effect of checkpoint activation on fork stalling. A fundamental question about the regulation of fork progression in response to DNA damage is whether it is a global or local effect [43]. The effect of checkpoint on origins is inherently a global response because origins distant to sites of damage are blocked from firing. However slowing of forks could be a local or a global effect. If all forks are slowed by checkpoint activation, irrespective of whether they encounter damage or not, then slowing is a global effect. On the other hand, if forks are slowed only when they encounter a lesion, then it is a local effect [43]. It should be possible to distinguish between global and local regulation of fork progression by examining the effects of different lesion densities. Previously, we showed that the extent of replication slowing correlated with the density of MMS lesion, suggesting a local effect on fork progression [44]. However, we used a bulk replication assay, which did not allow us to directly observe replication fork rates. Here, we have assayed checkpoint-dependent slowing of S phase in fission yeast in response to three DNA damaging drugs that activate the checkpoint at significantly different densities of lesions: methyl methanesulfonate (MMS), which mainly methylates purines and creates a relatively small adduct, 4-nitroquinoline 1-oxide (4NQO), which adds a quinoline group to purines resulting in a bulkier adduct [45–47] and bleomycin, which mainly creates single strand and double strand DNA breaks [48]. MMS and 4NQO create polymerase-stalling lesions and have been shown to activate the intra-S checkpoint [17,44,49–52]. The standard dose of 3. 5 mM (0. 03%) MMS causes about one lesion every 1 kb, whereas a physiologically similar dose of 1 μM 4NQO in CHO cells causes one lesion about every 25 kb and we estimate 16. 5 μM of bleomycin causes about one double-strand break per 50 kb (S1 Fig) [53–56]. Although these are only rough approximations of lesion density, they show that forks will encounter many more MMS lesions than 4NQO lesions or bleomycin-induced double-strand breaks. This wide disparity in lesion density allows us to address differences in global and local effects of the checkpoint. In case of 4NQO, since the lesions are rare, we expect few forks to encounter lesions. Therefore, if all forks slow in response to 4NQO then we can conclude that fork regulation by checkpoint is global in nature. On the contrary, if fork regulation is a local effect then very few forks will actually encounter the lesion and slow. Similarly since the double-strand breaks caused by bleomycin are infrequent we expect very few forks to be affected by the breaks unless the checkpoint actively regulates all forks. In case of MMS, since the lesions are frequent we expect all forks to encounter lesions, and thus to slow regardless of whether slowing is a local or global effect. By comparing the effects of these three drugs, we can differentiate between global and local effects on fork regulation by the checkpoint. We find that fork slowing is a local, checkpoint-independent effect, but that persistent fork stalling plays a more significant role in replication kinetics than previously appreciated.
To investigate the mechanism by which the checkpoint slows replication in response to MMS, 4NQO and bleomycin we used DNA combing. We pulsed cells in S phase with 5-chloro-2’-deoxyuridine (CldU) for 5 minutes and chased it with 5-iodo-2’-deoxyuridine (IdU) for 10 minutes, isolated and combed DNA, and visualized the CldU and IdU analogs with red and green antibodies, respectively. Fig 1 shows a sample fiber from the dataset. The fiber contains a rightward moving fork (red-green [RG]), an origin that fired during IdU labeling (green [G]), and three origins that fired during CldU labeling (green-red-green [GRG]). Further replication patterns observed in our combing dataset are shown with an interpretation of each pattern. The patterns observed were most simply interpreted as a leftward fork (green-red [GR]), a rightward fork (RG), origins that fired during CldU (GRG) or IdU (G), and terminations during IdU (red-green-red [RGR]) or CldU (red [R]). However, a more sophisticated analysis allowing the possibility of fork stalling reveals that many of these patterns—in particular termination during CldU (R) —are ambiguous, as described in the Methods section and Fig 2 and S2 Fig. For each experiment we collected about 25 Mb of DNA, which is about twice the size of the fission yeast genome, ensuring representation of most genomic loci in our analysis. From the combing data we estimated four parameters: origin firing rate, fork density, fork rate and fork stalling frequency (see Methods for details). Difficulty in identifying fork stalling events in combing data arises form the fact that stalls during the first pulse can result in ambiguous analog incorporation patterns [42]. Specifically, an isolated red patch can be interpreted as an elongating fork that stalled during the first pulse or an origin with both its forks stalled or as a termination event during the first pulse (Fig 2A). Therefore, we cannot use first-pulse events alone as rigorous evidence for stalled forks. However, using double-labeled combing data we can unambiguously identify fork stalling. In particular, two neighboring forks moving in the same direction show that the fork in between them moving in the opposite direction must have stalled (Fig 2B). Thus, a red-unlabeled-green (RUG) or green-unlabeled-red (GUR) pattern is diagnostic of a fork stall. Since the fibers in our datasets average over 400 kb, we observed many neighboring forks, allowing us to robustly measure fork stalling. Although stalled forks can be identified using the RUG and GUR patterns, some stalled forks still produce ambiguous patterns. For instance, an RUR pattern can be produced both by forks moving away from an origin and two converging forks that have stalled (Fig 2C). Thus, in order to approximate the fork stalling rate in our entire dataset, we used a probabilistic approach to estimate how many of the ambiguous patterns arise from fork stalls. We enumerated all possible incorporation patterns and classified them by the possibility that they arose from a fork stall (see Methods for details, S2 Fig). We then used the stall rate from the unambiguous signals (GUR or RUG) to calculate the likelihood of ambiguous events being due to stalled forks and used that frequency for our final estimation of stall rate for each fiber. To determine the effect of MMS, bleomycin, and 4NQO—three compounds that activate the checkpoint at very different densities of lesions—on the replication rate at a population level, we analyzed cells' response to them by flow cytometry. G1 synchronized cells were released into S phase with or without DNA damage, using the commonly used dose of 3. 5 mM (0. 03%) MMS or a dose of 1 μM 4NQO or 16. 5 μM bleomycin chosen to produce a similar slowing of bulk replication (S3 and S4 Figs). These doses of MMS, 4NQO and bleomycin cause lesions about once every 1 kb, 25 kb, and 50 kb respectively (S1 Fig) [53–56]. By flow cytometry, control cells completed replication by 80 minutes, while in the presence of either drug cells slowed replication, reaching only about 60% replicated by the end of the time course (Fig 3A and S5A Fig). Thus, despite the disparity in the number of lesions created by MMS, 4NQO and bleomycin, all three drugs led to a similar extent of replication slowing at the doses used. In all cases, the slowing in response to DNA damage was largely checkpoint-dependent. In the absence of the Cds1 checkpoint kinase, cells completed replication by 100 minutes even in the presence of damage, as reported previously (Fig 3B and S5B Fig) [52,57]. Although the three drugs appeared to have similar extent of slowing in wild-type, their effects in cds1Δ cells differed, with cds1Δ cells showing more checkpoint-independent slowing in MMS than in 4NQO or bleomycin (Fig 3B). We therefore examined if there is a difference in the mechanism by which DNA replication is slowed in response to MMS, 4NQO and bleomycin. As shown in Fig 3 we see similar levels of bulk slowing for all three damage treatments at 60 minutes in S phase (Figs 3 and 4A). We first analyzed the effect of DNA damage on origin firing. To measure the rate of origin firing, we directly measured the number of new origins fired during the CldU pulse (GRG patches) or IdU pulse (isolated green patches). The overall origin firing rate was calculated as the total number of origin firing events in each fiber normalized to total length of un-replicated DNA of that fiber and to the length of the analog pulse (Fig 4B). Additionally, the origin firing rate during first analog and second analog was determined separately (Fig 4D, 4E and 4F). In untreated controls, the rate of origin firing in the first analog was 2. 3±0. 7 origins per Mb per minute, similar to previous estimates of origin firing rates (S1 Table) [58]. In the presence of 4NQO, the origin firing rate in wild-type cells decreased to 47% of the untreated control (p = 5. 17x10-12, t tests were used for all statistical tests) (Fig 4B and S1 Table). Since, the origin firing rate only measures the origins that fired during the analog pulses, we also measured the density of active forks in the datasets in order to estimate the effect on origin firing prior to the analog pulses (see Methods for details). Fork density in 4NQO-treated cells was 64% of the untreated control (p = 1. 28x10-6) during the first analog pulse and 44% (p = 3. 08x10-13) in the second, consistent with the trend seen in origin firing rates (Fig 4D). Thus, the response to 4NQO included an immediate reduction in origin firing rate. We see a similar trend for bleomycin treated sample. The origin firing rate decreases to 58% (p = 7. 45x10-5) and the fork density decreases to 61% (p = 1. 33x10-5) (Fig 4B and 4C). Analog specific estimation shows that bleomycin treatment leads to reduction in the origin firing rate in the first analog as well as second analog to 69% (3. 91x10-4) and shows a corresponding decrease in fork density during the first (77%, p = 7. 65x10-4) and the second analog (55%, p = 8. 54x10-7) (Fig 4E). In case of MMS, the overall origin firing rate was reduced to 72% (p = 1. 18x10-6) (Fig 4B). Analyzing origin firing during the first and second pulse separately, we saw no statistically significant reduction in the first analog (90%, p = 0. 0748) followed by a stronger reduction to 56% (p = 7. 15x10-11) in the second analog (Fig 4F). Therefore, the effect of MMS on origin firing rate is delayed. This conclusion was supported by two observations. Firstly, the average fork density across both analogs in MMS showed a modest reduction to 78% (p = 2. 59x10-5) as compared to 53% (p = 5. 7x10-12) in 4NQO (Fig 4C). Second, the analog-specific fork density estimations for MMS followed a similar trend as the origin firing rate showing a greater reduction during the second analog (0. 93 v. 0. 62, Fig 4F). Hence, the effect of MMS-induced damage on origin firing is only manifest late in S-phase, after significant bulk slowing has already occurred, whereas 4NQO- and bleomycin-induced damage inhibit origin firing immediately in early S phase. Since 4NQO and MMS both create polymerase-blocking lesions [17,49–51], we next studied how these drugs affect fork speed. We measured fork rate in the combing data as the length of the green track (second analog) continuing from a red track (first analog), divided by the length of the chase time (10 minutes) (Fig 1). The average fork rate in our untreated samples was 0. 91±0. 02 kb/minute, which is within the range of previous estimates [42,59–62]. In MMS, the fork rate decreases to 76% of the untreated control (p = 6. 13x10-38) (Fig 4G and S9 Fig). The reduction in fork rate was expected since the lesions are so frequent that every fork encounters about ten of lesions during the 10 minutes pulse. In case of 4NQO, the lesions are significantly less common, so only about half of the forks should encounter a lesion during the second pulse. Consistent with the low density of lesions, fork rates in 4NQO treated cells were similar to untreated cells (Fig 4G). In fact, the fork rate showed a 27% increase in the presence of 4NQO as compared to untreated cells (p = 5. 75x10-4). The dose of bleomycin used appears to create about one break per 50 kb (S1B Fig). However, this estimate of lesion density may be an overestimation, because we have analyzed fibers on average 450 kb in length in bleomycin-treated wild-type samples by combing. Regardless, we do not see any statistically significant reduction in fork rate for bleomycin treated sample as compared to untreated sample (p = 0. 3138) (Fig 4G). In case of MMS treated sample, we only see a modest reduction in fork density as compared to 4NQO and bleomycin (Fig 4C). In particular, there is no statistically significant reduction in fork density in the MMS treated sample during the first analog (93%, p = 0. 1951) (Fig 4F). However by bulk assay we see the same kinetics of slowing for all three damage treatments (Figs 3A and 4A). A possible explanation for this discrepancy is that forks stall during the first pulse in response to damage and thus are not observed during the second pulse (Fig 2). We therefore interrogated our combing data for evidence of fork stalling. We first determined the absolute number of stalls per kb for each dataset (see Methods for details). The absolute number of stalls per kb determines the contribution of stalled forks towards total replication slowing. In MMS treated samples, we see a 3. 47-fold increase in the number of stalls as compared to untreated (p = 6. 23x10-27), while we see no significant increase in stalling in response to 4NQO (1. 12, p = 0. 604) and bleomycin (0. 99, p = 0. 346) (Fig 4H). Thus an increase in the total number of stalls per kb in response to MMS helps explain the slowing of total replication to similar levels as compared to 4NQO and bleomycin despite having a delayed effect on origin firing rate. Next we estimated the fork stall rate. We define the fork stall rate as the total number of stall events per fiber before and during the first analog pulse divided by the total number of ongoing forks in that fiber during the first analog. Although the absolute number of stalls per kb in response to 4NQO and bleomycin treatment is similar to the wild-type untreated sample, the treated samples have far fewer origins firing as compared to untreated sample (Fig 4B and 4H). Therefore the rate of stalling normalized to the origin firing rate is higher in the treated sample (Fig 4I). The average stall rate per fiber in the untreated sample was 14%. The fork stall rate showed a 1. 8-fold increase in response to 4NQO (p = 1. 2x10-4) and a 1. 9-fold increase in response to bleomycin (p = 2. 26x10-7), whereas MMS caused a 3. 8-fold increase relative to untreated cells (p = 1. 55x10-26) (Fig 4I). Combining our stall-rate data with estimates of lesion density, we estimate that forks stall at 5% of 4NQO lesions and at 0. 5% of MMS lesions, consistent with the fact that 4NQO creates a bulkier lesion than MMS. The stall rate per lesion is more complicated to interpret for bleomycin. Since the stalls we detect are not at the ends of our fibers, they cannot be at the sites of bleomycin-induced double-strand breaks. However, bleomycin is reported to create a 6 to 20-fold excess of single-strand nicks to double strand breaks, which, at the dose we used should produce one nick about every 2. 5–9 kb (S1B Fig) [48]. Since we expect that every leading-strand nick will lead to fork collapse, resulting in a single-ended DSB on one strand and a terminated nascent strand, which we would score as a fork stall, on the other, we presume that most nicks are rapidly repaired and estimate that forks encounter bleomycin-induced leading-strand nicks about every 300 kb (S1 Table). Although the stall events seem to be infrequent relative to lesion density, approximately 53% of forks in MMS-treated cells and 25% of forks in 4NQO- and bleomycin-treated cells stalled, contributing significantly to slowing, and ensuring that all treated cells had many stalled forks. To determine the role of the intra-S checkpoint in the observed DNA-damage dependent changes in replication kinetics, we repeated our combing experiments in checkpoint-deficient cds1Δ cells. In the presence of 4NQO, wild-type cells replicated 42% (p = 5. 55x10-16) as much as untreated cells, as assayed by flow cytometry at 60 minutes after release, whereas, in the absence of checkpoint, cds1Δ 4NQO treated cells replicated 77% as compared to untreated cells (p = 5. 82x10-8) (Fig 5A). 4NQO did not significantly reduce origin firing in cds1Δ cells (89%, p = 0. 206), as opposed to 47% in wild-type cells (p = 5. 17x10-12) (Fig 5B). Thus, the inhibition of origin firing in response to 4NQO is checkpoint-dependent (Figs 4B and 5B). Analog specific origin density and fork density followed similar trends as the total origin firing rate data (Fig 5D). During both the analog pulses, cds1Δ cells had a higher origin firing rate and fork density than wild-type cells (Fig 5D). Likewise, inhibition of origin firing in response to bleomycin and MMS is checkpoint-dependent (Fig 5E and 5F). In wild-type cells, the overall origin firing rate was reduced to 58% by bleomycin (p = 7. 45x10-5) and 72% by MMS (p = 1. 18x10-6) treatment (Figs 4B and 5B). In contrast, cds1Δ cells showed no significant decrease in origin firing relative to untreated cells when treated with bleomycin (96%, p = 0. 389) or MMS (114%, p = 0. 011) (Fig 5B). Analog specific estimation of fork density and origin firing rate in cds1Δ treated with bleomycin or MMS showed a similar trend (Fig 5E and 5F). cds1Δ cells had higher fork density and origin firing rate in response to bleomycin and MMS than wild-type cells for both analog pulses (Fig 5E and 5F). In the absence of DNA damage, the lack of Cds1 caused a 51% increase in the rate of origin firing (3. 5±0. 6 v. 2. 3±0. 6 origins firing/Mb/minute, S1 Table), consistent with previous reports in other systems of checkpoint inhibition of origin firing in unperturbed S phase [62,63]. Nonetheless, loss of Cds1 increases origin firing in damaged cells to the same level as in undamaged cells, showing that there is no checkpoint independent inhibition of origin firing (S6 Fig). In MMS-treated wild-type cells, fork rate was reduced to 76% of the untreated cells (p = 6. 14x10-38) (Figs 4G and 6A). This effect was not checkpoint-dependent. In fact, at 61% (p = 6. 13x10-66), cds1Δ cells showed a greater reduction in fork rate than seen in wild-type cells in response to MMS (Fig 6A). Thus, the observed reduction in fork rate in response to MMS seems to be due to the physical presence of the lesions and the checkpoint activation may facilitate efficient by-pass of the lesions. Fork rates showed similar increase in cds1Δ cells treated with 4NQO (126%, p = 1. 17x10-5) as they did in wild-type cells (127%, p = 5. 75x10-4) (Figs 4G and 6A). Although the lack of Cds1 does not seem to have an effect on the relative fork rate between untreated and 4NQO-treated cells, it does have a significant effect on the absolute fork rate in untreated cells. Forks moved significantly slower in untreated cds1Δ cells than in wild-type cells (0. 72 v 0. 91 kb/min, p < 10−9, S1 Table). This difference may be an indirect effect of the higher origin firing rate and fork density in cds1Δ cells (S1 Table). Several groups working in different systems have made the similar observation that fork rate is inversely correlated to the number of active forks, perhaps due to constrains on a limiting factor required for replication, such as the dNTP pool [64–66]. Similar to wild-type cells, we see an increase in fork stalls per kb in response to MMS treatment in cds1Δ but not in case of 4NQO and bleomycin (Figs 4H and 6B). The fork stalls per kb shows a 1. 9-fold increase in response to MMS as compared to untreated (p = 4. 42x10-10) cds1Δ cells. Fork stalls per kb does not change significantly between 4NQO (1. 39, p = 0. 987) and bleomycin (1. 05, p = 0. 839) treated cds1Δ cells as compared to untreated cds1Δ (Fig 6B). In wild-type cells, the fork stalling rate was increased 1. 8-fold by 4NQO (p = 1. 2x10-4), 1. 9-fold by bleomycin (p = 2. 26x10-7) and 3. 8-fold (p = 1. 55x10-26) by MMS treatment, relative to untreated cells (Figs 4H and 6C). In cds1Δ cells, we saw a 1. 7-, 2. 3- and 2. 8-fold increase of the stall rate in 4NQO (p = 2. 59x10-4), bleomycin (p = 7. 96x10-13) and MMS (p = 2. 93x10-15) treated cells respectively as compared to untreated cells (Fig 6C). This increase in stall rate is slightly lower in MMS-treated cds1Δ cells than in wild-type cells (2. 8 fold v 3. 8 fold, p = 9. 86x10-5), showing that there are checkpoint-dependent and -independent contributions to stalling in response to MMS, however the differences in stall rates for cds1Δ cells treated with 4NQO (1. 7 fold v 1. 8 fold, p = 0. 746) and bleomycin (2. 3 fold v 1. 89 fold, p = 0. 01) as compared to wild-type cells are less pronounced. Therefore the bulk of fork stalling events appears to be checkpoint-independent (Fig 6C). To test the possibility that the later inhibition of origin firing seen in response to MMS (Fig 4F) is due to delayed checkpoint activation, and to confirm that the doses of the various damaging agents we use cause comparable checkpoint activation, we assayed activation of the Cds1 S-phase checkpoint kinase in response to MMS, 4NQO and bleomycin. Cells with an HA-tagged Cds1 were synchronized in G1 and released into DNA damage (S7 Fig). Cells were harvested throughout S phase and Cds1 was immunopurified for in vitro kinase assays (Fig 7). We observe a significant and reproducible delay of Cds1 activation in response to MMS, relative to both 4NQO and bleomycin. These results are consistent with MMS taking longer to disrupt a sufficient number of replication forks to trigger a robust checkpoint response. To investigate the relationship between the fork stalling that we measured by DNA combing and the fate of forks in living cells, we visualized replication protein A (RPA) -GFP foci, which mark the accumulation of single-stranded DNA (ssDNA), in cells treated with MMS or 4NQO. Because RPA-coated single-stranded DNA is a trigger for the activation of the intra-S checkpoint [67–71], we hypothesized that stalls that accumulate substantial amounts of RPA are likely to correlate with checkpoint activation, but that stalls that lead to less RPA accumulation may be checkpoint silent. We analyzed 100 cells for each sample and sorted them into three categories based on the intensity and number of foci: strong foci, weak foci or no foci (see Methods for details). 21% of untreated wild-type cells have some foci, consistent with previous reports [72], but the majority of these are weak foci, consistent with the hypothesis that weak foci do not activate the checkpoint (Fig 8A). In the presence of damage, we saw distinctly different responses to MMS and 4NQO. In the presence of MMS, 36% of cells had strong foci (Fig 8A). Since all MMS-treated cells have stalled forks (S1 Table), we conclude that only a minority of MMS-induced stalls accumulate substantial ssDNA. On the contrary, essentially all 4NQO-treated cells had strong foci, suggesting a qualitatively different nature of 4NQO-induced fork stalls, in which the majority of 4NQO-induced lesions accumulated substantial ssDNA (Fig 8A). As previously reported [73], RPA did not accumulate at HU-stalled forks in checkpoint-proficient cells. In the absence of checkpoint all treated samples displayed strong foci in about 90% of cells. In particular, in MMS and HU treated cells we saw a large increase in the number of cells with strong foci in cds1Δ as compared to wild-type, suggesting that in response to MMS treatment the checkpoint plays a role in preventing accumulation of excess ssDNA at stalled forks, as it does in HU [73,74] (Fig 8A).
Consistent with previous studies from S. cerevisiae [9] and human cells [7,8, 15] reduction in origin firing in response to 4NQO, bleomycin and MMS is checkpoint-dependent. By flow cytometry, 4NQO, bleomycin and MMS lead to similar extent of slowing, however the combing data reveals important differences in the regulation of origin firing in response to the two drugs. In case of 4NQO and bleomycin reduction in origin firing is robust and immediate. The fork density during the first analog, which is a proxy for origin firing occurring prior to analog labeling, decreases to 64% (p = 1. 28x10-6) in case of 4NQO and 77% (p = 7. 65x10-4) in case of bleomycin, as compared to untreated cells (Fig 4D and 4E). Therefore, origin firing inhibition starts early in S phase in response to 4NQO and bleomycin. On the contrary, in the case of MMS, fork density in the first analog is barely reduced (93%, p = 0. 195) suggesting no significant reduction in origin firing during early S phase (Fig 4F). Furthermore, even during the second analog, the fork density is 40% higher in MMS than in 4NQO (62% v. 44%, Fig 4D and 4F). Therefore reduction of origin firing in MMS is delayed and modest as compared to 4NQO. The disparity in origin firing inhibition response can be explained by the observation that a threshold of damage has to be met for the activation of the intra-S checkpoint. A certain number of arrested forks are necessary for checkpoint activation during S phase [76]. Therefore robust reduction in origin firing in response to 4NQO in early S phase suggests that 4NQO lesions, although less frequent than MMS lesion at the concentrations used in our studies, have more severe effects on forks and hence are more efficient at activating the checkpoint than MMS. Consistent with this interpretation, 4NQO lesions lead to a 10- fold higher rate of fork stalls per lesion than MMS (5% v. 0. 5%) as detected by combing. The low density of stalls per lesion in response to MMS suggests that MMS takes longer to induce sufficient fork stalls to activate a checkpoint signal, and explains why the checkpoint mainly inhibits origin firing later in S phase. Our analysis of RPA accumulation is also consistent with this interpretation, showing that 4NQO-induced lesions accumulate RPA to a much greater extent than MMS-induced lesions (Fig 8). The fork rate in wild-type cells is reduced in response to MMS to 76% (p = 6. 14x10-38) (Fig 4G). The fork rate is also reduced to 61% in cds1Δ cells treated with MMS (p = 6. 13x10-66) (Fig 6A) consistent with previous reports [27]. Thus, reduction in fork rate is checkpoint-independent and seems to be simply due to the physical presence of the lesions. In fact, by combing we see a slower fork rate in response to MMS in cds1Δ cells than in wild-type cells as compared to untreated (61% vs 76%, Fig 6A). The previously discussed correlation between fork rate and fork density in undamaged cells notwithstanding [66], we do not believe this effect in MMS-treated cds1Δ cells is an indirect consequence of the increased origin firing, because we do not see a similar decrease in fork rate in the 4NQO-treated cells, which show much greater increase in origin firing and fork density in cds1Δ cells relative to wild-type cells (Figs 6A, 5D and 5F). Instead, we prefer the interpretation that the checkpoint facilitates fork progression across a damaged template. Recent work shows that even in the absence of the checkpoint, the replisome is intact at stalled forks [77]. Therefore, it has been speculated that the checkpoint does not affect the stability of the replisome per se, but instead helps maintain the replisome in a replication competent state at sites of DNA damage [78], a possibility supported by our data. In contrast, previous studies have reported similar extent of slowing in both wild-type and the checkpoint mutants in response to MMS [27,28]. One explanation for the discrepancy could be that the previous methods—density transfer approach and BrdU-IP-seq—offer an average profile at lower resolution and mask the difference between wild-type and checkpoint-deficient cells. Reduction of fork rate in response to MMS but not to 4NQO supports the model that slowing of forks is simply due to a transient physical slowing of forks at each lesion encountered and is not due to a global checkpoint-dependent effect on fork progression rates. Consistent with this model, we do not see a reduction in fork rate in response to bleomycin treatment (1. 06, p = 0. 31), which activates the checkpoint robustly (Figs 4G and 7). This result is consistent with our previous observations that slowing is dependent on MMS dose [44] and occurs in response to frequent UV-induced lesions, but not rare IR-induced double-strand breaks [57]. Since the lesions induced by 4NQO and bleomycin are 25 and 50 times rarer than those caused by MMS, the forks are less likely to encounter damage and slow down in 4NQO-treated cells. The corollary to this conclusion is that activation of the checkpoint does not slow replication forks, as demonstrated by the fact that forks in 4NQO- and bleomycin-treated cells fail to slow despite the strong Cds1 activation and checkpoint-dependent inhibition of origin function. The interaction of a fork with a DNA lesion is a first step towards recognition of damaged template during S phase and is a critical mechanism for checkpoint activation [67]. Activation of the intra-S checkpoint by stalled forks allows the cell to activate repair pathways, tolerate damage and prevent genomic instability [68]. Although it is believed that forks can pass polymerase-blocking lesions on the leading strand using translesion polymerases, leading-strand repriming or recombinational lesion bypass, the role of the checkpoint in such responses is unclear [21–26]. Here we show that forks can bypass both MMS- and 4NQO-induced lesions and that the checkpoint is not required for that bypass. However, in the case of MMS-induced lesion, the checkpoint seems to facilitate bypass, since forks move past lesions more slowly in cds1Δ cells as compared to wild-type (0. 61% v. 0. 76%, Fig 6A). We also show that, whereas forks can bypass most lesions efficiently, at a small fraction of lesions—0. 5% of MMS-induced lesions and 5% of 4NQO-induces lesions—forks stall for the duration of the experiment. Given the large number of lesions throughout the genome, even these small numbers lead to a large number of stalled forks: 53% in MMS-treated cells and 25% in 4NQO- and bleomycin-treated cells. However, since any stall that occurs prior to the beginning of the second label will be identified by our analysis (see Methods for details), only a fraction of the stalls will occur during either of the labeling pulses. Detection of fork stalling events is complicated due to lack of consensus on how to identify them in the DNA fiber datasets [42]. Generally, signal from the first (red) analog alone on a fiber (a unlabeled-red-unlabeled [URU] event) is presumed to be either an elongating fork that stalled or an origin that fired in the first pulse followed by stalling of both its forks (Fig 2A) [7,42,60,79,80]. Alternatively the events from the first analog alone can be interpreted as terminations (Fig 2A) [42,81–83]. However, both interpretations rely on heuristic arguments and are unable to quantitate ambiguous signals, such as URU. Therefore we developed a new, rigorous way of quantitating stalled forks in double-labeled data. We have used the context in which the first analog event occurs to define it as a stalled fork or not. As discussed in greater detail in the Results and Methods sections we have used RUG and GUR patterns as a diagnostic for fork stall event occurring during the first analog (Fig 2B). We then used a probabilistic approach to quantitate the frequency of stall events in other ambiguous patterns (Fig 2C and S2 Fig). It should be noted that we can detect a stall only if it occurs before or during the first (red) analog pulse and persists throughout the second (green) analog pulse. Fork stalling does not appear to be simply an extreme example of the transient fork pausing that leads to observed fork slowing. If it were, we would expect to see a continuum of fork pause lengths form very short pauses to full stalls. Such heterogeneity would lead to a greater variation in apparent fork speeds and, in particular an increase in the asymmetry of rates in fork pairs, neither of which we see (S9 and S10 Figs and S1 Table). Therefore, we conclude that there are two distinct possible fate for a fork that encounters damage. It can pause briefly as it bypasses the lesion or it can stall permanently. A permanent stall does not appear to be a catastrophic event, as a majority of MMS-treated cells, which all have many stalls (S1 Table), do not have strong RPA foci (Fig 8A). However, such stalls do appear to require the checkpoint to restrain ssDNA accumulation. 4NQO-induced lesions appear to cause more severe stalls, since 4NQO-treated cells have many more strong RPA foci, even though they have fewer stalls (S1 Table and Fig 8A). The replication fork dynamics that we observe in response to MMS- and 4NQO-induced DNA damage demonstrate that forks interact with DNA damage largely in a checkpoint-independent manner. Forks are able to bypass lesions that stall the replicative polymerases with only a modest reduction in speed (Fig 6A) and are no more prone to stall at lesions in the absence of checkpoint (Fig 6B and 6C). However, stalled forks do appear to accumulate more ssDNA in the absence of the checkpoint (Fig 8). These results suggest that the major role of the checkpoint is not to regulate the interaction of replication forks with DNA damage, per se, but to mitigate the consequences of fork stalling when forks are unable to successfully navigate DNA damage on their own.
The following strains used in this study were created by standard methods and grown in YES at 25°C [84]: yFS105 (h- leu1-32 ura4-D18), yFS940 (h+ ura4-D18 his7-366 cdc10-M17 leu1: : pFS181 (leu1 adh1: hENT1) pJL218 (his7 adh1: tk) ), yFS941 (h- ura4-D18 his7-366 cdc10-M17 cds1: : kanMX leu1-32: : pFS181 (leu1 adh1: hENT1) pJL218 (his7 adh1: tk) ), yFS956 (h+ leu1-32 ura4-D18 cdc10-M17 rpa1-GFP: : hph-MX6), yFS957 (h+ leu1-32 ura4-D18 cdc10-M17 cds1: : ura4 rpa1-GFP: : hph-MX6), yFS988 (h- ura4-D18 ade-? cdc10-M17 leu1-32: : pYJ294 (leu1 cds1-6his2HA). Cells were synchronized in G1 phase using cdc10-M17 temperature sensitive allele combined with centrifugal elutriation, which selects cells that have been arrested in G1 for as little time as possible [85]. Cells were grown to mid log phase at 25°C and arrested at 35°C for 2 hours followed by centrifugal-elutriation-based size selection at 35°C to collect cells that had most recently arrested in G1. The cells were then immediately released into S phase by shifting them to 25°C, untreated or treated with 3. 5mM MMS or 1 μM 4NQO or 16. 5 μM bleomycin. S-phase progression was followed by flow cytometry using a nuclei isolation protocol, as previously described [85] with some minor modifications. Briefly, 0. 6 O. D. of cells were pelleted every 20 minutes for 2 hours after release into S phase. Pelleted cells were fixed by resuspension in 70% ethanol and stored overnight at 4°C. Fixed cells were spheroplasted at 37°C for 1 hour in 0. 6 M KCl with 1 mg/ml Lysing enzyme (Sigma # L1412) and 0. 5 mg/ml Zymolyase 20T (Sunrise Science Products # N0766391). Cells were then washed with 0. 1 M KCl containing 0. 1% Triton X-100 followed by 20 mM Tris-HCl, 5 mM EDTA, pH 8. 0. The cells were then resuspended in 20 mM Tris-HCl, 5 mM EDTA, pH 8. 0 containing 250 μg/ml RNaseA and incubated at 37°C overnight. Cells were pelleted, chilled and sonicated for 7 seconds with a Sonifier (Branson Sonifier 450) equipped with a micro tip at power 5 and constant duty cycle to release nuclei. Nuclei were mixed with equal amount of 1x PBS containing 2 μM Sytox Green and analyzed by flow cytometry. S-phase progression values were obtained from the histograms as previously described [85]. We have used three different drugs (3. 5mM MMS, 1 μM 4NQO and 16. 5 μM bleomycin) to create significantly differing lesion densities (of 1 every 1 kb, 25 kb or 50 kb, respectively) in order to differentiate between global v. local regulation of forks. Conceivably, variable lesion density could be achieved by just titrating one of the drugs. However, we cannot achieve a 25-fold difference in lesion density by just titrating either drug alone. In case of 4NQO, increasing dose concentration to even 2 μM 4NQO almost inhibits replication (S3 Fig), while decreasing MMS dose below 3. 5mM to 0. 875mM greatly reduces the effect of damage on replication kinetics [44]. Cells were harvested from asynchronous log-phase cultures of yFS105 untreated or treated with 3. 5mM MMS for 1 hour. Genomic DNA was isolated by bead beating, followed by phenol-chloroform extraction and ethanol precipitation of DNA [84]. The genomic DNA was digested with EcoRV and ScaI at 37°C for 6–8 hours. The DNA was then heated at 65°C for 2 hours to create nicks at sites of MMS-induced lesions [53]. Restriction fragments were analyzed by alkaline agarose gel electrophoresis (0. 8%). The lanes were quantified using Image J software. The ratio of DNA molecules at peak1 and 2 remaining in the MMS-treated sample relative to the untreated samples was used to infer the lesion density, assuming a Poisson distribution of MMS-induced lesions and calculating the frequency of nicks predicted to result in the observed fraction of molecules with no nicks (S1A Fig). We have used the value estimated from peak1 (1 lesion per kb) for calculation. We maybe under-estimating the lesion density because we are not certain about the efficiency with which the 65°C treatment converts MMS-induced lesions to ssDNA breaks. Cells were harvested from asynchronous log-phase cultures of yFS105 untreated or treated with 16. 5 μM bleomycin for 1 hour. 10 O. D. of cells were pelleted, frozen in liquid N2 and stored at -80°C. Plugs were prepared as described before [38]. After the cell wall digestion, Proteinase K treatment and TE washes the plugs were processed and digested with EcoRV and ScaI as described [86]. The digested plugs were analyzed by neutral agarose gel electrophoresis (0. 8%). The lanes were quantified and analyzed as for MMS (S1B Fig). For S1B Fig, the length of DNA of peak1, which represents the exclusion volume of the gel, containing molecules from 10 to ~40 kb, was assumed to average 20kb. The value obtained for break per kb from both the peaks were averaged to get the estimate 1 break per 50 kb. However we suspect that we are over-estimating the number of breaks. perhaps due to shearing at single-strand nicks, since we are able to analyze fibers on average 450 kb long from bleomycin-treated wild-type samples by combing. | Faithful duplication of the genome is essential for genetic stability of organisms and species. To ensure faithful duplication, cells must be able to replicate damaged DNA. To do so, they employ checkpoints that regulate replication in response to DNA damage. However, the mechanisms by which checkpoints regulate DNA replication forks, the macromolecular machines that contain the helicases and polymerases required to unwind and copy the parental DNA, is unknown. We have used DNA combing, a single-molecule technique that allows us to monitor the progression of individual replication forks, to characterize the response of fission yeast replication forks to DNA damage that blocks the replicative polymerases. We find that forks pass most lesions with only a brief pause and that this lesion bypass is checkpoint independent. However, at a low frequency, forks stall at lesions, and that the checkpoint is required to prevent these stalls from accumulating single-stranded DNA. Our results suggest that the major role of the checkpoint is not to regulate the interaction of replication forks with DNA damage, per se, but to mitigate the consequences of fork stalling when forks are unable to successfully navigate DNA damage on their own. | Abstract
Introduction
Results
Discussion
Materials and methods | flow cytometry
cell cycle and cell division
cell processes
enzymology
dna-binding proteins
organisms
dna damage
fungi
model organisms
polymerases
dna replication
experimental organism systems
dna
molecular biology techniques
synthesis phase
enzyme inhibitors
schizosaccharomyces
research and analysis methods
proteins
schizosaccharomyces pombe
molecular biology
spectrophotometry
yeast
biochemistry
cytophotometry
kinase inhibitors
biomolecular isolation
cell biology
nucleic acids
dna isolation
genetics
biology and life sciences
yeast and fungal models
spectrum analysis techniques | 2017 | Replication fork slowing and stalling are distinct, checkpoint-independent consequences of replicating damaged DNA | 12,763 | 279 |
Despite recent progress in proteomics most protein complexes are still unknown. Identification of these complexes will help us understand cellular regulatory mechanisms and support development of new drugs. Therefore it is really important to establish detailed information about the composition and the abundance of protein complexes but existing algorithms can only give qualitative predictions. Herein, we propose a new approach based on stochastic simulations of protein complex formation that integrates multi-source data—such as protein abundances, domain-domain interactions and functional annotations—to predict alternative forms of protein complexes together with their abundances. This method, called SiComPre (Simulation based Complex Prediction), achieves better qualitative prediction of yeast and human protein complexes than existing methods and is the first to predict protein complex abundances. Furthermore, we show that SiComPre can be used to predict complexome changes upon drug treatment with the example of bortezomib. SiComPre is the first method to produce quantitative predictions on the abundance of molecular complexes while performing the best qualitative predictions. With new data on tissue specific protein complexes becoming available SiComPre will be able to predict qualitative and quantitative differences in the complexome in various tissue types and under various conditions.
Mass-spectrometry (MS) techniques solved many fundamental issues in the identification of protein complexes [1–3] and other high-throughput techniques allowed the identification of Protein-Protein Interactions (PPI) and Domain-Domain Interactions (DDI), which paved the way for computational methods to predict protein complexes [4,5]. Validation of these computational approaches is based on the existence of data on detected protein complexes in the budding yeast Saccharomyces cerevisiae [6–9] and on initial data on Homo sapiens [10,11]. Unfortunately, all existing complex prediction methods produce only qualitative results even though protein complexes are formed dynamically and in various amounts throughout cell life. Note also that proteins with low abundance and with many possible binding partners might limit complex formation [12]. Therefore, it is crucial to predict the quantity of protein complexes. Graph theory algorithms to predict clusters that match protein complexes [13–15] or replicate structural properties of protein complexes retrieved from in vitro experiments have been applied [14]. Recently a new clustering algorithm [15] considerably improved predictions by allowing the overlapping of protein complexes with a reference protein-protein interaction network (PPIN). Herein, we propose a method which simulates dynamic complex formation that relies on complementary binding sites of proteins and that considers absolute protein levels [16,17] as initial number of molecular entities, in order to predict both the existence of a particular complex and its quantity. Protein binding sites correspond to domains and merging DDI and PPI data we built a proteome-wide model of all interactions in S. cerevisiae and H. sapiens. We consider DDIs only between proteins with a corresponding PPI, but the same domain of a given protein can be bound by multiple proteins with matching DDI and PPI leading to competition for binding sites and limiting formation of unrealistically large complexes. This ensures that proteins with high number of possible interactors do not interact with all possible partners at the same time and limits the size of such complexes [18]. The method was tested on protein complex prediction and it produced both exceptional qualitative results and the first quantitative prediction on protein complexes. We have also examined how the addition of a drug (the proteasome inhibitor, bortezomib in this case) influences the complexome in a qualitative and quantitative fashion. This served as a proof of concept towards protein complex prediction based drug design [19,20].
Our approach considered protein domains, retrieved with SMART [21], together with their corresponding DDI [22] and PPI [23]. We ran stochastic simulations for a reaction-diffusion system where multiple instances of proteins (corresponding to the square root of detected protein levels [16]) move and interact randomly on a two-dimensional logical space inspired by the Gillespie MultiParticle algorithm (GMP) [24]. We considered the square root of the absolute protein expression levels and a 2D simulation environment to reduce the computational cost, while keeping the possibilities for all proteins to meet any other protein in a reasonable time (S1 Text). In classic Gillespie algorithm [25], space is not explicity considered and the diffusion of molecules is assumed to be only a limiting factor on the reaction rates. It is absolutely important to consider space when simulating protein complex formation since closely located proteins or proteins that already participate in the same complex should have higher probability to bind with each other. Therefore, simulation algorithms that do not consider space cannot capture the right behavior of complexation and decomplexation of proteins. We considered a two-dimensional simulation space instead of the real three dimensional structure (3D) of the cell, because a well-discretized 2D space is already enough to reduce the probability of distant proteins to bind each other. Consideration of the real 3D structure of the cell would make the simulations more realistic, but the increase of computational costs would be outweighed by the benefits of considering diffusion in the third dimension. We divided the simulation space into square lattices, called sub-volumes (SV), where proteins are diffused randomly between neighbour lattices at discrete time steps. As a further simplification, we used the same diffusion rate for each protein (this could be improved with proteome wide data on diffusion rates becoming available). Proteins are represented as complex objects with binding sites corresponding to domains similarly to the BlenX modelling language [26]. Complementary binding sites can interact to form complexes and their bonds can break and lead to sub-complexes (Fig 1A). The inclusion of domains as binding sites allows competition between proteins for a given binding site. As a result, in our simulations two proteins cannot bind to the same domain, which can have an impact on the formation of protein complexes when two competing proteins are present in different abundances. Due to molecular crowding and to the stochastic nature of interactions, the simulations might lead to results that depend on the initial position of each molecule. To reduce this effect, we consider multiple simulation runs with random initial conditions and protein complexes are extracted from multiple time points of the simulation. The reported results came from two simulations with different initial localizations of proteins in space. We collected the list of simulated protein complexes at two separate simulation time points (four points in total). We found that more than two simulation runs do not increase the overall performance of the method: more complexes can be found, but prediction accuracy decreases (S1 Fig). This finding could be explained by the combination of the robustness of our simulation based method (See S1 Text for details) and the limited information on protein complexes reported in reference datasets [6,9]. Indeed a single simulation is enough to identify 90% of reference complexes, and two more simulations increase the percentage only to 91% (S1 Text). A complication comes from discrepancies between PPIs and DDIs data: only 34% of protein pairs involved in a PPI have a corresponding DDI between identified domains of the involved proteins. To enable interactions between proteins involved in a PPI, but missing proper DDI pair, we added fictitious interacting domains. Various strategies were tested (S1 Text) and the best solution was found to be the addition of fictitious domains to a pair of interacting proteins only if they are involved in the same biological function, according to MIPS [6]. Therefore, we consider DDI information between a pair of proteins only if a corresponding PPI exists. This step increased the ratio of PPIs with corresponding DDI to 84%. Our yeast model consists of 1474 proteins and the model based on data from humans contains 2342 proteins (Table 1). The presence of fictitious domains in some of the predicted complexes cannot be used to reject a prediction but if the fraction of fictitious domains in a complex is low it strengthen the predictions, as it is based on known domain-domain interactions. The computation intensive simulations were run on GPUs supporting CUDA (details in S1 Text). The simulation produces a list of complexes together with their structure (Simulated Complexes, SCs). Many SCs are constituted of similar set of proteins (S1 Text). To quantify how many overlapping complexes we detected, we apply a refinement process based on a frequency matrix where each element represents how many times a pair of proteins interacted in SCs [14]. Clustering this matrix generates the refined complexes (RCs) and the total number of SCs associated with a RC gives the quantity of that RC. Abundance of a RC is used to further increase the performance of SiComPre by dropping complexes below a threshold abundance and above or below a threshold size (details in S1 Text). An overview of the algorithm that we named SiComPre for Simulation based Complex Prediction is shown in Fig 1A.
The quality check of predicted protein complexes was done by comparing them with experimentally detected complexes from various sources [6,9]. We used well accepted scoring methods to compare predicted and experimentally detected complexes: recall gives the fraction of properly predicted complexes [13]; maximal matching ratio (MMR) measures the ratio of one-to-one matching between reference and predicted complexes [15] and the geometric accuracy is a function of proper and improper protein associations to complexes [27] (S1 Text). A sum of these scores leads to a global measure (composite score) quantifying the performance of the prediction [15]. Qualitatively similar results measured by an alternative scoring system (called f-score) [13] are discussed in S1 Text. The scores of SiComPre and existing algorithms for budding yeast are presented in Fig 1B. We also show how SiComPre scores change at various steps of our prediction method (Fig 1A): stopping the process at the simulated complexes (SiComPre SIM) after clustering (SiComPre CL). The composite score of SiComPre CL are equal or higher than any previous methods (Fig 1B and S1 Text). Since we can quantify the abundance of each predicted complexes, we could evaluate how SiComPre performs when low abundance complexes are dropped from the list. Two alternative versions were tried by dropping low abundance large size complexes (SiComPre-LG) or low abundance small size complexes (SiComPre-SM) (S1 Text) and found that SiComPre-LG outperforms all other methods on the basis of the composite score and SiComPre-SM works the best in the alternative f-score prediction measurement system (S1 Text). This highlights that both scoring systems differentially penalize wrong predictions of large and small complexes but SiComPre still performs well in both systems. Other protein complex prediction methods could be investigated, but ClusterOne was already proved to perform better than each of these [15]. Note also that the clustering and dropping of low abundance complexes slightly reduces the recall, but increases accuracy, thus the direct simulation results could be used to predict higher fraction of complexes (0. 9055 instead of 0. 874), but with lower composite score (2. 2573 instead of 2. 3472). S1 Text show that use of alternative databases with somewhat differing PPIs [23] or changes in initial data or in prediction scoring [13] do not change the high performance of all versions of SiComPre. The full process generated 657 protein complexes. 248 of these have an overlap score ≤ 0. 25, thus these are considered as newly predicted protein complexes (Tables 1 and S1). We also tested whether the consideration of protein abundances can effectively improve the qualitative predictions of SiComPre. We ran simulations where the abundance of all proteins were set to the average (7491 molecules) of all protein abundances in the input dataset [16]. The qualitative results show a decrement of the composite score compared to simulation that uses actual protein abundances: composite score after the simulation step is 2. 16 compared to 2. 26 of using actual protein abundances (S2 Fig). Therefore, incorporating protein abundance information is improving qualitative protein complex predictions. SiComPre has the best scores available in the literature and it can also predict the abundance of protein complexes by counting the number of SCs overlapping a RC that can then be associated with experimentally identified complexes. S1 Table lists all our predicted complexes with their predicted abundances and their associated best matching reference complexes. Comprehensive validation of these quantitative predictions is impossible at the moment since we lack a reference dataset on protein complex abundances. However, some of the predictions can be validated according complex abundances published in the literature [28–33] (Table 2). For instance we predicted ~2,200 copies of RNA polymerase complex I, ~3,900 RNA polymerase complex II and 144 RNA polymerase complex III. Data is available for the RNA polymerase II holoenzyme in haploid yeast in the range of 2,000 to 4,000 complexes [30]. The proportion of polymerases is maintained with respect to Mus Musculus, where their quantification is ~30,000, ~60,000 and ~3,000 respectively [29]. Approximately 50,000 copies of ribosomes were detected in our simulations that were based on the initial protein abundance data of 7. 0 × 104 on average for all ribosomal subunits. In logarithmic growing yeast cells the estimated ribosome number is 187,000 ± 56,000 ribosomes [28], but this calculation was based on the average concentration of 3. 15 × 105 subunits per cell and assumes that all ribosomal rRNAs are involved in ribosome formation, thus our quantitative prediction could be realistic. Even after the clustering and optimization steps, we found that multiple RCs that differ in either size or exact structure (Fig 2) are associated to a single experimentally characterized complex. For instance several alternatives of the ribosomal large subunit were found, which could be different existing variants or be caused by the lack of rRNAs in the simulations. SiComPre also predicted several RCs that we could not associate with any characterized complexes. We identified some complexes containing up to six proteins and several of them showed high abundances (>1000 copies per cell). Two of these six-protein complexes (RC 222 and 272 in S1 Table) share four common proteins and, according to functional annotations, are related to nuclear transport processes. These and other predicted complexes in S1 Table call for further research on their possible existence and role in yeast cells. Similar overlaps between RCs are plotted in Fig 3. This graph represents all the RCs (nodes) and the protein content overlapping between them (edges). Larger protein complexes are associated with more alternative RCs by SiComPre. The alterative RCs of an existing protein complex could be merged to increase the precision of predictions. However, alternative RCs could also correspond to existing variation of the same complex and could thus lead to the discovery of other proteins that associate and have functional relevance in an already known complex. An example of possible auxiliary subunits of the chromatin remodelling RSC complex [34] is highlighted in Fig 3a. SiComPre predicted that the RSC complex often interact with the ISW1a complex [35] and a new module of four proteins (Fig 3a). A similar subdivision of protein complexes has been proposed by Gavin et al. [5], where proteins are either part of the core of a complex or are attachments or part of modules bound to the core proteins. The core group is preserved in most of the isoforms of the complex, while attachments and modules may give a different function to the complex. Their analysis is based on genome-wide mass spectrometry data thus can be directly used to validate SiComPre results. All but two of the proteins in the most abundant SiComPre predicted RSC complex are part of the core of the RSC complex of Gavin et al [5] and ISW1a and the four proteins in the new module of the SiComPre complex are all attachments of the Gavin RSC complex (Fig 3b). This shows that SiComPre can be used not only to quantitatively predict protein complex abundances but also to predict possible alternative compositions of these complexes. As a control, we investigated whether protein complex abundances can be predicted simply by averaging the abundance of all the constituting subunits. We found that on average there is a 14-fold difference between SiComPre quantitative predictions of protein complexes and the average abundance of their constitutive subunits with low correlation between them (Pearson 0. 159, Spearman’s 0. 006 (S1 Text) ). The importance of the use of actual protein abundances in predicting protein complex abundances can be also seen on the predictions of the few examples with literature data (Table 2). We also tested SiComPre on the human PPI [10] with protein abundances from a human osteosarcoma cell line (U2OS) [3]. We validated the results against the CORUM dataset of mammalian protein complexes [11] from which redundancies and complexes smaller than three proteins have been removed [10,11]. This resulted in a new dataset of 324 non redundant human protein complexes, 39 of which were identified by SiComPre. Despite this relatively low match, our predictions outperform any other existing methods [10] (Fig 1B), predicting four more complexes than ClusterOne (Fig 1C). The low number of matched complexes is due to the lack of comprehensive experimental data, which cannot be compensated for by any prediction methods. It is possible that multiple instances of a predicted complex correspond to an existing, but so far unknown, complex. For instance, the high abundance six-protein complexes RC 145 and 504 (S2 Table) share five common proteins, which are all associated with snRNA binding (without our addition of any fictitious domain) and thus suggests the existence of these complexes. Indeed this complex appears under the name of LSM-complex in the extended CORUM dataset of 1685 characterized complexes from which SiComPre matches 295 protein complexes. Similarly, Complex RC 259 cannot be associated to any of the CORUM complexes (S2 Table) but it matches the pyruvate dehydrogenase complex [36] based on Uniprot protein descriptions. Examples of complexes which cannot be associated to any characterized complex are the five-protein complexes RC 365 and RC 391. These share four components and are associated with Rab related GTPase activity, vesicle formation and transport [37]. A few of the constituting proteins participate in the RCP-RAB11 and Rab geranylgeranyltransferase complexes but the whole complex does not show significant overlap with either of these. It is important to note that interactions between the constituting proteins of these complexes are always supported by known DDI, thus no fictitious domain had been added to predict these complexes. It is hard to find validation for the predicted abundances of protein complexes in the considered osteosarcoma cell line (U2OS). SiComPre predicted 1. 3 × 106 ribosomes, which is in the same order of the number of ribosome identified in HeLa cells (3. 3 × 106) [38]. SiComPre predicted 4456 different types of budding yeast protein complexes after two stochastic simulations and considering two separate time points. This is a surprisingly low number considering all the possible complexes that could appear from the initial PPI sub-network of interacting proteins of the proteome-wide yeast network (1622 nodes and 9074 edges). 2983 complexes were predicted by the first simulation with random initial position of each protein (we considered two time points of that single run). The second stochastic simulation with new, random initial settings, shared 1462 complexes with the first run. Similarity increases to almost 2030 complexes (68% of total hit counts) when complex with similar counterpart (overlap > = 0. 75) between two simulations are considered. Moreover, only one simulation is enough to identify 90% of reference complexes, while the addition of a second simulation increases the percentage of predicted reference complexes by only 1% (S1 Fig). Addition of further simulations does not increase predictive capabilities (Fig 1). This suggests that most complexes robustly form independently of the stochastic noise in the initial layout of proteins in the various sub-volumes. Quite often the protein complex abundances also show extraordinary robustness. A good example is the methionyl glutamyl tRNA synthetase complex [39] with abundances 2212 and 2252 (actual simulated values to the square to predict real biological abundances) in two simulation runs with a perfectly predicted structure match (matching score = 1). Many other complexes also have small abundance variations between the two simulations (S1 Table) and only small fraction of yeast complexes show high sensitivity to noise in initial settings. For instance, the abundance of the mRNA cleavage factor complex (CFI) varied between 172 and 302 copies and the Pho85p/Pcl8p complex was not observed in the first simulation but 64 (8 simulated) complexes appeared in the second run. To get a broader picture for each complex we calculated the coefficient of variation (CV = standard deviation / mean) of its quantitative predictions. Only 20% of complexes show a CV > 0. 5) in the case of the human protein complex predictions after three runs (S3 Table). In yeast, where only two runs were performed, 36% of complexes have CV > 0. 5. Finally we compared the quantitative predictions resulted from two separate simulations and observed that quantitative predictions are also robust as the two simulations on the yeast data gave quantitative protein complex predictions with a Pearson’s correlation of 0. 997, while based on results from the human data this correlation was 0. 998 (S2 and S3 Tables). Tissue-specific protein data is emerging [40] and shows that protein expression and abundance can greatly vary between tissue [17]. SiComPre can take such tissue-specific information into account and thus give tissue-specific protein complex predictions, which could soon be useful in extending our knowledge of human protein complexes. Furthermore the tissue specific variations in protein levels in cancer and other diseases [41] could be translated into qualitative and quantitative predictions on protein complexes by SiComPre. These results could be used to associate complex abundances and compositions with diseases as novel therapeutic targets [20]. For instance the administration of a drug can influence the abundance of complexes or allow the formation of new complexes. As a proof of concept, we performed simulations on the human SiComPre model with the addition of Bortezomib, a proteasome inhibitor [42] (details in S1 Text), which is a highly characterized drug with known to affect the formation of protein complexes. Drug—protein interactions were collected from the Stitch database [43] and after performing a domain enrichment of interacting proteins we estimated the domains bound by the drug (S1 Text). We set the abundance of Bortezomib to 50002 molecules, which is roughly the abundance of the most abundant protein we consider for human cells. We mapped protein complexes predicted with and without Bortezomib by finding the best matching complexes in the normal case to the complexes found after Bortezomib addition. Abundances of matching protein complexes were analysed by a t-test to find complexes which were perturbed in their abundance by Bortezomib (p-value <0. 05). Protein complexes without a best matching complex were considered qualitatively altered by Bortezomib. We observed that the abundance of the Proteasome, the Anaphase-Promoting Complex, Prefoldin and the Multisynthetase complex were greatly perturbed by Bortezomib. We also observed that the composition of the above discussed snRNA binding LSM complex and several other predicted complexes were modified by drug treatment (S3 Table). Several of the altered complexes are involved in transcriptional regulation (constitutive proteins are known transcription factors [44]). We searched the literature for validation of the involvement of SiComPre predicted transcriptional complexes in Bortezomib treatment and found numerous transcription factors that could be implicated (S4 Table). These predictions cannot be trivially inferred from the direct interactions of Bortezomib [43] as most of the candidate transcription factors are part of larger complexes that are perturbed by Bortezomib. Thus, we can conclude that SiComPre could be used to predict qualitative and quantitative changes to complexomes upon drug treatment. Several protein complexes perform essential biological functions slowing down their evolution and allowing only co-evolution of their components [45]. We investigated how protein complex compositions and abundances change between organisms with the example of the Anaphase Promoting Complex (APC). Fig 4 shows the structures of SiComPre predicted APC in yeast, human and in human after Bortezomib treatment. All three SiComPre complexes show high overlap with the experimentally identified complexes [46]. The SiComPre predicted yeast APC complex shows an overlap of 0. 68 with the core of the APC complex found by Gavin et al [5] and with one protein exception, fully matches the orthologs of the predicted human APC (Fig 4). The Bortezomib treatment seems to cause a loss of ANAPC10 (the yeast ortholog is Doc1) from some of the SiComPre simulated APC complexes (Fig 4). Such loss of ANAPC10 could cause an S-phase block in the cell cycle [47]. The constituting proteins are not the only variations revealed by SiComPre. As expected, SiComPre predicted the abundance of APC in human almost one order of magnitude higher than in yeast. Unexpectedly SiComPre also predicts that the addition of Bortezomib further doubles the abundance of APC in human (S2 and S3 Tables), although the majority of these complexes might be defective due to the lack of ANPC10. Thanks to simulated complexes it was possible to estimate the number of unbound proteins. This can help us to identify which proteins are fully bound in complexes, thus might limit the formation of other complexes. As expected there is a negative correlation between fraction of unbound proteins and the number of interactions of proteins (Pearson’s correlation -0. 43 for yeast and -0. 39 for human data) as with more possible interactors there is a higher chance of ending up in a complex. Interestingly there is no correlation between the fraction of free subunits and the abundances (Pearson’s correlation 0. 03). A high number of proteins are present in high abundance have a few interactors, but fully used up in complexes (Fig 5 and S5 Table). For instance, in yeast, TEF4 (YKL081W) has only 6 interactions, present in 102,000 copies, which are all bounded in complexes. This and several others with low (many cases 0) free abundance could be limiting factors in protein complexes.
All SiComPre applications and datasets are provided to the community on a dedicated website (www. cosbi. eu/research/prototypes/sicompre). The scripts are available as supplementary S1 File. | Most proteins are biologically active only when part of a complex with other proteins of the same or other type. Hence, to unravel biological functions of proteins, it is important to identify the type of complexes they can form. Multiple copies of each protein are present in cells and some of these could be involved in multiple complexes, thus it is a challenging task to identify protein complex compositions and abundances of all possible complexes. In this article we propose an integrative computational approach able to predict protein complexes from existing data sources on protein-protein and domain-domain interactions and protein abundances. By merging this information we built a computational model of all proteins and their dynamic interactions. Using cell-specific data we performed multiple stochastic simulations to predict protein complexes specific to budding yeast and human cells. Our predictions on protein complex compositions are consistent with a manually curated dataset and, for the first time, provide an approximation of their abundances. Our simulations can also predict how perturbations by a drug can influence the composition and abundance of protein complexes. | Abstract
Introduction
Materials and Methods
Results
Discussion | 2015 | Qualitative and Quantitative Protein Complex Prediction Through Proteome-Wide Simulations | 6,310 | 237 |
|
Presented here is a genome sequence of an individual human. It was produced from ∼32 million random DNA fragments, sequenced by Sanger dideoxy technology and assembled into 4,528 scaffolds, comprising 2,810 million bases (Mb) of contiguous sequence with approximately 7. 5-fold coverage for any given region. We developed a modified version of the Celera assembler to facilitate the identification and comparison of alternate alleles within this individual diploid genome. Comparison of this genome and the National Center for Biotechnology Information human reference assembly revealed more than 4. 1 million DNA variants, encompassing 12. 3 Mb. These variants (of which 1,288,319 were novel) included 3,213,401 single nucleotide polymorphisms (SNPs), 53,823 block substitutions (2–206 bp), 292,102 heterozygous insertion/deletion events (indels) (1–571 bp), 559,473 homozygous indels (1–82,711 bp), 90 inversions, as well as numerous segmental duplications and copy number variation regions. Non-SNP DNA variation accounts for 22% of all events identified in the donor, however they involve 74% of all variant bases. This suggests an important role for non-SNP genetic alterations in defining the diploid genome structure. Moreover, 44% of genes were heterozygous for one or more variants. Using a novel haplotype assembly strategy, we were able to span 1. 5 Gb of genome sequence in segments >200 kb, providing further precision to the diploid nature of the genome. These data depict a definitive molecular portrait of a diploid human genome that provides a starting point for future genome comparisons and enables an era of individualized genomic information.
Each of our genomes is typically composed of DNA packaged into two sets of 23 chromosomes; one set inherited from each parent whose own DNA is a mosaic of preceding ancestors. As such, the human genome functions as a diploid entity with phenotypes arising due to the sometimes complex interplay of alleles of genes and/or their noncoding functional regulatory elements. The diploid nature of the human genome was first observed as unbanded and banded chromosomes over 40 years ago [1–4], and karyotyping still predominates in clinical laboratories as the standard for global genome interrogation. With the advent of molecular biology, other techniques such as chromosomal fluorescence in situ hybridization (FISH) and microarray-based genetic analysis [5,6] provided incremental increases in the resolution of genome analysis. Notwithstanding these approaches, we suspect that only a small proportion of genetic variation is captured for any sample in any one set of experiments. Over the past decade, with the development of high-throughput DNA sequencing protocols and advanced computational analysis methods, it has been possible to generate assemblies of sequences encompassing the majority of the human genome [7–9]. Two versions of the human genome currently available are products of the Human Genome Sequencing Consortium [9] and Celera Genomics [7], derived from clone-based and random whole genome shotgun sequencing strategies, respectively. The Human Genome Sequencing Consortium assembly is a composite derived from haploids of numerous donors, whereas the Celera version of the genome is a consensus sequence derived from five individuals. Both versions almost exclusively report DNA variation in the form of single nucleotide polymorphisms (SNPs). However smaller-scale (<100 bp) insertion/deletion sequences (indels) or large-scale structural variants [10–15] also contribute to human biology and disease [16–18] and warrant an extensive survey. The ongoing analyses of these DNA sequence resources have offered an unprecedented glimpse into the genetic contribution to human biology. The simplification of our collective genetic ancestry to a linear sequence of nucleotide bases has permitted the identification of functional sequences to be made primarily through sequence-based searching alignment tools. This revealed an unexpected paucity of protein coding genes (20,000–25,000) residing in less than 2% of the DNA examined, suggesting that alternative transcription and splicing of genes are equally important in development and differentiation [19,20]. The sequencing of DNA of various eukaryotic genomes, such as for murine [21,105] and primate [22,23] as well as many others, has enabled a comparative genomics strategy to refine the identification of orthologous genes. These genomic datasets have also enabled the identification of additional functional sequence such as cis-regulatory DNA [24–29] as well as both noncoding and microRNA [30–34]. Building on the existing genome assemblies, numerous initiatives have explored variation at the population level, in particular to generate markers and maps as a means of understanding how sequence variation evolves and can contribute to phenotype. The initial drafts of the two human genomes provided an excess of 2. 4 million SNPs [7,8] providing a platform for the initial phase of the HapMap project [35]. This ambitious project initially catalogued genetic variation at more than 1. 2 million loci in 269 humans of four ethnicities, enabling a definition of common haplotypes and resulting in tag SNP sets for these populations. The use of these data has already allowed the mapping and identification of susceptibility genes and loci involved in complex diseases such as asthma [36], age related macular degeneration [37], and type II diabetes [38]. Notwithstanding, there are limitations with current SNP-based genome-wide association studies, because they rely on reconstructing haplotypes based on population data and can be uninformative or misleading in regions of low linkage disequilibrium (LD). Further, association studies have been designed to detect common disease variants and are not optimized to detect rare etiological variants [39]. The ability to generate a diploid genome structure via haplotype phasing for the HapMap samples is limited by the SNPs that were genotyped and their spacing. By using LD measures, it was possible to identify diploid blocks of DNA averaging 16. 3 kb for Caucasians (CEU), 7. 3 kb for Yorubans (YRI), and 13. 2 kb for grouped Han Chinese and Japanese (CHB+JPT) [35]. However, LD varies across the genome, and regions of low LD, i. e. , high recombination, cannot be represented by haplotype blocks. Furthermore, these diploid blocks are incomplete because there may be unknown variants between the SNP loci sampled. These results do not permit a comprehensive definition of the sequence present at each allele nor the information that produces the relevant allelic combinations, which are essential in identifying the differences of biological information encoded by the diploid state. The ability to perform, in a practical manner, whole-genome sequencing in large disease populations would enable the construction of haplotypes from individuals' genomes, thus phasing all variant types throughout the genome without assumptions about population history. Clearly, to enable the forthcoming field of individualized genomic medicine, it is important to represent and understand the entire diploid genetic component of humans, including all forms of genetic variation in nucleotide sequences, as well as epigenetic effects. To understand fully the nature of genetic variation in development and disease, indeed the ideal experiment would be to generate complete diploid genome sequences from numerous controls and cases. Here we report our endeavor to fully sequence a diploid human genome. We used an experimental design based on very high quality Sanger-based whole-genome shotgun sequencing, allowing us to maximize coverage of the genome and to catalogue the vast majority of variation within it. We discovered some 4. 1 million variants in this genome, 30% of which were not described previously, furthering our understanding of genetic individuality. These variants include SNPs, indels, inversions, segmental duplications, and more complex forms of DNA variation. We used the variant set coupled with the sequence read information and mate pairs to build long-range haplotypes, the boundaries of which provide coverage of 11,250 genes (58% of all genes). In this manner we achieved our goal of the construction of a diploid genome, which we hope will serve as a basis for future comparison as more individual genomes are produced.
The individual whose genome is described in this report is J. Craig Venter, who was born on 14 October 1946, a self-identified Caucasian male. The DNA donor gave full consent to provide his DNA for study via sequencing methods and to disclose publicly his genomic data in totality. The collection of DNA from blood with attendant personal, medical, and phenotypic trait data was performed on an ongoing basis. Ethical review of the study protocol was performed annually. Additionally, we provide here an initial foray into individualized genomics by correlating genotype with family history and phenotype; however, a more extensive analysis will be presented elsewhere. The donor' s three-generation pedigree is shown in Figure 1A. The donor has three siblings and one biological son, his father died at age 59 of sudden cardiac arrest. There are documented cases of family members with chronic disease including hypertension and ovarian and skin cancer. According to the genealogical record, the donor' s ancestors can be traced back to 1821 (paternal) and the 1700s (maternal) in England. Genotyping and cluster analysis of 750 unique SNP loci discovered through this project support that the donor is indeed 99. 5% similar to individuals of European descent (Figure 1B), consistent with self-reporting. This is further corroborated by an extensive five-generation family history provided by the donor (unpublished data). Cytogenetic analysis through G-banded karyotyping and spectral karyotypic chromosome imaging reveals no obvious chromosomal abnormalities (Figure 2) that need to be considered in interpretation of genome assembly results or phenotypic association analyses. The assembly, herein referred to as HuRef, was derived of approximately 32 million sequence reads (Table S1) generated by a random shotgun sequencing approach using the open-source Celera Assembler. The approach used is similar in many respects to the whole-genome shotgun assembly (WGSA) reported previously [40], but there are three major differences: (i) HuRef was assembled entirely from shotgun reads from a single individual, whereas WGSA was based on shotgun reads from five individuals [7,40,41], albeit the majority of reads were from the same individual as HuRef; (ii) the approximate depth of sequence coverage for HuRef was 7. 5 versus 5. 3 for WGSA, although the clone coverage was about the same for both (Table 1) [7,40]; and (iii) the release of Celera Assembler as an open-source project has allowed us and others to continue to improve the assembly algorithms. As a consequence, we made modifications for the specification of consensus sequence differences found at distinct alleles. The multiple sequence alignment methodology was improved and reads were grouped by allele, thus allowing the determination of alternate consensus sequences at variant sites (see Materials and Methods). HuRef is a high-quality draft genome sequence as evidenced from the contiguity statistics (Table 2). Improving the assembly algorithms and increasing the sequencing depth of coverage (compared to WGSA) resulted in a 68% decrease in the number of gaps within scaffolds from 206,552 (WGSA) to 66,815 (HuRef) as previously predicted [40]. We also observed a more than 4-fold increase in the N50 contig size (the length such that 50% of all base pairs are contained in contigs of the given length or larger) to 106 kb (HuRef) from 23 kb (WGSA). We used a fairly standard, but arbitrary, cutoff of 3,000 bp (similar to what was used for WGSA) to distinguish between scaffolds that were part of the HuRef assembly proper versus partially assembled and poorly incorporated sequence (see Materials and Methods). This resulted in 4,528 scaffolds (containing 2,810 Mb) of which 553 scaffolds were at least 100 kb in size (containing 2,780 Mb), whereas WGSA had 4,940 scaffolds (containing 2,696 Mb) of which 330 scaffolds were at least 100 kb (containing 2,669 Mb). The scaffold lengths for HuRef (N50 = 19. 5 Mb) were somewhat shorter than WGSA (N50 = 29 Mb) primarily due to the difference in insert size for bacterial artificial chromosome (BAC) end mate pairs—HuRef 91 kb versus WGSA > 150 kb (Table 2) [41]. We determined that 144 of the 553 large HuRef scaffolds could be joined by two or more of the WGSA BAC mate pairs, and 98 more by a single WGSA BAC mate pair (see Materials and Methods), suggesting that use of large insert BAC libraries (>150 kb) would generate larger scaffolds. Genomic variation was observed by two approaches. First, we identified heterozygous alleles within the HuRef sequence. This variation represents differences in the maternal and paternal chromosomes. In addition, a comparison between HuRef and the National Center for Biotechnology Information (NCBI) version 36 human genome reference assembly, herein referred to as a one-to-one mapping, also served as a source for the identification of genomic variation. These comparisons identified a large number of putative SNPs as well as small, medium, and large insertion/deletion events and some major rearrangements described below. For the most part, the one-to-one mapping showed that both sequences are highly congruent with very large regions of contiguous alignment of high fidelity thus enabling the facile detection of DNA variation (Table S2). The one-to-one mapping to NCBI version 36 (hereafter NCBI) was also used to organize HuRef scaffolds into chromosomes. HuRef scaffolds were only mapped to HuRef chromosomes if they had at least 3,000 bp that mapped and the scaffold was mostly not contained within a larger scaffold. With the exception of 12 chimeric joins, all scaffolds were placed in their entirety with no rearrangement onto HuRef chromosomes. The 12 chimeric regions represent the misjoining of a small number of chimeric scaffold/contigs by the Celera Assembly [40], as detected with mate pair patterns [7,42], and are also apparent by comparison to another assembly (Materials and Methods). The 12 chimeric joins in the HuRef scaffolds were split when these scaffolds were assigned to build HuRef chromosomes. Inversions and translocations within the nonchimeric scaffolds relative to NCBI are thus maintained within the HuRef chromosomes. The final set of 24 HuRef chromosomes were thus assembled from 1,408 HuRef assembly scaffolds and contain 2,782 Mb of ordered and oriented sequence. The NCBI autosomes are on average 98. 3% and 97. 1% represented by runs and matches, respectively, in the one-to-one mapping to HuRef scaffolds (Table S3). A match is a maximal high-identity local alignment, usually terminated by indels or sequence gaps in one of the assemblies. Runs may include indels and are monotonically increasing or decreasing sets of matches (linear segments of a match dot plot) with no intervening matches from other runs on either axis. The Y chromosome is 59% covered by the one-to-one mapping due to difficulties when producing comparison between repeat rich chromosomes. In addition, the Y chromosome is more poorly covered because of the difficulties in assembling complex regions with sequencing depth of coverage only half that of the autosomal portion of the genome. The X chromosome coverage with HuRef scaffolds is at 95. 2%, which is typical of the coverage level of autosomes (mean 98. 3% using runs). However it is clear that the X chromosome has more gaps, as evidenced by the coverage with matches (89. 4%) compared with the mean coverage of autosomes using matches (97. 1%). The overall effects of lower sequence coverage on chromosomes X and Y are clearly evident as a sharp increase in number of gaps per unit length and shorter scaffolds compared to the autosomes (Figure 3). Similarity between the sex chromosomes is another source of assembly and mapping difficulties. For example, there is a 1. 5-Mb scaffold that maps equally well to identical regions of the X and Y chromosomes and therefore cannot be uniquely mapped to either (see Materials and Methods and Figure 3). From our one-to-one mapping data, we are also able to detect the enrichment of large segmental duplications [10] on Chromosomes 9,16, and 22, resulting in reduced coverage based on difficulties in assembly and mapping (Table S3). Since NCBI, WGSA, and HuRef are all incomplete assemblies with sequence anomalies, assembly-to-assembly mappings also reflect issues of completeness and correctness. We compared three sets of chromosome sequences to evaluate this issue (see Materials and Methods): NCBI with the exclusion of the small amount of unplaced sequences, HuRef, and WGSA (Table S2) were thus compared in a pairwise manner. The comparison of WGSA and HuRef revealed 83 Mb more sequence in HuRef in matched segments of these genomes. This sequence is predominantly from HuRef that fills gaps in WGSA. Comparisons of HuRef and WGSA to NCBI showed the considerable improvement of HuRef over WGSA. Correspondingly, in HuRef there are approximately 120 Mb of additional aligned sequence, composed of 47 Mb of HuRef sequence that aligns to NCBI that was not aligned in WGSA and 73 Mb within aligned regions that fill gaps in WGSA. This comparison also showed an improvement factor of two in rearrangement differences (order and orientation) from WGSA to HuRef when mapped to the NCBI reference genome at small (<5 kb), medium (5–50 kb), and large (>50 kb) levels of resolution (Table S2). HuRef includes 9 Mb of unmatched sequence that fill gaps in NCBI or are identified as indel variants. An additional 14 Mb of HuRef chromosome sequence outside of aligned regions with NCBI represents previously unknown human genome sequence. The large regions of novel HuRef sequence are identified to be either: (a) gap filling or insertions, (b) unaligned NCBI chromosome regions, or (c) large scaffolds not mapped to NCBI chromosomes. Some of these were investigated using FISH analysis and are discussed below. Although we were able to organize HuRef scaffolds into HuRef chromosome sequence, all of the subsequent analyses in this report were accomplished using HuRef scaffold sequences. To examine sequence diversity in the genome, we estimated nucleotide diversity using the population mutation parameter θ [43]. This measure is corrected for sample size and the length of the region surveyed. In the case of a single genome with two chromosomes, θ simplifies to the number of heterozygote variants divided by the number of base pairs (see Materials and Methods). We define θSNP as the nucleotide diversity for SNPs (number of heterozygous SNPs/number of base pairs) and θindel as the diversity for indels (number of heterozygous indels/number of base pairs) [44]. For both θSNP and θindel, the 95% confidence interval would be [0,3θ] due to the small number of chromosomes (n = 2) being sampled (see Materials and Methods). Across all autosomal chromosomes, the observed diversity values for SNPs and indels are 6. 15 × 10−4 and 0. 84 × 10−4 respectively. When restricted to coding regions only, θSNP = 3. 59 × 10−4 and θindel = 0. 07 × 10−4, indicating that 42% of SNPs and 91% of indels have been eliminated by selection in coding regions. The strong selection against coding indels is not surprising, because most will introduce a frameshift and produce a nonfunctional protein. Our observed θSNP falls within the range of 5. 4 × 10−4 to 8. 3 × 10−4 that has been previously reported by other groups [44–47]. Our observed θindel (0. 84 × 10−4) is approximately 2-fold higher than the diversity value of 0. 41 × 10−4 that was reported from SeattleSNPs (http: //pga. gs. washington. edu), which was derived from directed resequencing of 330 genes in 23 individuals of European descent [44]. The values of θindel in repetitive sequence regions are 1. 2 × 10−4 for regions identified by RepeatMasker (http: //www. repeatmasker. org) and 4. 9 × 10−4 for regions identified by TandemRepeatFinder [48], respectively. Thus, the indel diversity in repetitive regions is between 1. 4 and 5. 8 times higher than the genome-wide rate. This suggests that the high value of θindel over all loci is likely mediated by the abundance of indels in repetitive sequence. It is also possible that repetitive regions in genic sequence are under stronger selective pressure and therefore have lower indel diversity. These are precisely the regions that have been targeted in previous resequencing projects [44] from which indel diversity values have been determined. Additionally, repetitive regions also have more erroneous variant calls due to technical difficulties in sequencing and assembly of these types of regions. Therefore, our estimate for θindel is likely a combination of both a true higher mutation rate in repetitive regions and sequencing errors. Values of θindel are consistent among the chromosomes (Figure 5). Chromosomes with high θindel values also have a larger fraction of tandem repeats. For example, Chromosome 19 has the highest θindel (1. 1 × 10−4 compared with the chromosomal average of 0. 86 × 10−4), and it also has the highest proportion of tandem repeats (13% compared with the chromosomal average of 7%). The fraction of tandem repeats of a chromosome is positively correlated with the value of θindel for each chromosome (r = 0. 73), so that the diversity of indels is associated with the underlying sequence composition. The SNP variants identified in the HuRef genome include a larger-than-expected number of homozygous variants than those commonly observed in population-based studies (compare ratios of heterozygous SNP: homozygous SNP in Table 5). Our homozygous variants are detected as differences between the HuRef genome and the NCBI genome. One common interpretation of a homozygous variant is that given a common allele A and a rare allele B, the homozygous SNP is BB. However, because not all variant frequencies are known, we cannot determine if a position may carry the minor B allele in homozygous form. We analyzed ENCODE data using this definition and found the ratio of heterozygous SNPs to homozygous SNPs is 4. 9 in an individual [49]. For our dataset, the observed ratio of heterozygous to homozygous SNP, where our “homozygous” SNPs are detected as bases differing from the NCBI human genome, is 1. 2. To resolve this discrepancy, we examined the homozygous positions in the HuRef assembly and found that the increased frequency of homozygous SNPs results from the presence of minor alleles (BB) in the NCBI genome assembly. We observed that 75% of the homozygous positions in HuRef also had a SNP identified by the ENCODE [49]. A comparison of the alleles at these positions revealed that in 56% of the instances the HuRef genome had the more common allele, whereas the NCBI genome contained the minor allele. The remaining homozygous SNPs tended to be common minor alleles (76% had minor allele frequency [MAF] ≥ 0. 30), consistent with their observation in homozygous form in the HuRef genome. Therefore, we confirmed that a large fraction of homozygous alleles from HuRef are real, and that differences between the HuRef and NCBI assemblies are due to NCBI containing the minor allele at a given SNP position, or HuRef containing a common SNP in homozygous form. We also modeled the inter/intraindividual genome comparison using directed resequencing data from SeattleSNPs data (see Materials and Methods) to determine if our variant detection frequencies were commonly found for different types of variants. By sampling and comparing the genotypes of two individuals from the SeattleSNPs data, we were able to simulate the conditions for calling “heterozygous” and “homozygous” variants as we have defined them in an independently generated set (Table 5). The ratio of heterozygous variants to homozygous variants from the modeled SeattleSNPs is lower in the HuRef genome compared with the SeattleSNPs data. This suggests that there are an overabundance of homozygous variants and/or an under-representation of heterozygous variants, and this trend is more pronounced for indels compared to SNPs. A possible explanation for this is that homozygous genotypes are actually heterozygous and the second allele is missed due to low sequence coverage. Our attempts to explain this phenomenon using statistical modeling did support our hypothesis that low sequence coverage resulted in excess homozygous over heterozygous variant calls. Indeed, our modeling provided us with a bound on the missed heterozygous calls for both indels (described below) and SNPs (see section below titled: Experimental Validation of SNP Variants). In an attempt to explain the discrepancy in the heterozygous to homozygous indel ratio (Table 5), we modeled the rate of identification of true heterozygous variants given the depth of coverage of HuRef sequencing reads and the various variant filtering criteria. This enabled us to determine that between 44% and 52% of the time, heterozygous indels will be missed due to insufficient read coverage at 7. 5-fold redundancy and these indels be erroneously called homozygous. Therefore, the projection for the true number of homozygous indels is between 418,731 and 459,639, a reduction of 17%–25% from the original number of 559,473 homozygous indels, and the corresponding ratio of heterozygous to homozygous indels is between 1: 1 and 1. 3: 1. Furthermore, our modeling also allowed us to determine that approximately 20× sequence coverage would be required to detect a heterozygous variant with 99% probability in unique sequence given our current filtering criteria of random shotgun sequence reads. Another further explanation for the overabundance of homozygous indels is the error-prone nature of repeat regions. Using a subset of genes (55) completely sequenced by SeattleSNPs, we found that 28% of the potential 92 HuRef homozygous indels overlap with indels in these genes, as opposed to 75% confirmation rate for homozygous SNPs described earlier. When one categorizes the repeat status of a homozygous indel, a higher confirmation rate (46%) is seen for indels excluded from regions identified by RepeatMasker or TandemRepeatFinder. The confirmation rate for an indel in a transposon or tandem repeat region is much lower at 16%. Therefore, indels in nonrepetitive loci have a higher probability of authenticity than indels in repeat regions. The ratio of SNPs to indels is lower in the HuRef assembly than what is observed by the SeattleSNPs data (Table 5), indicating that relatively fewer SNPs or relatively more indels are called. This is likely due to relatively more indels being identified, as discussed above. We note that a large fraction of indels occur in repeat sequence (Table 6), which has higher indel frequency as well as higher incidence of sequencing error. Moreover, SeattleSNPs resequencing data is focused on variant discovery in genic regions, which may not reflect genome-wide indel rates. We identified in the HuRef assembly 263,923 heterozygous indels spanning 635,314 bp, with size ranges from 1 to 321 bp. The characteristics of the indels we detected, their distribution of sizes <5 bp, and the inverse relationship of the number of indels to length are similar to previous observations [50,51] (Figure 6A and 6B). As noted previously (Table 6), there are 2-fold more homozygous indels (559,473) than heterozygous indels, and these span 5. 9 Mb and range from 1 to 82,771 bp in length. We observe that genome-wide, even-length indels are more frequent than odd-length indels (Figure 6C and 6D, χ2 = 12. 4; p < 0. 001, see Materials and Methods). One possible explanation for these results is that tandem repeats often have motif sizes that occur in even numbers, such as through the expansion of dinucleotide repeats. In fact, based on RepeatMasker, the majority of simple repeats are composed of even-numbered–sized motifs rather than odd-numbered–sized motifs (73%). Furthermore, of the heterozygous indels that occur in simple repeats identified by RepeatMasker, 79% occur in even-numbered bp repeats. This suggests that the preponderance of even-base–sized indels likely results from the inherent composition of simple repeats. There are 6,535 homozygous indels that are at least 100 bases in length for which both flanks of the indel can be located precisely on HuRef and NCBI assemblies. These comprise 3,431 insertions uniquely occurring on HuRef, totaling 2. 13 Mb, and 3,104 deletions, totaling 1. 82 Mb, found only on NCBI (Figure 7). These homozygous indels have a higher representation of repetitive elements (66%–67%) than the overall HuRef and NCBI assemblies (each 49%). This enrichment derives mainly from a higher relative content of short interspersed nuclear elements (SINEs), simple repeats, and unclassified SVAs (Table 7). For 657 (19% of the total) insertions with a minimum length of 100 bp, at least 50% of the segment length (mean = 95%) is composed of a single SINE insertion. Most of these SINE insertions (88%) belong to the youngest Alu family (AluY), for which insertion polymorphisms are well documented in the human genome [52,53]. Similarly, for 26% of deletions at least 100 bp in length, an average of 95% of the segment consists of a single SINE element, and 92% of these elements are classified as AluY. Interestingly, the combined total of 1,316 AluY insertions that differ between HuRef and NCBI include 703 (53%) that are not currently identified in the most comprehensive database of human bimorphic SINE insertions, the database of retrotransposon insertion polymorphisms in human (dbRIP; 1625 loci; http: //falcon. roswellpark. org: 9090/) (Table S4) [54]. To evaluate the accuracy and validity of SNP calling from the sequencing reads, the donor DNA was interrogated using hybridization-based SNP microarrays: the Affymetrix Mapping 500K Array Set, which targets 500,566 SNP markers, and the Illumina HumanHap650Y Genotyping BeadChip, which targets 655,362 SNPs. The Affymetrix array experiment was performed twice to provide a technical replicate for genotyping error estimation, and 0. 12% of genotype calls were discordant. Of the 92,144 assays with an annotation in dbSNP that overlap between the two different platforms, 99. 87% were concordant (0. 13% discordant). Thus, the discordance rate between platforms was similar to that between Affymetrix technical replicates. Genotype calls that were discordant between technical replicates or between the Affymetrix and Illumina platforms were excluded from further analysis. This resulted in 1,029,688 nonredundant SNP calls from the two genotyping platforms, which were then compared to the HuRef assembly and to the single nucleotide variants extracted from the sequencing data. Of these, 943,531 genotypes (91. 63%) were concordant between the genotyping platforms and the HuRef assembly (Table 8). Of the 86,157 discordant genotype calls, the vast majority (83. 9%) were identified as heterozygous in the merged genotyping platform data, but called as homozygous in the HuRef assembly (Table 9). This is consistent with a predictable effect of finite sequence coverage in the HuRef dataset: assuming uniform random sampling of both haplotypes, 21. 6% of true heterozygous SNPs are expected to be missed given 7. 5× coverage of the diploid genome and the requirements for calling a heterozygous SNP (i. e. , at least two instances of each allele and ≥20% of reads confirming the minor allele). This is close to the observed false-negative error of 24. 6% (Table 9 and Figure 8). Consistent with this explanation, the level of coverage is significantly lower for the missed heterozygous SNPs than for the heterozygous SNPs detected in the HuRef assembly (average read depth 5. 2 and 8. 8, respectively) (Figure 9). Another possible form of error would be to erroneously call a truly homozygous position a heterozygous variant. Of the 65,337 homozygote calls that were concordant between the Affymetrix and Illumina platforms, none were called as heterozygous in the HuRef assembly. Therefore, the upper bound for the false-positive rate is 0. 0046% (one-tailed 95% confidence interval), and one would expect false-positive heterozygote calls approximately once every 22 kb from the upper bound of this confidence interval. However, this estimate may be lower than the genome-wide false-positive error, because it is based on the positions chosen by the microarray platforms, which tend to be biased away from repetitive, duplicated, and homopolymeric regions. Approximately three-quarters of the novel heterozygous SNPs (73%) and novel heterozygous indels (75%) are in a region identified by RepeatMasker, TandemRepeatFinder, or a segmental duplication. Therefore, approximately three-quarters of the novel heterozygous variants are in regions that are most likely underrepresented in the microarrays. Consequently, we cannot readily extrapolate the false-positive error determined from the microarrays to be the discovery rate of the HuRef variant set. The repetitive regions are likely to have a higher false-positive rate due to sequencing error and misassembly. Further, they are not represented in the current estimate of the false-positive rate. However, they also exhibit a higher rate of authentic variation. Homozygous and heterozygous insertions and deletions identified in the HuRef assembly were computationally validated by comparison to previously published datasets. As indicated in Figure 4, the homozygous insertion and deletions variants are operationally defined as either inserted or deleted sequence in the HuRef genome respectively since there is no other read evidence for heterozygosity. The homozygous nature of these variants does not imply any notion of ancestral allele. The largest set of indel variants that has been published is based on mapping of trace reads to the NCBI human genome reference assembly [55]. This approach can be used to identify deletions of any size and insertions that are small enough to be spanned by sequence reads. In this analysis, the 216,179 deletions and 177,320 insertions from Mills et al. [55] were compared to the insertions and deletions identified from the HuRef assembly. Based on this analysis, we found support for 37,893 homozygous deletions and 46,043 homozygous insertions that overlapped between the two datasets (Table 11). Comparison with the heterozygous deletions and insertions from the HuRef assembly yielded support for 9,431 deletions and 7,738 insertions, respectively (Table 10). These values represent a lower limit due to possible alignment issues in regions with tandem repeats. This dataset produced the largest overlap with the HuRef variant set compared to all others discussed below. However the Mills et al. published dataset used reads from the NCBI TraceArchive that we also used during assembly (i. e. , Celera reads, donor HuBB). This suggests that essentially the same dataset used by two different groups produced an overlapping result by using different methods. As a consequence, we cannot determine which part of the overlapped variants with the Mills et al. data came from non-Celera sources, and therefore we cannot comment on novelty or polymorphic supporting evidence for HuRef variants. Next, the HuRef homozygous deletions were compared to three other sets of previously identified deletion polymorphisms [56–58]. However, the overlap with these datasets was minimal, possibly due to the larger size of these variants (Table 11). Finally, the set of HuRef homozygous insertions was compared to those variants identified in an assembly comparison approach [59], and support was found for additional 243 insertion variants. We sought further evidence in support of the longest indels identified by the one-to-one HuRef–NCBI mapping. We focused on the 20 longest insertions (9–83 kb) and the 20 longest deletions (7–20 kb) and examined the presence of these large indels in the genomes of eight other individuals by identifying fosmid clones that map to these 40 loci (Table S5). The fosmid mapping provided support for all 20 insertions, and 17 of 20 deletions. The lack of support for two of the deletions (Unique Identifiers 1104685056026,1104685093410) is likely due to their location at the ends of HuRef scaffolds, which greatly reduces the possibility of mapping fosmids that span the insertion site. Support from multiple fosmids provides the strongest evidence for variation in indels between individuals. For example, the presence of a 24 kb insertion on Chromosome 22 (Unique Identifier 1104685552590) is supported by 13–17 fosmids in three individuals (with no evidence for absence), whereas its absence is supported by 19 fosmids in another individual (with no evidence for presence). These data suggest that the majority of large indels defined by the one-to-one HuRef–NCBI mapping are genuine variations among human genomes. We selected 19 non-genic heterozygous indels in a nonrandom manner, ranging in length from 1 to 16 bp, for experimental validation using PCR coupled with PAGE detection of allelic forms. We ensured that the read depth coverage was in an acceptable range (not greater than 15 reads), suggesting that these loci were not in segmental duplications and would therefore not produce spurious PCR amplification. Three Coriell DNA samples and HuRef donor DNA were examined, and 15 out of 19 PCR assays assessed generated results consistent with the positive and negative controls. The indel lengths that yielded experimental data ranged from 1 to 8 bp in length. In four out of 15 indels, the heterozygote variant was identified in all four DNA samples, and in three out of 15, it was only found the HuRef donor DNA. For the remaining eight out of 15 cases, the indels were differentially observed among the four DNA samples (Figure S1). We selected 51 putative homozygous HuRef insertions in a nonrandom manner for validation in 93 Coriell DNA samples based on their proximity to annotated genes, their size range of 100–1,000 bp, the absence of transposon repeat or tandem repeat sequence, uniqueness in the HuRef genome, and the absence of any similarity to chimpanzee sequence. The experimental results (Table S6) indicated that for 43 of 51 insertions (84%), we were able to generate specific PCR products for which the size of PCR products were as predicted and fell within the detectable range of the gel. For 84% of these 43 cases, insertions were identified in HuRef and additional DNA samples, and most follow Hardy-Weinberg equilibrium in CEU samples. Approximately 7% of the insertions tested (3 of 43) were false positives, because the HuRef donor DNA and all the 93 Coriell DNAs were homozygous for no insertion. In four insertions (9%), all of the tested Coriell samples displayed normal Hardy-Weinberg equilibrium; however, the insertion was absent in the HuRef sample. The inability to observe the insertion in the HuRef sample in these instances might be due to allelic dropout in the PCR process for the HuRef sample. This could be caused by specific SNPs at the primer annealing sites that were not accounted for during the primer design process. In 22 (61%) confirmed experiments, the HuRef donor bears homozygous insertions in agreement with our computational analyses. There are four insertions in this set, among the 22, where the HuRef donor and all 93 Coriell DNA donors tested were homozygous for insertions. This suggests that these sequences were either not assembled in the NCBI human genome assembly or that the NCBI donor DNA sequenced had a rare deletion in these regions. For the remaining 14 insertions (39%), the HuRef donor was heterozygous for the insertion instead of homozygous as was predicted by our indel detection pipeline. We searched for these alternative shorter alleles in the HuRef assembly and observed that two of the alternative alleles matched degenerate scaffolds and two matched singleton unassembled reads. These are sequence elements that are typically small or unassembled elements respectively, signifying that the assembly process selected one allele. We note that many of the insertions tested (84%) are polymorphic in the Coriell panel tested, and although many are intronic, there are instances of UTR and exonic insertions whose impact on function may be more directly ascertained. It has previously been shown that extended regions of high sequence identity complicate de novo genome assembly [10,60,61]. An analysis was undertaken to assess how well the segmental duplications (identified as regions of >5 kb with >90% sequence identity) annotated in the NCBI assembly are represented in the HuRef genome sequence. We analyzed the NCBI sequence (90. 1 Mb) external to the one-to-one mapping with the NCBI assembly for segmental duplication content by comparison to the Human Segmental Duplication Database (http: //projects. tcag. ca/humandup/) [61]. More than 70% of these nucleotides (63. 6 Mb) are contained within segmental duplications, compared with 5. 14% across the entire NCBI assembly. This suggests that the regions of the NCBI assembly that are not aligned to HuRef likely result from the absence of assembled segmental duplication regions in HuRef. This is further supported by the fact that only 57. 2% of all regions annotated as segmental duplications in NCBI are present in HuRef. Clearly, these are some of the most difficult regions of the genome to represent accurately with a random shotgun approach and de novo assembly. However, it is also important to note that at least 25% of segmental duplication regions differ in copy number between individuals [62], and the annotation of such sequences will certainly differ between independent genomes. Copy number variants (CNVs) have been identified to be a common feature in the human genome [11,15,62–64]. However, such variants can be difficult to identify and assemble from sequence data alone, because they are often associated with the repetition of large segments of identical or nearly identical sequences. We tested for CNVs experimentally to compare against those annotated computationally, and also to discover others not represented in the HuRef assembly. We used comparative genomic hybridization (CGH) with the Agilent 244K array and Nimblegen 385K array, as well as comparative intensity data from the Affymetrix and Illumina SNP genotyping platforms (using three analysis tools for Affymetrix and one for Illumina). In total, 62 CNVs (32 losses and 30 gains) were identified from these experiments (Table S7). It is noteworthy that the Agilent and Nimblegen CGH experiments, as well as the analysis of Affymetrix data using the GEMCA algorithm, were run against a single reference sample (NA10851). Therefore, a subset of the regions reported as variant may reflect the reference sample rather than the HuRef donor, even though all previously identified variants in the reference sample [62] were removed from the final list of CNV calls in the present study. The majority of the variant regions were detected by only one platform, reflecting the difference in probe coverage and sensitivity among various approaches [12,62]. As an independent form of validation, the CNVs detected here were compared to those reported in the Database of Genomic Variants (DGV) [63], and 54 of the variants (87%) have been described previously (with the thresholds used for these analyses we expect approximately 5% of calls to be false positive). A summary of the genomic features overlapped by these CNVs is presented in Table 12. Approximately 55% of the CNVs overlap with annotated segmental duplications, which is slightly higher than reported in previous studies [63,64]. The CNVs also overlap 95 RefSeq genes, seven of which are described in the Online Mendelian Inheritance in Man database (OMIM) as linked to a specific phenotype (Table S7). These include blood group determinants such as RHD and XG, as well as a gain overlapping the coagulation factor VIII gene. Numerous HuRef sequences that span the entire or partial scaffolds did not have a matching sequence in the NCBI genome. Some had putative chromosomal location assignments (e. g. , sequences extending into NCBI gaps), whereas others were unanchored scaffolds with no mapping information. We selected sequences >40 kb in length with no match to the NCBI genome and identified fosmids (derived from the Coriell DNA NA18552) mapping to these sequences based clone end-sequence data. The fosmids were then used as FISH probes with the aim of confirming annotated locations for anchored sequences and assigning chromosomal locations to unanchored scaffolds. Fosmids were hybridized to metaphase spreads from two different cells lines. At least 10 metaphases were scored for each probe, and a differentially labeled control fosmid was included for each hybridization. For 23 regions, there was no mapping information available from mate-pair data or the one-to-one mapping comparison. Of the remaining 26 regions, 24 had a specific chromosomal location assigned at the nucleotide level (Figure 10A and 10B), whereas two regions were assigned to specific chromosomes but lacked detailed mapping information. The results of the FISH experiments are outlined in Table S8. Of the 23 regions with no prior mapping information, 13 gave a single primary mapping location (Figure 10C). The majority of the remaining 10 regions located to multiple centromeric regions (Figure 10D), suggesting that there are large euchromatic-like sequences present as low-copy repeats in the current centromeric assembly gaps. For the 26 regions with mapping information, the expected signal was observed for 22 (85%). However, in six of these hybridizations, there were additional signals of equal intensity at other locations. Ten of the scaffolds chosen for FISH extend into contig or clone gaps in the current reference assembly. Of these 10 regions, the expected localization was corroborated for seven. The combined data indicate that the HuRef assembly contributes significant amounts of novel sequence important for generating more complete reference assemblies. Haplotypes have more power than individual variants in the context of association studies and predicting disease risk [65–67] and also permit the selection of reduced sets of “tagging” SNPs, where linkage disequilibrium is strong enough to make groups of SNPs largely redundant [68,69]. The potential for shotgun sequences from a single individual to be used to separate haplotypes has been examined previously [70,71]. For a given polymorphic site, sequencing reads spanning that variant can be separated based on the allele they contain. For data from a single individual, this amounts to separation based on chromosome of origin. When two or more variant positions are spanned by a single read, or occur on paired reads derived from the same shotgun clone, alleles can be linked to identify larger haplotypes. This is sometimes known as “haplotype assembly. ” When single shotgun reads are considered, the problem is computationally tractable [70,71] but the resulting partial haplotypes would be quite short with reads produced by existing sequencing technology, given the observed density of polymorphisms in the human genome (R. Lippert, personal communication). Mate pairing has the potential to increase the degree of “haplotype assembly, ” but finding the optimal solution in the presence of errors in the data has been shown to be computationally intractable [71]. Nevertheless, we show that the character and quality of the data is such that heuristic solutions, while not guaranteed to find the best possible solution, can provide long, high-quality phasing of heterozygous variants. The set of autosomal heterozygous variants described above (n = 1,856,446) was used for haplotype assembly. The average separation of these variants on the genome was ∼1500 bp (twice the average read length). Fewer than 50% of variants could be placed in “chains” of six or more variants where successive variants were within 1 kb of one another. Consequently, single reads cannot connect these variants into large haplotypes. However, the effect of mate pairing is substantially greater than would be observed simply by doubling the length of a read, as shown in Figure 11: variants are linked to an average of 8. 7 other variants. Using this dataset, haplotype assembly was performed as described in Materials and Methods. Half of the variants were assembled into haplotypes of at least 401 variants, and haplotypes spanning >200 kb cover 1. 5 Gb of genome sequence. The full distributions of haplotype sizes, both in terms of bases spanned and in terms of numbers of variants per haplotype, are shown in Figure 12. Although haplotypes inferred in this fashion are not necessarily composed of continuous variants, haplotypes do in fact contain 91% of the variants they span. More than 75% of the total autosomal chromosome length is in haplotypes spanning at least four variants, and 89% of the variants are in haplotypes that include at least four heterozygous HapMap (phase I) variants. Both internal consistency checks and comparison to HapMap data indicate that the HuRef haplotypes are highly accurate. Comparing individual clones against the haplotypes to which they are assigned, 97. 4% of variant calls were consistent with the assigned haplotype. Moreover, the HuRef haplotypes were strongly consistent with those inferred as part of the HapMap project [35]. Where a pair of variants is in strong LD according to the HapMap haplotypes, the correct phasing of the HuRef data would be expected to match the more frequent phasing in the HapMap set in most cases. Exceptions would require a rare recombination event, convergent mutation in the HuRef genome, or an error in the HapMap phasing in multiple individuals. We accessed the 120 phased CEU haplotypes from HapMap and identified the subset of heterozygous HuRef SNP variants that also coincided with the HapMap data. For adjacent pairs of such variants that were in strong LD (r2 ≥ 0. 9; n = 197,035), fewer than 1 in 40 of the HuRef-inferred haplotypes conflicted with the preferred HapMap phasing. Figure 13 shows more generally the consistency of HuRef haplotypes with the HapMap population data as a function of r2 and D′. Because the inference of HuRef haplotypes is completely independent of the data and methods used to infer HapMap haplotypes, this is a remarkable confirmation of the HuRef haplotypes. The restriction to variants in strong LD has no clear selection bias with respect to our inferred haplotypes. On the other hand, it provides only weaker confirmation for the HapMap phasing, since it is restricted to the easiest cases for phasing using population data—namely only those pairs of variants in strong linkage disequilibrium. The lengths and densities of the inferred HuRef haplotypes described above are possible due to the use of paired end reads from a variety of insert sizes. Given the relatively simple means that were used for separating haplotypes, the high accuracy of phasing is likewise due to the quality of the underlying sequence data, the genome assembly, and the set of identified variants. The rate of conflict with HapMap with regard to variants in high LD can be further decreased by filtering the variants more aggressively (particularly excluding indels; unpublished data), although at the expense of decreasing haplotype size and density. It is also possible to improve the consistency measures described above by using more sophisticated methods for haplotype separation. One possibility we have explored is to use the solutions described above as a starting point in a Markov chain Monte Carlo (MCMC) algorithm. This produces solutions for which the fraction of high LD conflicts with HapMap is reduced by ∼30%. This approach has other advantages as well: MCMC sampling provides a natural way to assess the confidence of a partial haplotype assignment. Assessment of this and other measures of confidence is a topic for future investigation. We used the generated haplotypes to view how well they span the current gene annotation. We were able to identify 84% (19,407 out of 23,224 protein coding genes) of Ensembl version 41 genes partially contained within a haplotype block and 58% of protein coding genes completely contained within a haplotype block. We note that in population-based haplotypes, denser sampling of SNPs in regions of low LD leads to reduction in the size of the average haplotype block [72]. In contrast to this finding, detection of additional true heterozygous variants through personal sequencing, regardless of LD, would lead to larger partial haplotypes, because additional variants increase the density of variants and thus their linkage to one another. The sequencing, assembly, and cataloguing of the variant set and the corresponding haplotypes of the HuRef donor provided unprecedented opportunity to study gene-based variation using the vast body of scientific literature and extensively curated databases like OMIM [73] and Human Genetic Mutation Database (HGMD, [18]). A preliminary assessment indicates that 857 OMIM genes have at least one heterozygous variant in the coding or UTR regions, and 314 OMIM genes have at least one nonsynonymous SNP (Figure 14A). Overall, we observed 11,718 heterozygous and 9,434 homozygous coding SNPs and 236 heterozygous and 627 homozygous coding indels (Figure 14B). In addition, 4,107 genes have 6,114 nonsynonymous SNPs indicating that at least 17% (4,107/23,224) of genes encode differential proteins. The nonsynonymous SNPs define a lower limit of a potentially impacted proteome, because 44% of genes (10,208/23,224) have at least one heterozygous variant in the UTR or coding region and these variants could also affect protein function or expression. Therefore, almost half of the genes could have differential states in this diploid human genome, and this estimate does not include variation in nonexonic regions involved in gene regulation such as promoters and enhancers. Understanding potential genotype-to-phenotype relationships will require many more extensive population-based studies. However, the complexities of assessing genotype–phenotype relationships begin to emerge even from a very preliminary glimpse of an individual human genome (Table 13). For Mendelian conditions such as Huntington disease (HD), the predictive nature of the genomic sequence is more definitive. Our data reveal the donor to be heterozygous (CAG) 18/ (CAG) 17 in the polymorphic trinucleotide repeat located in the HD gene (HD affected individuals have more than 29 CAG repeats) [74]. The genotype matches the phenotype in this case, since the donor does not have a family history of Huntington disease and shows no sign of disease symptoms, even though he is well past the average onset age. The HuRef donor' s predisposition status for multifactorial diseases is, as expected, more complicated. For example, the donor has a family history of cardiovascular disease prompting us to consider potentially associated alleles. The HuRef donor is heterozygous for variants in the KL gene; F352V (r9536314) and C370S (rs9527025). It has previously been observed that these heterozygous alleles present a lower risk for coronary artery disease [75]. However, the donor is also homozygous for the 5A/5A in rs3025058 in the promoter of the matrix metalloproteinase-3 (MMP3) [76]. This genotype is associated with higher intra-arterial levels of stromelysin and has a higher risk of acute myocardial infarction. This observation highlights the forthcoming challenge toward assessing the effects of the complex interactions in the multitude of genes that drive the development and progression of phenotypes. On occasion, these variant alleles may provide either protective or deleterious effects, and the ascertainment of resulting phenotypes are based on probabilities and would need to account for impinging environmental effects. In our preliminary analysis of the HuRef genome, we also identified some genetic changes related to known disease risks for the donor. For example, approximately 50% of the Caucasian population is heterozygous for the GSTM1 gene, where the null mutation can increase susceptibly to environmental toxins and carcinogens [77–79]. The HuRef assembly identifies the donor to be heterozygous for the GSTM1 gene. Currently, it is not possible without further testing (including somatic analysis) and comparison against larger datasets to determine if this variant contributes to the reported health status events experienced by the donor, such as skin cancer. We also found some novel changes in the HuRef genome for which the biological consequences are as yet unknown. For example, we found a 4-bp novel heterozygous deletion in Acyl-CoA Oxidase 2 (ACOX2) causing a protein truncation. ACOX2 encodes an enzyme activity found in peroxisomes and associates intimately with lipid metabolism and further was found to be absent from livers of patients with Zellweger syndrome [80]. The deletion identified would likely abolish peroxisome targeting, but the biological function of the mutation remains to be tested. We have also been able to detect inconsistencies between detected genotypes in the donor' s DNA and the expected phenotype based on the literature given the known phenotype of the HuRef donor. For example, the donor' s LCT genotype should confer adult lactose tolerance according to published literature [81], but this does not match with the self-reported phenotype of the donor' s lactose intolerance. Apparent inconsistencies of this nature may be explained by considering the modifying effect of other genes and their products, as well as environmental interactions.
We describe the sequencing, de novo assembly, and preliminary analysis of an individual diploid human genome. In the course of our study, we have developed an experimental framework that can serve as a model for the emerging field of en masse personalized genomics [82]. The components of our strategy involve: (i) sample consent and assessment, (ii) genome sequencing, (iii), genome assembly, (iv) comparative (one-to-one) mapping, (v) DNA variation detection and filtering, (vi) haplotype assembly, and (vii) annotation and interpretation of the data. We were able to construct a genome-wide representation of all DNA variants and haplotype blocks in the context of gene annotations and repeat structure identified in the HuRef donor. This provides a unique glimpse into the diploid genome of an individual human (Poster S1). The most significant technical challenge has been to develop an assembly process (points ii–v) that faithfully maintains the integrity of the allelic contribution from an underlying set of reads originating from a diploid DNA source. As far as we know, the approach we developed is unique and is central to the identification of the large number of indels less than 400 bp in length. We attempted de novo recruitment of sequence reads to the NCBI human reference genome, using mate pairing and clone insert size to guide the accurate placement of reads [83]. Although this approach can produce useful results, it does limit variant detection to completed regions of the reference genome and, like genome assembly, can be confounded by segmentally duplicated regions. The genome assembly approach with allelic separation allows the detection of heterozygous variants present in the individual genome with no further comparison. The one-to-one mapping of our HuRef assembly against a nearly completed reference genome permits the detection of the remaining variants. These variants arise from sequence differences found within and also outside the mapped regions, where the precision of the compared regions is being provided by the genome-to-genome comparison [59]. The ability to provide a highly confident set of DNA variants is challenging, because more than half of the variants are a single base in length but include both SNPs and indels. A filtering approach was used that accounts for the positional error profile in a Sanger sequenced electropherogram in relation to the called variant. Additional filtering considerations necessitated minimal requirements for read coverage and for the proportional representation of each allele. The filtering approaches were empirical and used the large amounts of previously described data on human variation (dbSNP). The utility of using paired-end random shotgun reads and the variant set defined on the reads via the assembly enabled the construction of long-range haplotypes. The haplotypes are remarkably well constructed given that the density of the variant map is comparable to those used in other studies [35], reflecting the utility of underlying sequence reads beyond just genome assembly. To understand how an individual genome translates into an individual transcriptome and ultimately a functional proteome, it is important to define the segregation of variants among each chromosomal copy. While several new approaches for DNA sequencing are available or being developed [84–86], we chose to use proven Sanger sequencing technology for this HuRef project. The choice was obviously motivated in part for historical reasons [7], but not solely. We attached a high importance to generating a de novo assembly including maximizing coverage and sensitivity for detecting variation. We further anticipated that long read lengths (in excess of 800 nucleotides), compatibility with paired-end shotgun clone sequencing, and well-developed parameters for assessing sequencing accuracy would be required. High sequence accuracy is essential to avoid calling large numbers of false-positive variants on a genome-wide scale. Long paired-end reads are especially useful for achieving the best possible assembly characteristics in whole-genome shotgun sequencing and for providing sufficient linkage of variants to determine large haplotypes. We have been able to categorize a significant amount of DNA variation in the genome of a single human. Of great interest is the fact that 44% of annotated genes have at least one, and often more, alterations within them. The vast majority—3,213,401 events (78%) of the 4. 1 million variants detected in the HuRef donor—are SNPs. However, the remaining 22% of non-SNP variants constitute the vast majority, about 9 Mb or 74%, of variant bases in the donor. Using microarray-based methods, we also detected another 62 copy number variable regions in HuRef, estimated to add some 10 Mb of additional heterogeneity. Given these potential sources of measured DNA variation, we can, for the first time, make a conservative estimate that a minimum of 0. 5% variation exists between two haploid genomes (all heterozygous bases, i. e. , SNP, multi-nucleotide polymorphisms [MNP], indels, [complex variants + putative alternate alleles + CNV]/genome size; [2,894,929 + 939,799 + 10,000,000]/2,809,547,336) namely those that make up the diploid DNA of the HuRef assembly. We also note that there will be significantly more DNA variation discovered in heterochromatic regions of the genome [87], which largely escaped our analysis in this study. We had mixed success when attempting to find support for the experimentally determined CNVs in the HuRef assembly itself or the data from which it was derived. More than 50% of the CNVs overlapped segmental duplications, and these regions are underrepresented in HuRef, which complicated the analysis. We attempted to map the sequence reads onto the NCBI human genome and then identify CNVs by detecting regions with significant changes in read depth. However, we found significant local fluctuations in read depth across the genome, limiting the ability for comparison and suggesting that a higher coverage of reads may be required to use this approach effectively. As we have emphasized throughout, a major difference of the genomic assembly we have described is our approach to maintain, wherever possible, the diploid nature of the genome. This is in contrast to both the NCBI and WGSA genomes, which are each consensus sequences and, therefore, a mosaic of haplotypes that do not accurately display the relationships of variants on either of the autosomal pairs. For BAC-based genome assemblies such as the NCBI genome assembly, the mosaic fragments are generally genomic clone size (e. g. , cosmid, PAC, BAC), with each clone providing contiguous sequence for only one of the two haplotypes at any given locus. Moreover, there are substantial differences in the clone composition of different chromosomes due to the historical and hierarchical mapping and sequencing strategies used to generate the NCBI reference assemblies [7,8]. In contrast, for WGSA, the reads that underlie most of the consensus sequence are derived from both haplotypes. This can result in very short-range mosaicism, where the consensus of clustered allelic differences does not actually exist in any of the underlying reads. To address this issue, the Celera assembler was modified to consider all variable bases within a given window and to group the sequence forms supporting each allele before incorporation into a consensus sequence (see Materials and Methods). In our experience, this reduces the incidence of local mosaicism, although, between windows, the consensus sequence remains a composite of haplotypes. Efforts to build haplotypes from the genome assembly (Haplotype Assembly) will likely lead to future modification of the assembler, allowing it to output longer consensus sequences for both haplotypes at many loci. Clearly, a single consensus sequence for a diploid genome, whether derived from BACs or WGS, has limitations for describing allelic variants (and specific combinations of variants) within the genome of an individual. Partial haplotypes can be inferred for an individual from laboratory genotype data (e. g. , from SNP microarrays) in conjunction with population data or genotypes of family members. However, at least in the absence of sets of related individuals (e. g. , family trios), it is difficult to determine haplotypes from genotype data across regions of low LD. We have shown that sequencing with a paired-end sequencing strategy can provide highly accurate haplotype reconstruction that does not share these limitations. The assembled haplotypes are substantially larger than the blocks of SNPs in strong LD within the various populations investigated by the HapMap project. In addition to being larger, haplotypes inferred in our approach can link variants even where LD in a population is weak, and they are not restricted to those variants that have been studied in large population samples (e. g. , HapMap variants). We note that in addition to the implications for human genetics, this approach could be applied to separating haplotypes of any organism of interest—without the requirement for a previous reference genome, family data, or population data—so long as polymorphism rates are high enough for an acceptable fraction of reads or mate pairs to link variants. There are several avenues for extending our inference of haplotypes. As noted, although the naive heuristics used here give highly useful results, other approaches may give even more accurate results, as we have observed with an MCMC algorithm. There are various natural measures of confidence that can be applied to the phasing of two or more variants, including the minimum number of clones that would have to be ignored to unlink two variants, or a measure of the degrees of separation between two variants. The analysis presented here provides phasing only for sites deemed heterozygous, but data from apparently homozygous sites can be phased as well, so we can tell with confidence whether a given site is truly homozygous (i. e. , the same allele is present in both haplotypes) or whether the allele at one or even both haplotypes cannot be determined, as occurs as much as 20% of the time with the current dataset. Lastly, it should be possible to combine our approach with typical genotype phasing approaches to infer even larger haplotypes. Our project developed over a 10-year period and the decisions regarding sample selection, techniques used, and methods of analysis were critical to the current and continued success of the project. We anticipated that beyond mere curiosity, there would be very pragmatic reasons to use a donor sample from a known consented individual. First and foremost, as we show in a preliminary analysis, genome-based correlations to phenotype can be performed. Due to the still rudimentary state of the genotype-phenotype databases it can be argued that at the present time, DNA sequence comparisons do not reveal much more information than a proper family history. Even when a disease, predisposition, or phenotypically-relevant allele is found, further familial sampling will usually be required to determine the relevance. Eventually, however, populations of genomes will be sequenced, and at some point, a critical mass will dramatically change the value of any individual initiative providing the potential for proactive rather than reactive personal health care. In a simple analogy, absent of family history, genealogical studies can now be quite accurate in reconstructing ancestral history based purely on marker-frequency comparisons to databases. Here, with a near-unlimited amount of variation data available from the HuRef assembly, we can reconstruct the chromosome Y ethno-geneographic lineage (Figure 15), which is not only consistent with, but better defines the self-reported family tree data (Figure 1A and unpublished data). There are always issues regarding the generation and study of genetic data and these may amplify as we move from what are now primarily gene-centric studies to the new era where genome sequences become a standard form of personal information. For example, there are often concerns that individuals should not be informed of their predisposition (or fate) if there is nothing they can do about it. It is possible, however, that many of the concerns for predictive medical information will fall by the wayside as more prevention strategies, treatment options, and indeed cures become realistic. Indeed we believe that as more individuals put their genomic profiles into the public realm, effective research will be facilitated, and strategies to mitigate the untoward effects of certain genes will emerge. The cycle, in fact, should become self-propelling, and reasons to know will soon outweigh reasons to remain uninformed. Ultimately, as more entire genome sequences and their associated personal characteristics become available, they will facilitate a new era of research into the basis of individuality. The opportunity for a better understanding of the complex interactions among genes, and between these genes and their host' s personal environment will be possible using these datasets composed of many genomes. Eventually, there may be true insight into the relationships between nature and nurture, and the individual will then benefit from the contributions of the community as a whole.
We used the assembled chromosome sequence of the human genome available as NCBI version 36. The gene annotation of this genome was provided by Ensembl (http: //www. ensembl. org) version 41, which incorporates dbSNP version 126. Haplotype map data was obtained from http: //www. hapmap. org, Release version 21a. Celera-generated chromatograms for the HuBB individual [7] were obtained from the NCBI trace archive. These included reads from two tissues sources: blood and sperm. Sequence reads were generated from these traces using Phred version 020425. c [88] and a modified version of Paracel TraceTuner (http: //sourceforge. net/projects/tracetuner/). This reprocessing significantly improved accuracy and quality in the 5′ portion of the reads, increasing their usable length by 7%, and reducing variants encoding spurious protein truncations, as well as reducing apparent heterozygous variants in the assembly. 200-μl aliquots of thawed, whole blood were processed using the MagAttract DNA Blood Mini M48 Kit and the MagAttract DNA Blood >200 μl Blood protocol on the BioRobot M48 Workstation running the GenoM-48 QIAsoft software (version 2. 0) (Qiagen; http: //www. qiagen. com). Tris: EDTA (10: 0. 1) was used for the final 200 μl elution step. A260/A280 readings (SPECTRAmax Plus spectrophotometer (Molecular Devices; http: //www. moleculardevices. com) or an ND-1000 spectrophotometer (NanoDrop Technologies; http: //www. nanodrop. com), and gel images were used to quantify the DNA and to confirm that high-quality, high–molecular weight DNA was available for downstream processing. 1. 0 μl of extracted DNA was run on a 0. 8% agarose gel containing ethidium bromide, for 4 h at 60 V and imaged using Gel Doc and Quantity One Software (Bio-Rad Laboratories; http: //www. bio-rad. com). Phytohemagglutin-stimulated lymphocytes from peripheral blood were cultured for 72 h with thymidine synchronization. G-banding analysis was performed on metaphase spreads from peripheral blood lymphocytes using standard cytogenetic techniques. Spectral karyotyping was performed on metaphase spreads from cultured lymphocytes. SkyPaint probes were used according to manufacturer' s instructions (Applied Spectral Imaging; http: //www. spectral-imaging. com). Metaphases were viewed with a Zeiss epifluorescence microscope and spectral images were acquired with an SD300 SpectraCube system and analyzed using SkyView software 1. 6. 2 (Applied Spectral Imaging). Plasmid and Fosmid Library Construction. We nebulized genomic DNA to produce random fragments with a distribution of approximately 1–25 kb, end-polished these with consecutive BAL31 nuclease and T4 DNA polymerase treatments, and size selected using gel electrophoresis on 1% low–melting-point agarose. After ligation to BstXI adapters, we purified DNA by three rounds of gel electrophoresis to remove excess adapters, inserted fragments into BstXI-linearized medium-copy pBR322 plasmid vectors, and inserted the resulting library into GC10 cells by electroporation. To ensure that plasmid libraries contained few clones without inserts and no clones with chimeric inserts, we used vectors (pHOS) that include several features: (i) the sequencing primer sites immediately flank the BstXI cloning site to avoid sequencing of vector DNA, (ii) there are no strong promoters oriented toward the cloning site, and (iii) BstXI sites for cloning facilitate a high frequency of single inserts and rare no-insert clones. Sequencing from both ends of cloned inserts produced pairs of linked sequences of ∼800 bp each. We constructed fosmid libraries with approximately 30 μg of DNA that was sheared using bead beating and repaired by filling with dNTPs. We used a pulsed-field electrophoresis system to select for 39–40 kb fragments, which we ligated to the blunt-ended pCC1FOS vector. Clone Picking and Inoculation. Libraries were propagated on large-format (16 × 16 cm) diffusion plates and colonies were picked for template preparation using a Q-bot or Q-Pix colony-picking robots (Genetix; http: //www. genetix. com) and inoculated into 384-well blocks. DNA Template Preparation. We prepared plasmid DNA using a robotic workstation custom built by Thermo CRS, based on the alkaline lysis miniprep [89], modified for high-throughput processing in 384-well plates. The typical yield of plasmid DNA from this method was approximately 600–800 ng per clone, providing sufficient DNA for at least four sequencing reactions per template. Sequencing Reactions. Sequencing protocols were based on the di-deoxy sequencing method [90]. Two 384-well cycle-sequencing reaction plates were prepared from each plate of plasmid template DNA for opposite-end, paired-sequence reads. Sequencing reactions were completed using Big Dye Terminator (BDT) chemistry version 3. 1 Cycle Sequencing Ready Reaction Kits (Applied Biosystems) and standard M13 forward and reverse primers. Reaction mixtures, thermal cycling profiles, and electrophoresis conditions were optimized to reduce volume and extend read lengths. Sequencing reactions were set-up by the Biomek FX (Beckman Coulter; http: //www. beckmancoulter. com) pipetting workstations. Templates were combined with 5-μl reaction mixes consisting of deoxy- and fluorescently labeled dideoxynucleotides, DNA polymerase, sequencing primers, and reaction buffer. Bar coding and tracking promoted error-free transfer. Amplified reaction products were transferred to a 3730xl DNA Analyzer (Applied Biosystems). The Celera Assembler Software (https: //sourceforge. net/projects/wgs-assembler/) [7,40,91] generated contiguous sequences (contigs) that could be linked via mate-pair information into scaffolds. It has a phase for splitting initial apparently chimeric contigs (referred to as unitigs), but this process is not repeated for the final set of contigs and scaffolds as with some other assemblers (Arachne 2 [92]). This leaves a small number of chimeric scaffolds, which can be detected and split as described below. All assemblers fail to discriminate alternate alleles in polymorphic regions from distinct regions of the genome. These polymorphic regions, containing highly repetitive sequence with short unique anchoring sequence and simple algorithmic failures, result in a number of small scaffolds that are highly redundant. Although there are valuable data in these small scaffolds, they are usually not treated as part of the assembled sequence. For this project we made specific modifications to the Celera Assembler to enable the grouping of reads into separate alleles when heterozygous variants were encountered. Instead of taking a column-by-column approach to determine the consensus sequence from a set of aligned reads, the region of variation was considered as a whole, defined as that between at least 11 bp nonvariant columns. In practice, variant regions would most frequently be single columns (SNPs), but the new algorithm only applied to longer regions. The reads spanning a variant region were split between alleles. An allele, for this purpose, was one or more spanning reads sharing an identical sequence for the variant region, and was considered confirmed if represented by two or more reads. Each allele was assigned a score equal to the sum of average quality values for the spanning portions of its reads. The highest-scoring confirmed allele was used for the consensus sequence. Alternate confirmed allele sequences were reported separately. As expected, there were usually two confirmed alleles in each region of sequence variation. Regions with more than two apparent confirmed alleles represented either collapsed repetitive sequence or a group of reads with systematic base calling error, rather than true genetic variation. The set of The Institute for Genome Research (TIGR) BAC ends [41] used in the WGSA [40] assembly were aligned to the 553 HuRef scaffolds of at least 100 kb in length. We kept BAC ends that mapped uniquely to a single scaffold and near the end of a scaffold, such that their mate was likely to reside outside of the scaffold. Mate pairs were kept if both BAC ends passed the above criterion, and these indicated a possible joining of two scaffolds in a certain orientation. There were 144 consistent scaffold joins with at least two supporting mate pairs and 98 with one supporting mate pair. Using these scaffold joins would result in 409 or 311 scaffolds, respectively, of at least 100 kb, with a concomitant increase in the scaffold N50 length. We used open-source software (http: //sourceforge. net/projects/kmer/) [40,93,94] to generate a one-to-one comparison between HuRef and NCBI human genome reference assembly. For sequences that do not contain very large, nearly identical duplications, this mapping is accurate [93]. Nearly identical duplicated regions tend to be underrepresented in whole-genome shotgun assemblies such as HuRef [10]. Segments that are duplicated in one sequence but not the other (for instance when failing to merge overlapping contigs) cannot be fully included in any one-to-one mapping. For example the first few megabases of NCBI version 36 Chromosomes X and Y are identical; therefore, a 1. 5-Mb scaffold from HuRef that maps to both of these regions is not part of the one-to-one mapping. Tandem repeats with variable unit copy number are also problematic for a one-to-one mapping. For each one-to-one mapping we determined three levels: matches, runs, and clumps. A match is a maximal high-identity local alignment, usually terminated by indels or sequence gaps in one of the assemblies. Runs may include indels, and are monotonically increasing or decreasing sets of matches (linear segments of a match dot plot) with no intervening matches from other runs on either axis. Clumps are similar to runs but allow small intervening matches/runs (such as small inversions) to be skipped over. The total number of base pairs in matches is a measurement of how much of the sequence is shared between assemblies. Within a run, the number of base pairs in each assembly is different, because indels are allowed among matches in the run. These could be gaps that are filled in one assembly but not the other, polymorphic insertions or deletions, or artifactual sequence. Runs span regions in both assemblies that have no rearrangements with respect to each other, providing a direct measure of the order and orientation differences between a pair of assemblies. Clumps provide a similar measure of rearrangement but allow for small differences that may be due to noise or polymorphic inversions. Remaining sequence may be unique to one assembly or the other, but some will also be large repetitive regions without good one-to-one mapping but present in some copy number in both assemblies. Apparently unique sequence may also represent some form of contaminant. We determined an initial set of potentially chimeric scaffolds by finding those that contained more than one clump of at least 5,000 bp relative to NCBI version 36. By mapping all HuRef and Coriell fosmid mate pairs to NCBI human reference genome and to HuRef, we assessed whether mate pair constraints were violated at the potentially chimeric junctions. Accordingly, we split 12 scaffolds. DNA variants were characterized by alignment of sequencing reads in the HuRef assembly and by comparison of regions of difference in the one-to-one HuRef to NCBI reference genome map. The contribution of each sequence read to a single position in the HuRef consensus was evaluated both during and after the assembly process to identify positions that contain more than one allele. This process identified heterozygous SNPs and indel polymorphisms, and typically two or more reads were required for the initial identification of an alternate allele. Homozygous SNPs and MNPs were identified when (respectively) single or multiple contiguous loci differed in the one-to-one mapping, and all underlying HuRef reads supported one allele. Finally, homozygous insertion or deletion loci were identified where the HuRef assembly had or lacked sequence relative to the NCBI assembly, respectively. These were commonly referred to as homozygous indels unless it was relevant for analysis purposes, computational or experimental, to refer to a homozygous insertion or deletion as a way of indicating presence or absence of the sequence, respectively, in the HuRef assembly. DNA variations were identified by examining the base changes within the HuRef assembly multialignment and between the HuRef assembly and the NCBI reference human genome. 5,061,599 SNPs and heterozygous variants were identified initially, after which filters were applied to eliminate erroneous calls. For a potential SNP, each read supporting that SNP was considered, and if the QV was <15 at the putative SNPs position in the read, then the read was considered invalid and was discarded as evidence for that particular variant. We also observed that deletions were overcalled at the beginnings and ends of reads, and insertions were overcalled at the ends of reads (Figure S2). By using the relative positions in the read where overcalling was detected, we were able to invalidate reads contributing to indel variant calls. We further observed that the relative read positions at which overcalling occurred was dependant on whether the read source was produced at Celera or The J. Craig Venter Institute (JCVI). Thus, any Celera read containing a putative deletion at a relative read position ≤0. 18 or ≥0. 76 was considered invalid for that particular deletion. Correspondingly, any JCVI read containing a putative deletion, at the relative read position ≤0. 07 or ≥0. 81 was deemed invalid in contributing to that particular variant call. Any Celera read was deemed invalid if it contained an insertion at a relative read position ≥0. 70, and any JCVI read with an insertion at relative read position ≥0. 77 was discarded as evidence. These thresholds were determined by plotting the frequency of insertions and deletions with respect to read position, and choosing the value where the call frequency was twice that of baseline (Figure S2). Subsequent to the quality value and read location filtering the remaining variants were inspected for the percentage, number, and directionality of reads supporting the alternate alleles. Additionally these variants were inspected for the total number of reads in their assembled locus and the repeat sequence status (transposon and tandem repeat). Transposon repeats were identified using the RepeatMasker program (http: //www. repeatmasker. org), and tandem repeats were identified using the Tandem RepeatFinder program [48]. The distribution of the percentage of reads containing the minor allele for heterozygous SNP and indels in Figure S3 shows that a large fraction of those putative variants that are found in dbSNP version 126 have a “minor allele frequency” (fraction of reads supporting the allele with fewer reads) of at least 20% and 25% for SNPs and indels, respectively. Therefore, we decided to apply the following filters separately to the QV and read location filtered variants, calculating at each filter step the fraction of passing variants that could be found in dbSNP. The filters applied to allow variants to be counted as bona-fide were: (i) 20% reads support minor allele for heterozygous SNP and 25% reads support minor allele for heterozygous indels, and (ii) two or more reads supporting the variant. The results of this analysis are presented in Table 3 and discussed in the Results section. Manual inspection showed that some neighboring variants identified within the one-to-one mapping of HuRef to the NCBI genome reference would be more precisely represented as one larger variant after realignment. To address these regions of clustered variants, we identified these problematic regions by clustering SNPs within 2 bp of each other or any non-SNP variants with 10 bp of another variant. For these variable regions, we recalled the variant (s) using the variant calling algorithm developed as part of the consensus sequence generation found in the Celera assembler. Homozygous insertion/deletions were filtered in the same manner as SNPs and heterozygous variants. All variants that were not confirmed by two or more reads were eliminated, as were those that did not fulfill minimal requirements of at least one spanning mate pair, and that the inserted sequence on the HuRef assembly or deleted sequence on the NCBI assembly not contain any ambiguous bases We estimate the population mutation parameter (θ) [43] as: where K is the number of variants identified, L is the number of base pairs, and n is the number of alleles. For indels, K is the number of indel events. In the case of a single diploid genome, n = 2, so a and b reduce to 1. Then θ = K/L, which is simply the number of heterozygous variants divided by the length sequenced. The standard deviation of θ reduces to θ: Thus, the 95% confidence interval for θ is [0, θ+2θ] or [0,3θ]. Two individuals of European ancestry were randomly selected from the SeattleSNPs data (http: //pga. gs. washington. edu/) [95]. For the first individual, we constructed a haploid representation (without phasing) by randomly choosing one allele at each variant position. This reconstructed sequence is analogous to the NCBI genome sequence that we used to call HuRef homozygous variants. For the second individual, all variant positions were examined and scored. If the second individual was heterozygous at a position, then the heterozygous count was incremented by one. If the second individual had a homozygous genotype that did not match the allele seen in the reconstructed sequence then the homozygous variant was incremented by one. The second individual is analogous to the HuRef assembly sequence, and this procedure mimics our variant-calling algorithm and our definitions for heterozygous and homozygous variants. One caveat is that the NCBI human genome sequence, while only being one sequence, represents multiple individuals, and thus possibly contains more rare alleles in its sequence. We developed a statistical model based on our assembly read coverage in the single diploid genome and on the filtering criteria used for calling high confidence variants. We assumed that chromosomes containing each of the two alleles are equally likely to be sampled and that allele loci are independent. At a given heterozygous locus, the probability of observing both alleles in at least x reads follows the binomial distribution with p = 0. 50 and n = depth of coverage, where x is defined by the filtering criteria. To calculate the false-negative rate genome wide, a Poisson distribution is also incorporated to estimate sequence depth at different loci, where λ is set to the genome sequence coverage (7. 5 for SNPs, 5. 5 for insertions, 4. 9 for deletions, after read filtering is taken into account). A number of heterozygous indels between 1 and 20 bp were manually selected for experimental validation by verifying trace quality in the region of the indel, read coverage depth, and repeat sequence status. In order to detect heterozygous indels from the HuRef assembly, we ran PCR-amplified genomic DNA on PAGE to look for homoduplex and heteroduplex bands. Large insertions and deletions were also recognized by this process. Primers were designed by centering the targeted indel to produce amplicons 150–250 bp in length with the melting temperatures of these amplicons ranging between 70 °C and 86 °C. PCR for polymorphism analysis was carried out in 10-μl volume reactions containing 30 ng of purified genomic DNA, 1× PCR buffer, 20 μM deoxynucleoside triphosphates, 2 mM MgCl2,8% glycerol, 0. 18 μM primers, and 0. 0375 U AmpliTaq Gold DNA polymerase. Post-amplification treatment of each sample involved digestion with shrimp alkaline phosphatase (0. 5 U) and exonuclease I (1. 76 U) for 45 min at 37 °C, 15 min at 50 °C, with heat inactivation for 15 min at 72 °C. PAGE was carried out at room temperature for 4 h at 650 V (constant) in a standard vertical gel measuring 1 mm thick, 20 cm wide, and 30 cm long (apparatus Model SG-400–20, CBS Scientific Company Inc, http: //www. cbssci. com). The native gel consisted of 10% acrylamide with the 40% acrylamide stock solution having an acrylamide/ N, N′-methylenebisacrylamide ratio of 29: 1. The running buffer consisted of 1× TBE. A loading dye consisting of 2× BlueJuice (Invitrogen) was added to each amplified sample and 5 μl was loaded per gel lane. After electrophoresis, the DNA bands were visualized by staining with a 1: 10,000 dilution of SYBR Gold (Invitrogen). Fifty-one apparent homozygous insertions in the HuRef assembly were selected based on assembly structure (appropriate read depth coverage and supporting mate pair evidence), their proximity to annotated genes, and their size. The insertion sequences were from 100 to 1,200 bp with few repeat sequences, and no detectable alignments to human (NCBI 36) or chimpanzee [22] genomes. We tested 93 Coriell DNA donors in addition to the HuRef DNA sample: 21 samples of European origin (CEU - NA06985, NA07056, NA11832, NA11839, NA11840, NA11881, NA11882, NA11992, NA11993, NA11994, NA11995, NA12057, NA12156, NA12239, NA12750, NA12751, NA12813, NA12814, NA12815, NA12891, NA12892), 12 Han Chinese samples (NA18524, NA18526, NA18537, NA18545, NA18552, NA18562, NA18566, NA18572, NA18577, NA18609, NA18621, NA18635), 11 Japanese (Tokyo) samples (NA18940, NA18942, NA18945, NA18949, NA18953, NA18961, NA18964, NA18967, NA18981, NA18994, NA18998), 22 samples of Hispanic origin (NA17438, NA17439, NA17440, NA17441, NA17442, NA17443, NA17444, NA17445, NA17446, NA17448, NA17449, NA17450, NA17451, NA17452, NA17453, NA17454, NA17456, NA17457, NA17458, NA17459, NA17460, NA17461,15 samples of African American origin (NA17101, NA17102, NA17103, NA17104, NA17105, NA17106, NA17107, NA17108, NA17109, NA17110, NA17111, NA17112, NA17113, NA17114, NA17115) and 12 samples of Yoruban origin (NA18502, NA18504, NA18855, NA18870, NA19137, NA19144, NA19153, NA19200, NA19201, NA19203, NA19223, NA19238). A 200-bp amplicon was designed for each insertion. By design, a homozygous insertion sequence yielded a single high–molecular weight band of (200 bp + the insertion size) on the agarose gel. Absence of the insertion would be detected as a single low molecular band of 200 bp alone and a heterozygous indel would be detected as presence of both bands. The amplicons were classified according to theoretical melting temperatures (Tm). Standard GC content and high GC content amplicons (82 °C < Tm < 87 °C) were processed separately in the laboratory using optimized high-throughput PCR protocols enabling all amplifications to be performed in 384-well plates in a volume of 10 μl. The standard GC content PCR protocol was composed of 3. 0 μl of 0. 4 μM mixed forward and reverse primers, 3. 0 μl of DNA (1. 67 ng/μl) and 0. 05 μl (0. 25 Us) of AmpliTaq Gold DNA polymerase (Applied Biosystems). The high-GC PCR protocol comprised 3. 0 μl of 1. 2 μM mixed forward and reverse primers, 3. 0 μl of DNA (10. 0 ng/μl), and 0. 075 μl (0. 375 U) of AmpliTaq Gold DNA polymerase (Applied Biosystems). PCR was set up using a Biomek FX (Beckman Coulter) pipetting robot and a Pixsys 4200 (Cartesian Technologies; http: //www. cartesiantech. com/) nanoliter dispenser. All PCR amplifications were performed on dual 384-well GeneAmp PCR System 9700 thermal cyclers (Applied Biosystems) under the following program: 96 °C for 5 min (1×); 94 °C for 30 s, 60 °C for 45 s, 72 °C for 45 s (40×); 72 °C for 10 min (1×); and a 10 °C final hold. 2. 0 μl of PCR product was combined with 5. 0 μl of diluted loading dye (Invitrogen) and run on a 2. 0% agarose gel, containing ethidium bromide. Gels were run for 45 min at 90 V and imaged using a Gel Doc and Quantity One Software (Bio-Rad Laboratories). Gel images were manually evaluated for the presence or absence of expected products. Segments of the human genome that were found exclusively in either HuRef or NCBI version 36 represent potential misassemblies or genuine variations. In order to distinguish between these possibilities, we attempted to confirm the existence of the largest one-to-one HuRef–NCBI indels in a collection of fosmid clones, derived from eight individuals (see Table S5 legend). Fosmid end reads were downloaded from the Trace Archive, and mapped to HuRef and NCBI human reference genome using Snapper (http: //sourceforge. net/projects/kmer/). To avoid short allelic variants of single loci, the HuRef assembly included only scaffolds that spanned at least 30 kb. The initial alignments required a unique best score with at least 90% nucleotide identity for at least 25% of the read length. Pairs of end read alignments were then filtered sequentially to retain only those that mapped to the same scaffold (HuRef genome) or chromosome (NCBI reference genome), in a tail-to-tail orientation, and within three standard deviations of the mean insert length. First, regions of the HuRef genome that failed to map to NCBI reference genome in the one-to-one mapping and were spanned to an average depth of 10x by fosmids that failed to map to the NCBI reference genome were identified as potentially novel segments. Their sequences were aligned to NCBI using ncbi-blastn (-W 100), and novelty was defined by the absence of nucleotide identity (≥98%) for lengths of ≥1 kb in spans of at least 35 kb. Second, the mapping coordinates of clones that mapped discordantly to either HuRef or NCBI were intersected with the 40 largest one-to-one HuRef-NCBI–derived indels to identify fosmid clones that support the existence of these indels in other human genomes. To define inserted DNA, we required one fosmid end to map within the insert exclusive to one assembly, the other to map within flanking sequence common to both assemblies, and inconsistent mapping to the genome assembly that lacked the insertion. Defining absence of inserted DNA required the fosmid mapping to span the putative insertion point in the assembly that lacked the insertion, and inconsistent mapping to the assembly that contains the insertion. Haplotypes of heterozygous variants were inferred using a greedy heuristic with iterative refinement of the initial solution. Data Encoding. An SNP matrix (rows = reads or mate pairs, columns = variants) was constructed as follows: for each variant location, reads whose sequence matched the consensus sequence were assigned state “0, ” while reads not matching the consensus were assigned state “1. ” A pair of mated reads was merged into a single row only if they were in the same scaffold, with the expected orientation and separated by the expected distance (± 3 SD). Thus, a row in the matrix correspond to one of the following: (i) a pair of mated reads with consistent placements and (ii) a single unmated read or single mated read whose mate is not consistently placed. Initial Haplotype Construction. Initial partial haplotypes were constructed by repeating the following sequence of steps until all rows were assigned. From the remaining set of unassigned rows (initially all), choose the row with fewest missing elements. Use this row to seed a partial haplotype pair (i. e. , assign the row to one haplotype, which is initialized with the non-missing states from this row, and initialize the other haplotype with the complementary states). Until no more rows share non-missing information, identify the row that has the strongest signal (i. e. , number of columns indicating one haplotype minus number of columns indicating the other haplotype is maximal), and assign that row to the indicated haplotype, extending the haplotypes to include any additional columns that are non-missing for that row. When no unassigned rows overlap the current haplotypes, consider this pair of partial haplotypes final and go back to the beginning. Iterative Haplotype Refinement. When all rows have been assigned to partial haplotypes, each haplotype pair and the rows it includes can be refined iteratively, repeating the following two steps until no changes result. First, for each column (variant position) in the haplotypes, determine by majority rule the state assignment of each haplotype. Second, for each row (read or mate pair), determine the haplotype assignment by majority rule. Measurement of Haplotype Sizes. For each pair of partial haplotypes, two measures of size are natural: the number of variants that are phased and the distance in bp from the first to the last variant. In addition to the average of such values, the N50 statistic indicates a haplotype size that encompasses at least half of the variants. Comparison of Phasing to HapMap. Consistency of HuRef haplotypes with HapMap haplotypes was assessed as follows. Within each partial HuRef haplotype, variants that were present in Phase I HapMap data were identified (henceforth “HapMap variants”). For each pair of HapMap variants that were adjacent in a HuRef haplotype, two measures were determined. The first was the degree of LD between the paired variants from the HapMap CEU panel. The second was the conditional probability of observing the HuRef haplotype in the CEU panel given the observed genotypes. When r2 ≥ 0. 9 and the conditional probability was <0. 5, this was considered a clear conflict of HuRef and HapMap haplotypes. The HuRef sample was genotyped in duplicate on each of the GeneChip Human (500K) Mapping NspI and StyI Array Sets (Affymetrix; http: //www. affymetrix. com), according to the manufacturer' s instructions and as described previously [96]. Each array contains an average of 250,000 SNP markers. The arrays were scanned using the Gene Chip Scanner 3000 7G and Gene Chip Operating System. The call rate was >96% for all four all hybridizations; 0. 1% discordant genotype calls between the technical replicates were excluded from further analysis. The NspI and StyI array scans were analyzed for copy number variation using a combination of DNA Chip Analyzer (dChip) [97], Copy Number Analysis for GeneChip (CNAG) [98], and Genotyping Microarray-based CNV Analysis (GEMCA) [99]. Analysis with dChip (http: //www. dchip. org) was performed using a Hidden Markov Model (HMM) as previously described [100], and a set of 50 samples run in the same facility were used as reference. For analyses with CNAG version 2. 0 (http: //www. genome. umin. jp), the copy number changes were inferred using a HMM built into CNAG [98]. GEMCA analysis was performed essentially as described [99], except that we used one designated DNA samples (NA10851) as reference for pair-wise comparison. This sample has been screened for CNVs in a previous study [62] and the CNVs known to be present in the reference genome were excluded. The HuRef sample was genotyped using the Sentrix HumanHap650Y Genotyping BeadChip according to the manufacturer' s instructions. All chips were scanned using the Sentrix Bead-Array reader and the Sentrix Beadscan software application. The results from the BeadChip were analyzed for CNV content using QuantiSNP as previously described [101]. The Agilent human genome CGH array contains 244,000 60mer probes on a single slide. The experiment was run using 2. 5 μg of genomic DNA for Cy3/Cy5 labeling for each hybridization, with a standard dye-swap experimental design. DNA sample NA10851 was used as a reference. The slides were scanned at 5-μm resolution using the Agilent G2565 Microarray Scanner System (Agilent Technologies; http: //www. agilent. com). Feature extraction was performed using Feature Extraction v9. 1 and results were analyzed using CGH Analytics v3. 4. 27. CGH was performed using the Nimblegen human genome CGH array. The array contains 385,000 isothermal probes yielding a median spacing of 6 kb across the human genome. The experiment was performed as previously described [102] with a standard dye-swap experimental design. Results were analyzed using the CNVfinder algorithm [103]. One of the dye-swap experiments did not meet the quality control cut-offs, and because of this, the Nimblegen CNV calls were only employed for confirmation of CNV identified by the other platforms, and not used for identification of additional CNVs FISH analysis was performed to find the location of DNA segments present in the HuRef DNA but either missing or represented by gaps in HuRef assembly. The FISH analysis was performed as previously described [104]. Initially, fosmids representing 107 different regions were chosen and end-sequenced to confirm that they mapped to the intended scaffolds. After excluding fosmids for which the original mapping was erroneous or uncertain, 88 fosmids remained. The entire sequence for each fosmid was then computationally excised from the scaffolds sequence and analyzed for repeat content using RepeatMasker. Fosmids with more than 6 kb (∼17%) satellite repeat content were excluded from further analysis. All fosmids that passed these filtering criteria were analyzed on metaphase spreads from two different cell lines (GM10851 and GM15510) to determine the chromosomal location of the fosmid probe. At least 10 metaphases were scored for each probe, all in duplicate by two experienced cytogeneticists.
The GenBank (http: //www. ncbi. nlm. nih. gov/Genbank) accession number for sequences discussed in the paper are: AADD00000000 (WGSA) and ABBA01000000 (the consensus sequences of both HuRef scaffolds and chromosomes). | We have generated an independently assembled diploid human genomic DNA sequence from both chromosomes of a single individual (J. Craig Venter). Our approach, based on whole-genome shotgun sequencing and using enhanced genome assembly strategies and software, generated an assembled genome over half of which is represented in large diploid segments (>200 kilobases), enabling study of the diploid genome. Comparison with previous reference human genome sequences, which were composites comprising multiple humans, revealed that the majority of genomic alterations are the well-studied class of variants based on single nucleotides (SNPs). However, the results also reveal that lesser-studied genomic variants, insertions and deletions, while comprising a minority (22%) of genomic variation events, actually account for almost 74% of variant nucleotides. Inclusion of insertion and deletion genetic variation into our estimates of interchromosomal difference reveals that only 99. 5% similarity exists between the two chromosomal copies of an individual and that genetic variation between two individuals is as much as five times higher than previously estimated. The existence of a well-characterized diploid human genome sequence provides a starting point for future individual genome comparisons and enables the emerging era of individualized genomic information. | Abstract
Introduction
Results
Discussion
Materials and Methods
Supporting Information | primates
mammals
human
genetics and genomics | 2007 | The Diploid Genome Sequence of an Individual Human | 26,003 | 284 |
The advent of induced pluripotent stem cells (iPSCs) revolutionized human genetics by allowing us to generate pluripotent cells from easily accessible somatic tissues. This technology can have immense implications for regenerative medicine, but iPSCs also represent a paradigm shift in the study of complex human phenotypes, including gene regulation and disease. Yet, an unresolved caveat of the iPSC model system is the extent to which reprogrammed iPSCs retain residual phenotypes from their precursor somatic cells. To directly address this issue, we used an effective study design to compare regulatory phenotypes between iPSCs derived from two types of commonly used somatic precursor cells. We find a remarkably small number of differences in DNA methylation and gene expression levels between iPSCs derived from different somatic precursors. Instead, we demonstrate genetic variation is associated with the majority of identifiable variation in DNA methylation and gene expression levels. We show that the cell type of origin only minimally affects gene expression levels and DNA methylation in iPSCs, and that genetic variation is the main driver of regulatory differences between iPSCs of different donors. Our findings suggest that studies using iPSCs should focus on additional individuals rather than clones from the same individual.
Research on human subjects is limited by the availability of samples. Practical and ethical considerations dictate that functional molecular studies in humans can generally only make use of frozen post mortem tissues, a small collection of available cell lines, or easily accessible primary cell types (such as blood or skin cells). The discovery that human somatic cells can be reprogrammed into a pluripotent state [1–3] and then be differentiated [4] into multiple somatic lineages, has the potential to profoundly change human research by providing access to a wide range of cell types from practically any donor individual. Though much progress has been made since the initial development of iPSC reprogramming technology, and human iPSCs have been used in a wide range of studies [5–8], the usefulness of iPSCs as a model system for the study of human phenotypes is still extensively debated [9–11]. The principal issue is the extent to which reprogrammed iPSCs retain epigenetic and gene expression signatures of their cell type of origin. A residual epigenetic signature of the original precursor cell in the reprogrammed iPSCs is often referred to as ‘epigenetic memory’ [12]. The common view, established by a few early studies in mice and humans, is that epigenetic memory is a significant problem in iPSCs [10,12–18]. In mice, methylation profiles in iPSCs and in the precursor somatic cells from which the iPSCs were generated were found to be more similar than expected by chance alone [12,14]. The extent of this similarity, however, could not be benchmarked against genetic diversity because the somatic cells and the iPSCs were all from genetically identical mice. In turn, methylation profiles in human iPSCs reprogrammed from different somatic cell types were found to be quite distinct from each other [15,16]. However, the somatic cells were provided by different donor individuals, hence epigenetic memory and differences due to genetic diversity were confounded. Additionally, concerns were initially raised about residual epigenetic memory in iPSCs by studies that considered iPSCs generated using retroviral vectors [12,14–16]. Retroviral reprogramming is characterized by random integrations that vary in copy number and genomic location across lines. Furthermore, it has been shown that viral vectors commonly utilized in iPSC generation preferentially integrate into active gene bodies, strong enhancers or active promoters [19,20], this process of preferential integration into open chromatin would likely lead to a strong cell type of origin signature. In contrast to retroviral reprograming, the more recent episomal approaches to establish iPSCs are associated with much lower rates of genomic integration [21,22]. Indeed, one recent study has concluded that when properly controlling for genetic variation and using integration free methodology to establish iPSCs, the effect of cell type of origin on gene expression in iPSCs is low compared to inter-individual genetic contributions [23]. However, this study did not consider matched epigenetic markers, the supposed drivers of the suspected phenomenon of residual cell type of origin memory in reprogrammed iPSCs. We thus designed a study to directly and effectively address this issue. We focused on two cell types that are the source for the majority of human iPSCs to date, and the most easily collected tissue samples from humans: skin fibroblasts, and blood cells. Specifically, we collected skin biopsies and blood samples from four healthy Caucasian individuals (two males and two females). Dermal fibroblasts were isolated from dissociated skin biopsies and maintained in culture until reprogramming. We isolated the buffy coat from whole blood and subsequently used Epstein–Barr virus to transform B cells into immortalized lymphoblastoid cell lines (LCLs), one of the most common cell types used in genomic studies.
Once the quality of the iPSCs was confirmed, we extracted RNA and DNA from LCLs, fibroblasts, LCL derived iPSCs (L-iPSCs), and fibroblast derived iPSCs (F-iPSCs) from all four individuals (S1 Table). We then used the Illumina Infinium HumanMethylation450 array and the Illumina HumanHT12v4 array to measure DNA methylation and gene expression levels, respectively. Our data processing approach is described in detail in the methods. Briefly, considering the methylation data, we first excluded data from loci that were not detected either as methylated or unmethylated (no signal; detection P > 0. 01) in more than 25% of samples. We then applied a standard background correction [24] and normalized the methylation data using SWAN [25] (S5 Fig), which accounts for the two different probe types in the platform. Finally, we performed quantile normalization (S6A and S6B Fig). Following these steps we retained methylation data from 455,910 CpGs. Considering the expression data, we first excluded probes whose genomic mapping coordinates overlapped a known common SNP. We then retained all genes that were detected as expressed in any cell type in at least three individuals (S7 Fig). We then quantile normalized the gene expression data (S6C and S6D Fig). Following these steps we retained expression data for 11,054 genes. To examine overall patterns in the data, we initially performed unsupervised clustering based on Euclidean distance. As expected, using gene expression or methylation data, samples clustered based on cell type (LCLs, fibroblasts, and iPSCs) without exception. Interestingly, using the methylation data, iPSCs clustered perfectly by individual, not cell type of origin (Fig 2A). Within individual, however, data from L-iPSCs are more similar to each other than to data from F-iPSC in three of the four individual clusters. These results are consistent with a small proportion of the regulatory variation being driven by cell type of origin. The clustering pattern is less clear when we consider the gene expression data, although the iPSCs again tend to cluster by individual more than they do by cell type of origin (Fig 2B). The property of imperfect clustering of iPSC gene expression data by individual is consistent with previous observations by Rouhani and Kumasaka et al. [23]. We believe that a possible explanation for this observation is that overall regulatory variation between iPSCs–even across individuals–is small. Given the large number of sites interrogated (particularly on the methylation array), we also examined the clustering of iPSCs using only the top 1,000 most variable measurements across lines, similar to the approach of Kim et al. 2011 [16]. Our clustering remained largely unchanged using this subset of variable sites for both methylation data (S8A Fig) and expression data (S8B Fig). Clustering based on pairwise Pearson correlations rather than Euclidian distance produced nearly identical results (S8C–S8F Fig). We also examined patterns in the data using principal components analysis (PCA; S9 Fig) The results from the PCA are not as easily interpretable as those from the clustering analysis, but it is clear that the major components of variation are not driven by cell type of origin. We next considered methylation and expression patterns at individual loci and genes, respectively. We first focused on differences in CpG methylation between the cell types. Using limma [26] (see methods), we identified 190,356 differentially methylated (DM) CpG loci between LCLs and fibroblasts (FDR of 5%). Similarly, we identified 310,660 DM CpGs between LCLs and L-iPSCs and 226,199 DM loci between fibroblasts and F-iPSCs (Fig 3A). In contrast, at the same FDR, we only classified 197 CpG loci (0. 04% of the total sites tested; S10 Fig) as DM between L-iPSCs and F-iPSCs (S2A–S2D Table). Moreover, the 197 DM loci were not all independent; they clustered into 53 genomic regions, 37 of which are located near or within annotated genes. Of these 37 genes, 24 had measurable gene expression data (Fig 3C). The observation of small number of significant DMs associated with cell type of origin does not preclude a persistent but small difference between the epigenetic landscapes of L-iPSCs and F-iPSCs. We therefore asked, for each CpG classified as DM between LCLs and fibroblasts, whether the sign of the mean methylation difference between L-iPSCs and F-iPSCs is the same as the sign of the mean difference between the cell types of origin. We found a slight but significant enrichment of a consistent sign (50. 5% of the loci; binomial test; P < 10−6) in these two contrasts. This observation confirms that while epigenetic memory in iPSCs can be detected, the magnitude of such effect is small. Of the 197 DM loci between L-iPSCs and F-iPSCs, 133 loci were also DM between LCLs and fibroblasts (a highly significant overlap; χ2 test; P < 10−15). Moreover, 122 of these 133 DM loci showed a difference in methylation between LCLs and fibroblasts that was in the same direction as the one seen between L-iPSCs and F-iPSCs (sign test; P < 10−15). In principle, these observations support the idea of epigenetic memory, namely that a subset of epigenetic differences between the somatic cells persists in the reprogrammed iPSCs. Yet our results indicate that epigenetic memory persists in a remarkably small number of loci. We turned our attention to the gene expression data. We again used limma to identify (at an FDR of 5%) 7,281 differentially expressed (DE) genes between LCLs and fibroblasts, 8,008 DE genes between LCLs and L-iPSCs, and 7,420 DE genes between fibroblasts and F-iPSCs (Fig 3B). In contrast, at the same FDR, we classified only a single gene (TSTD1) as DE between L-iPSCs and F-iPSCs. These results are consistent with recent observations [23]. More generally, we found nearly no evidence for departure from a null model of no differences in gene expression levels between L-iPSCs and F-iPSCs (Fig 4, S11 Fig; S3A–S3D Table). We proceeded by performing a sign test, considering the sign of the mean gene expression difference between L-iPSCs and F-iPSCs in genes that were classified as DE between LCLs and iPSCs. We found fewer consistent signs than expected by chance alone (47. 8%; binomial test: P = 10−4). The single DE gene between L-iPSCs and F-iPSCs, TSTD1 (P = 6. 28 x 10−7; FDR 0. 69%), is also DE between the LCLs and fibroblasts precursor cells. Moreover, 11 of 19 CpG sites that are located near the TSTD1 gene, and are assayed by the methylation array, are among the 197 DM loci between L-iPSCs and F-iPSCs. We observed a decreased fold change of TSTD1 expression when comparing between LCLs and fibroblasts (log2 fold change of 2. 06) and L-iPSCs and F-iPSCs (log2 fold change of 1. 34). This may be a case of epigenetic memory that maintains a gene expression residual difference, but it appears to be the only such case in our data. We found no evidence that any of the other DM loci are associated with gene expression differences between L-iPSCs and F-iPSCs (Fig 3C). This is true even when we conservatively accounted for multiple tests by only considering the number of tests that involved genes that are associated with DM loci between L-iPSCs and F-iPSCs (S11 Fig). Our observations indicate that remarkably little residual memory of the precursor somatic cell affects gene expression and methylation patterns in the reprogrammed iPSCs. To formally evaluate this we estimated the contribution of inter-individual differences and cell type of origin effects on variation in methylation and gene expression levels (see methods). The mean proportion of variance explained by donor individual is 16. 2% and 15. 5%, for the methylation and expression data, respectively; while the mean proportion of variance explained by cell type of origin is 6. 6% and 6. 7%, respectively (T-test; P < 10−15; KS test P < 10−15; Fig 5). Interestingly, when we focus on gene and CpGs whose expression and methylation levels in LCLs were previously associated with genetic variation (eQTLs and meQTLs, respectively), the mean proportion of variance explained by donor individual is significantly higher (21. 2% and 19. 9%, for the methylation and expression data, respectively; T-test P < 10−15; KS test P < 10−15; S13 Fig), while the mean proportion of variation explained by cell type of origin is roughly similar (6. 28% and 6. 34% for methylation and expression data, respectively).
To date, the common view is that iPSCs derived from somatic cells retain robust epigenetic traces of the precursor cells [10,12–17,24]. Yet, in our data, a remarkably small amount of the observed regulatory variation in iPSCs is driven by cell type of origin. Our observations are consistent with genetic background being a major driver of regulatory variation in iPSCs. While our results challenge the common view that epigenetic memory is prevalent in iPSCs, a careful examination of the literature suggests that our data are in fact consistent with previous studies, though our interpretation is not. The principal difference between previous studies and ours is that we were able to benchmark epigenetic memory against other sources of variation. Previous studies either characterized iPSCs from a single individual [12,14], or were not able to distinguish between genetic and cell type of origin effects [15,16]. For example, though Kim et al. [16] reported a similar number of DM loci (137–370) between iPSCs derived from different cell types as we observed in our study, Kim et al. interpreted their observation as evidence for a marked effect of the donor cells. Yet, our observation that DNA methylation is quite homogenous across all iPSCs (both within replicates and between L-iPSCs and F-iPSCs; S8C and S8D Fig), is not in disagreement with the observations of Kim et al. Indeed, our study explicitly models the contribution of genetic background to variation in DNA methylation levels in iPSCs. When we consider DNA methylation in the context of variation explained by inter-individual differences, we find a remarkably small effect associated with cell type of origin. Moreover, even unsupervised clustering (based on either DNA methylation or gene expression data) indicated that samples largely clustered by individual. We found little evidence of clustering by cell type of origin. When we turned our attention to individual loci, only 197 (0. 043%) tested CpGs were classified as DM between L-iPSCs and F-iPSCs, compared with 190,356 (41. 7%) loci that were classified as DM between LCLs and fibroblasts. Our observation that only a handful of DM sites may drive regulatory differences between iPSCs from different origins is consistent with recent work by Rouhani and Kumasaka et al. [23] where a similar study design was employed examining only gene expression levels. Indeed, as in Kumasaka et al. , we found that individual genetic background captures a much larger proportion of gene regulatory variation than cell type of origin using both the DNA methylation and gene expression data. Future work needs to address additional pertinent questions. First, our study was limited to methylation and gene expression levels in iPSCs. Future studies should focus on additional epigenetic and regulatory markers. Second, we focused on regulatory differences between iPSCs, but did not study differentiated cell types. This needs to be addressed in the future because the degree to which iPSCs retain regulatory signatures of their cell type of origin ultimately is expected to influence the extent to which iPSCs can be used as a model system for studying complex traits in differentiated cell types. In conclusion, our study demonstrated that when accounting for individual, the impact of cell type of origin on DNA methylation and gene expression in iPSCs is limited to a small number of CpGs, which cluster into an even smaller number of genomic loci, and a single gene, with almost no detectable influence genome-wide. Our observations further confirm the usefulness of iPSCs for genetic studies regardless of the original somatic cell type. The high correlation of DNA methylation and gene expression levels (S8C and S8D Fig) between individuals, demonstrate the faithfulness of the model, though as we pointed out–similar studies in differentiate cells are required to generalize these conclusions. While cell type of origin should continue to be carefully documented, our data also suggest that future studies should focus on collecting more individuals rather than establishing multiple iPSC clones from the same individual.
Skin punch biopsies and blood were collected from the same individual within 20 minutes under University of Chicago IRB protocol 11–0524 (samples from four individuals were collected over three collection dates; samples from individuals 3 and 4 were collected on the same date). Skin and blood samples from an individual were processed at the same time (S1 Table). Fibroblast isolation and culture was conducted using the approach described in detail in Gallego Romero et al [27]. Briefly, skin punch biopsies (3mm) were digested using 0. 5% collagenase B (Roche), isolated fibroblasts were cultured in DMEM (Life Technologies) supplemented with 10% fetal bovine serum (FBS; JR Scientific), 0. 1mM NEAA, 2mM GlutaMAX (both from Life Technologies), 1% penicillin/streptomycin (Fisher), 64mg/L L-ascorbic acid 2-phosphate sesquimagnesium salt hydrate (Santa Cruz Biotechnology), at 5% CO2 and 5% O2. All other cell culture was performed at 5% CO2 and atmospheric O2. For LCL generation, whole blood was drawn (within 20 minutes of obtaining skin punch biopsies) into two 8. 5mL glass yellow top tubes (Acid Citrate Dextrose Solution A tubes; BD). Blood tubes were stored at room temperature and processed within 12 hours of collection. To isolate lymphocytes, we diluted whole blood with an equal amount of RPMI 1640 (Corning), diluted blood was slowly layered onto Ficoll-Paque (GE Lifescience) in 50 mL centrifuge tubes. This gradient was centrifuged at 1700 rpm for 30 minutes without acceleration or braking. Leukocytes and platelets formed a white band at the interface between the blood plasma and the Ficoll (called the buffy coat). We collected the buffy coat using a Pastette and to that added 10mL of PBS. The collected buffy coat was then washed three times with PBS. For EBV transformation, 4 x 106 fresh lymphocytes collected as described above were resuspended in a total of 4. 5 ml of RPMI 1640 culture medium (Corning) containing 20% FBS and 1: 100 phytohemagglutinin (PHA-M; LifeTechnologies) and transferred to a T-25 flask. EBV supernatant produced by the B95-8 cell lines (provided by the Ober lab) was added at 1: 10 to the culture flask. Cells were left undisturbed for three to five days before adding fresh media. Flasks were subsequently examined weekly for changes in cell growth as indicated by acidic pH (yellow color) and the appearance of clumps of cells growing in suspension. Once growth was established (21–35 days), cells were diluted or split to several flasks. When the cell density reached 8 x 105 to 1 x 106 cells per mL they were cryopreserved at a density of 10 x 106 cells per ml of freezing media in cryovials. All LCLs using this study were transformed with the same lot of EBV supernatant. To establish iPSCs we transfected LCLs (Amaxa Nucleofector Technology; Lonza) and fibroblasts (Neon Transfection System; Life Technologies) with oriP/EBNA1 PCXLE based episomal plasmids that containing the genes OCT3/4, SOX2, KLF4, L-MYC, LIN28, and an shRNA against p53 [21]. We supplemented these plasmids with an in vitro-transcribed EBNA1 mRNA transcript to promote exogenous vector retention following electroporation of the episomal vector [28,29]. Fibroblasts from all individuals were reprogrammed in two batches (see details in S1 Table). LCLs were reprogrammed in four batches (S1 Table). The first three batches contained LCLs from all four individuals. Individual 4 failed reprogramming in batches one and three. A final fourth batch was therefore done with only individual 4 (replicates A and C; S1 Table). We plated a range of 10,000–40,000 transfected cells per well in a 6-well plate. Within 21 days colonies were visible and manually passaged onto a fresh plate of irradiated CF1 mouse embryonic fibroblasts (MEF). We passaged these new iPSC colonies on MEF in hESC media (DMEM/F12 (Corning) supplemented with 20% KOSR (LifeTechnologies), 0. 1mM NEAA, 2mM GlutaMAX, 1% Pen/Strep, 0. 1% 2-Mercaptoethanol (LifeTechnologies) ). Fibroblast derived iPSCs were supplemented with 100ng/mL human basic fibroblast growth factor, versus 25ng/mL for LCL derived iPSCs; all other culture conditions were identical. After 10 passages of growth we transitioned the cultures to feeder-free conditions and cultured them for an additional three passages before collecting cell pellets for analysis. Feeder-free cultures were grown using 0. 01mg/cm2 (1: 100) hESC-grade Matrigel (BD Sciences) and Essential 8 media (LifeTechnologies). Passaging was done using DPBS supplemented with 0. 5mM EDTA. All RNA and DNA were isolated using Zymo dual extraction kits (Zymo Research) with a DNase treatment during RNA extraction (Qiagen). All iPSC lines were characterized as described previously [27]. Briefly, we initially confirmed pluripotency using PluriTest [30], a classifier that assigns samples a pluripotency score and novelty score based on genome-wide gene expression data. All samples were classified as pluripotent and had a low novelty score (S1 Fig). We next performed qPCR using 1 μg of total RNA, converted to cDNA, from all samples to confirm the endogenous expression of pluripotency genes: OCT3/4, NANOG, and SOX2 (S2A–S2C Fig). Additionally, we tested for the presence and expression of the EBV gene EBNA-1 using PCR (primers and cycling conditions can be found in S5 Table) (S2D and S3 Figs). We tested all samples for both genomic integrations and vector-based EBV. We did this using primers designed to amplify the EBNA-1 segment found in both the episomal vectors and the EBV used to transform LCLs. If the cell was positive (a single positive case was found: Ind4 F-iPSC), we further tested the origin of the EBV (genomic or episomal) using primers specific to the LMP-2A gene found in EBV or part of the sequence specific to the episomal plasmid (S3 Fig). Finally, we confirmed the ability of all iPSC lines to differentiate into the three main germ layers using the embryoid body (EB) assay. The EBs were imaged for the presence of all three germ layers (S4 Fig). It should also be noted that gene expression and DNA methylation levels are extremely similar between iPSC lines. This relative homogeneity further demonstrates the quality of our iPSC lines. In summary, all iPSC lines established in this study showed expression of pluripotent genes quantified by qPCR, generated EBs for all three germ layers, and were classified as pluripotent based on PluriTest. Extracted DNA was bisulphite-converted and hybridized to the Infinium HumanMethylation450 BeadChip (Illumina) at the University of Chicago Functional Genomics facility. To validate the array probe specificity, probe sequences were mapped to an in silico bisulfite-converted genome using the Bismark aligner [31]. Only probes that mapped uniquely to the human genome were retained (n = 459,221). We further removed data from probes associated with low signal (detection P-value > 0. 01) in more than 25% of samples (retained data from n = 455,910 loci). Raw output from the array (IDAT files) were processed using the minfi package [24] in R. We performed standard background correction as suggested by Illumina [24], and corrected for the different distribution of the two probe types on the array using SWAN [25] (S5 Fig). Additionally, we quantile normalized the red and green color channels (corresponding to methylated and unmethylated signal respectively) separately (S6A and S6B Fig). To calculate methylation levels (reported as β-values) we divided the methylated signal by the total signal from both channels. β-values were considered estimates of the fraction of alleles methylated at that particular locus in the entire cell population. RNA quality was confirmed by quantifying sample’s RNA Integrity Number (RIN) on an Agilent 2100 Bioanalyzer (Agilent Technologies). All samples had a RIN of 10. The extracted RNA from all samples was hybridized to the Illumina HT12v4 Expression BeadChip array (Illumina) at the University of Chicago Functional Genomics facility. Sample processing was performed using the lumi package in R [32]. We excluded data from a subset of probes prior to our analysis: First, we mapped the probe sequences to the human genome hg19 and kept only those with a quality score of 37, indicative of unambiguous mapping (n = 40,198; note that we also explicitly pre-filtered the 5,587 probes which were annotated as spanning exon-exon junctions to avoid mapping errors). Second, we downloaded the HapMap CEU SNPs (http: //hapmap. ncbi. nlm. nih. gov/downloads/genotypes/2010-08_phaseII+III/forward/) and converted their coordinates from hg18 to hg19 using the UCSC liftOver utility [33]. We retained only those probes that did not overlap any SNP with a minor allele frequency greater than 5% (n = 34,508). Third, we converted the Illumina probe IDs to Ensembl gene IDs using the R/Bioconductor package biomaRt [34] and retained only those probes that are associated with exactly one Ensembl gene ID (Ensembl 75—Feb 2014; n = 22,032). The full pipeline was implemented using the Python package Snakemake [35]. We defined a gene as expressed in a given sample if at least one probe mapping to it had a detection P-value < 0. 05. In the case of L-iPSCs, we defined a gene as expressed in an individual if any associated probes had a detection P-value < 0. 05 in at least one biological replicate. Using these criteria, we identified all genes expressed in at least three individuals in at least one cell type (S7 Fig; n = 14,111 probes associated with 11,054 annotated genes). In the case that multiple expressed probes were associated with the same ENSEMBL gene (n = 3,057), we only retained data from the 3' -most detected probe. Following these filtration steps, we obtained estimates of expression levels in all samples across 11,054 genes. Data from the 11,054 genes were quantile normalized using the lumiExpresso function in lumi [32] (S6C and S6D Fig). Only data from autosomal probes were retained for the hierarchical clustering analyses in order to reduce bias towards clustering by individual or sex (n = 10,648 expression, and n = 445,277 methylation). We calculated a matrix of pairwise Euclidean distances between samples from the methylation and expression data separately. From these matrices we performed hierarchical clustering analyzing using the complete linkage method as implemented in the R function hclust. The observed dendrograms remained consistent regardless of the linkage method chosen (complete, single, or average). The 1,000 most variable loci were defined by taking the loci with the highest variance in iPSCs. Clustering based on the 1,000 most variable probes were processed in an identical manner as above. Heatmaps were generated from matrices of pairwise Pearson correlations between samples using data from autosomes and sex chromosomes. Data from probes on both autosomes and sex chromosomes were included in this analysis, given that individuals were balanced across cell types (n = 455,910 CpGs; n = 11,054 genes). Additionally, we anticipated that sites on the sex chromosomes may be particularly sensitive to mis-regulation during reprogramming [36]. Differential expression and methylation analyses were performed using linear modeling and empirical Bayes methods as implemented in the limma package [26]. We tested for differential methylation and expression, using locus-specific models, between L-iPSCs and F-iPSCs; L-iPSCs and LCLs; F-iPSCs and fibroblasts; and between fibroblasts and LCLs. We considered a locus DM or DE at an FDR < 5% (Benjamini Hochberg). We also tested for DE genes between L-iPSCs and F-iPSCs using only genes that were classified as DE between L-iPSCs and LCLs; F-iPSCs and fibroblasts; and LCLs and fibroblasts (S11 Fig). We estimated FDRs separately each time we considered only subsets of the data. Due to the imbalance of L-iPSC samples to F-iPSC samples we repeated our analyses using data from a reduced set of samples. Namely, we randomly sampled a single replicate of the L-iPSC from each individual. As expected, reducing the number of L-iPSC samples greatly reduces the number of loci classified as DM between L-iPSCs and F-iPSCs as well as between L-iPSCs and LCLs. However, the number of DM loci was reduced across all other contrasts as limma models the entire matrix together (S12 Fig). Interestingly, we found that different combinations of replicates yielded DE genes other than TSTD1. Therefore, we sampled all possible combinations and overall, found six genes that were classified as DE (FDR 5%; see S4 Table) in at least one of the combinations of reduced samples. Of note, we never classify TSTD1 as DE (FDR 5%) in the reduced data set. The most common DE gene, INPP5F, is the only gene that also has nearby DM CpGs (five of the 25 nearby loci). Additionally, in the full model, INPP5F has the second lowest P value (uncorrected P = 6. 84 x 10−5; FDR 38%). However, INPP5F was not DE between LCLs and fibroblasts, but was DE between LCLs and L-iPSCs and also fibroblasts and F-iPSCs (S3A–S3D Table; Fig 3C). We employed two strategies to identify enrichments of DM loci between L-iPSCs and F-iPSCs in regulatory features. First, we used the regulatory states defined by Ernst et al. [37]. We tested for enrichments in all regulatory categories using a χ-square test comparing the number DM loci and total probes within each regulatory class to the number DM loci and total probes outside the regulatory class. We found no significant enrichment for any of the defined regulatory states. Next, we used the UCSC_RefGene_Group annotation as supplied by Illumina. These annotations detail the location of probes in relation to genes (1st Exon, 3' UTR, 5' UTR, Gene Body, within 1. 5kb of a TSS or within 200bp of a TSS). We identified significant enrichments of DM loci within 1. 5kb of a TSS and gene bodies. However, there are six probes classified as both within a gene body and within 1. 5kb of a TSS. We chose to report both results because it is difficult to deconvolute these categories. We also considered the position of DM loci in relation to genes. The annotations were defined by Illumina. We were able to identify 37 genes associated with DM loci, but we only had corresponding gene expression data for 24 of these genes. We attempted to identify signals of enrichment in DE levels between L-iPSCs and F-iPSCs in these 24 genes. To this end, we compared the log fold changes in gene expression between L-iPSCs and F-iPSCs from genes with nearby DM loci between L-iPSCs and F-iPSCs to 10,000 random samplings of log fold change in expression between L-iPSCs and F-iPSCs from all genes and found no enrichment for increased log fold changes. To estimate the proportion of variance explained by individual and cell type of origin we performed a linear mixed model with a fixed effect for cell type of origin and a random effect for individual. Only data from autosomes were included in this analysis so that the results would not be biased toward differences in individuals (n = 10,648 expression, and n = 445,277 methylation). To calculate the proportion of variance explained we divided the variance components of each term by the total variance in gene expression (Fig 5). When focusing on CpGs and genes with previously identified genetic associations (eQTLs and meQTLs, respectively) we used genes with at least one eQTL identified by Lappalainen et al. 2013 [38] and CpGs with at least one meQTL identified by Banovich et al. 2014 [39] (S11 Fig). The expression and methylation data sets supporting the results of this article are available in the Gene Expression Omnibus (GEO) under accession GSE65079 (http: //www. ncbi. nlm. nih. gov/geo/query/acc. cgi? acc=GSE65079). All individuals consented to study participation under University of Chicago IRB protocol 11–0524. | Induced pluripotent stem cells (iPSCs) are a new and powerful cell type that provides scientists the ability to model complex human diseases in vitro. These cells can be cryopreserved and later expanded, providing a renewable source of cells from the same individual. iPSCs can be made from a variety of somatic cells in the body and many labs have created them from blood and skin cells. We asked whether the cell type of origin impacts methylation and gene expression patterns in the reprogrammed iPSCs. Our findings indicate that there are remarkably few regulatory remnants of the cell type of origin in the iPSCs. In other words, most of the variation between iPSCs can be attributed to individual genetics. Our findings suggest that studies using iPSCs should focus on obtaining additional individuals rather than additional clones from the same individual. We caution that our current findings are limited to iPSCs and further studies are needed to address the question of somatic memory in differentiated cell types. | Abstract
Introduction
Results
Discussion
Materials and Methods | medicine and health sciences
pathology and laboratory medicine
pathogens
microbiology
fibroblasts
cytoplasmic staining
viruses
stem cells
dna viruses
connective tissue cells
epigenetics
dna
dna methylation
chromatin
herpesviruses
research and analysis methods
specimen preparation and treatment
staining
epstein-barr virus
cell potency
chromosome biology
animal cells
medical microbiology
gene expression
microbial pathogens
connective tissue
biological tissue
chromatin modification
dna modification
pluripotency
genetic loci
biochemistry
cell biology
nucleic acids
anatomy
viral pathogens
genetics
biology and life sciences
cellular types
organisms | 2016 | Genetic Variation, Not Cell Type of Origin, Underlies the Majority of Identifiable Regulatory Differences in iPSCs | 8,699 | 236 |
Much of the complexity of biochemical networks comes from the information-processing abilities of allosteric proteins, be they receptors, ion-channels, signalling molecules or transcription factors. An allosteric protein can be uniquely regulated by each combination of input molecules that it binds. This “regulatory complexity” causes a combinatorial increase in the number of parameters required to fit experimental data as the number of protein interactions increases. It therefore challenges the creation, updating, and re-use of biochemical models. Here, we propose a rule-based modelling framework that exploits the intrinsic modularity of protein structure to address regulatory complexity. Rather than treating proteins as “black boxes”, we model their hierarchical structure and, as conformational changes, internal dynamics. By modelling the regulation of allosteric proteins through these conformational changes, we often decrease the number of parameters required to fit data, and so reduce over-fitting and improve the predictive power of a model. Our method is thermodynamically grounded, imposes detailed balance, and also includes molecular cross-talk and the background activity of enzymes. We use our Allosteric Network Compiler to examine how allostery can facilitate macromolecular assembly and how competitive ligands can change the observed cooperativity of an allosteric protein. We also develop a parsimonious model of G protein-coupled receptors that explains functional selectivity and can predict the rank order of potency of agonists acting through a receptor. Our methodology should provide a basis for scalable, modular and executable modelling of biochemical networks in systems and synthetic biology.
A goal of biology is to understand the structure and function of the biochemical networks that underpin cellular decision-making. One organizing principle is that these networks are inherently modular [1]–[3], with specific functions ascribed to a subset of proteins in the network. Yet, like logic gates in electronic circuits, even individual proteins can perform sophisticated computations and integrate multiple inputs [4]–[7]. In engineering, a modular approach to the analysis of a system scales well with the size of the system and its complexity. Indeed, engineers design systems hierarchically with modules comprising other modules. If molecular biology is similarly modular, which structures are the “atomic” modules from which larger modules are constructed? In signalling networks, we may plausibly ascribe this role to protein subunits and domains [8], [9]. Their function as elementary modules often depends on allosteric transitions: an interaction at one site alters the structure at a distant site via a conformational change. Indeed, allostery increases the information-processing ability of a network because it transforms proteins from passive substrates to dynamic computational elements [10]. A modular approach to the analysis and design of biochemical networks should therefore explicitly describe the computations performed by individual allosteric proteins. Efforts to tackle complexity in biochemical networks should also exploit the modularity of protein structure. Protein structure is hierarchical, and a given protein often has domains also present in other proteins or repeated subunits. For example, many signalling proteins contain SH2 or PDZ domains, and many receptors, ion channels and enzymes are multimers. In genetic networks, transcription factors are also often multimers or have a common DNA-binding domain, such as a zinc finger or homeobox. The re-use of protein domains is both a simplifying and confounding feature: once a domain has been characterized, that characterization can be used again, but it is also necessary to model molecular cross-talk between signalling pathways that contain proteins with similar structures. In vivo, protein interactions can generate both combinatorial and regulatory complexity. Combinatorial complexity is an “explosion” in the number of possible species in a system as the number of proteins and interactions in the system increases. It arises because the number of states of a module dramatically increases as its proteins bind ligands as well as each other and as different residues are covalently modified [11], [12]. For example, p53, the so-called cellular “gatekeeper”, has 37 known modification sites and so potentially 237 states [13]. Thus, a complete description of the system potentially requires a combinatorially large number of chemically distinct species and reactions. In contrast, “regulatory complexity” is a combinatorial increase in the number of parameters required to describe the regulatory interactions within a system as the number of interactions increase. This complexity arises because the strength of protein interactions depends on the state of a module, and each state of the module potentially requires a unique set of parameters to characterize interactions within the module, with other modules in the network, and with molecules external to the network. Measuring this number of parameters in vivo is challenging. Rule-based modelling addresses combinatorial complexity and allows biologists to specify the regulatory logic of a system [14]. Examples include BioNetGen [15], Kappa [16], Moleculizer [17] and StochSim [18]. Rather than explicitly enumerating each species and reaction in the network, a rule-based model describes a system as a collection of biomolecules interacting according to a set of rules. Each rule is a template for a reaction that specifies the reactants, products and all relevant biochemical parameters. Thus, combinatorially complex systems are compactly described because a large number of distinct reactions are subsumed in the template encoded by a single rule. An algorithm may automatically infer a complete reaction network prior to simulation or, if the combinatorial complexity is too great, use alternative techniques to simulate the system [19], [20]. Importantly, some rules also specify contextual conditions that constrain when an interaction can occur and hence encode the regulatory logic of the network. For example, a rule may allow only a doubly phosphorylated MAP kinase to phosphorylate its substrate. Rule-based formalisms can describe complex biochemical systems, but inherently offer little guidance on avoiding a number of methodological problems. First, using rules to specify the regulatory logic of a system does not address the system' s regulatory complexity. Consider G protein-coupled receptors (GPCRs), which allosterically couple an extracellular ligand-binding site to an intracellular G protein-binding site [21]. GPCRs can be promiscuous, binding multiple intracellular targets [22], [23]. Supposing a given GPCR can bind one of L different drugs or endogenous ligands and one of G different G proteins, then in principle we require LG pair-wise cooperativity parameters to describe how each ligand regulates the GPCR' s affinity for each G protein. Thus, the number of regulatory parameters scales with LG, and the number of rules also scales with LG because each parameter is part of a rule with distinct contextual constraints. Promiscuous allosteric proteins can therefore require a large number of rules and parameters to characterize their interactions. Second, a module should have a well-described function and be easily re-used and “portable” between systems, but most rule-based formalisms are not inherently modular. Modellers typically treat proteins as “black boxes” and define interactions using biochemical equations. In such “interaction-centric” approaches, the regulation of proteins is encoded by rules with ad hoc (system-specific) conditions that no longer apply when the proteins interact with different partners. These ad hoc rules obfuscate the mechanism underlying allosteric regulation because they do not show explicitly how the intrinsic structural and thermodynamic properties of allosteric proteins generate their functional properties. In contrast, a “biomolecule-centric” approach would encode regulatory logic in the proteins themselves. Fewer changes to rules would then be required to define how a new set of interaction partners regulates the protein' s activity. If a model includes protein domains and subunits, re-use of these components would also be simplified. Finally, models generated by rule-based methods should be thermodynamically correct. In biochemical networks, there are often sets of reversible reactions that connect into a closed loop, forming a thermodynamic cycle. In many of these cycles no free energy is consumed: for example, when proteins bind multiple ligands, when ligands bind several conformations of a protein, or when ion channels bind multiple agonists and have closed, open, and desensitized states. Thermodynamics imposes a mathematical relationship between the equilibrium constants for all the reactions involved in such cycles: their product must be unity. Equilibrium constants cannot therefore be assigned independently. A thermodynamically correct methodology should ensure that a model satisfies this constraint, ideally by construction. Here, we present a modular and scalable modelling methodology that alleviates the regulatory as well as the combinatorial complexity of biochemical networks. We first describe our modelling framework, which uses a thermodynamically grounded treatment of allostery in which ligands distinguish only the conformational state of allosteric proteins. We also introduce a rule-based modelling tool that implements our methodology: the Allosteric Network Compiler (ANC). We use ANC to examine how allostery can make macromolecular assembly more efficacious. We then show how our modelling framework describes common mechanisms of allostery by mapping the regulatory properties of a protein onto conformational changes in the protein itself and demonstrate how we can ease the analysis of multiple ligands interacting through an allosteric protein. Next, we discuss how our approach reduces regulatory complexity and thereby increases a model' s modularity. Finally, we use our framework to develop a model of G protein-coupled receptors whose regulatory complexity scales with (L+G) instead of LG and consequently has greater predictive power. While our major goal is to introduce a new modular modelling methodology rather than its implementation, we have made ANC and the models we discuss available at: http: //swainlab. bio. ed. ac. uk/anc.
Our method is based on the Monod-Wyman-Changeux (MWC) paradigm of allostery [24]. We assume that allosteric proteins are dynamic and have one or more structural components, such as domains or subunits, with distinct conformations. These conformational states have different biological activity – for example, a basal state with poor affinity to downstream proteins and an active state that can bind these proteins. Thermal fluctuations cause these allosteric components to transition between their two conformations in either a concerted or sequential fashion [25]. Ligands and other molecules interact non-cooperatively with each conformation of the protein, distinguishing only its conformational state, and so contribute independently to the equilibrium of the allosteric transition. Each such contribution is parameterized by a “regulatory factor” Γ, which gives the fold-change of the equilibrium constant generated by interactions with the ligand. We make a similar independence assumption with respect to the transition state of the allosteric transition such that ligands also contribute independently to the transition' s kinetic rate-constants. These contributions to the kinetics are parameterized in terms of one or more “Φ-values”, which give the effect of modifiers on the forward and reverse rates of the allosteric transition (see Methods). Thus, an allosteric protein can be seen as a modular and dynamic computational device, and we can define the input and output of each allosteric component. The input is a “modifier”, a molecule that binds to and locally perturbs the structure of the component; the output is the fraction of time the component spends in each conformation when the allosteric transition is at equilibrium (see Methods). Activation corresponds to biasing the equilibrium in favour of the biologically active state; inhibition corresponds to biasing towards the inactive state. Depending on the system, modifiers may be ligands [24], covalent modifications [26], [27], the conformational state of another component [25], [28], [29], or mutations [30], [31]. An ANC model consists of a set of modular structures and interaction rules. Using our rule-based approach (Tables 1–5 of Text S1, Figures 1–9 of Text S1) and building on the thermodynamic framework described in Methods, each molecule in a system is described using an ANC construct called a structure, which captures the true structure of a protein in terms of its components (Figure 1A). ANC-structures contain two types of components: hierarchical components and interaction sites. Hierarchical components have two roles. The first is that of containment and composition: a hierarchical component typically contains interaction sites but can also contain other hierarchical components. Hierarchical components may represent a unit of tertiary structure, such as a protein domain, or of quaternary structure, for example, a protein with multiple subunits. Their second role is to undergo conformational transitions if designated as allosteric, following the two-state model described in Methods. Interaction sites are of three types: catalytic sites (such as a kinase or phosphatase), sites that can be covalently modified, and ligand-binding sites. Next, rules specify the interactions between sites and how the strength of these interactions depends on the conformational state or the covalent modifications of a protein (Figure 1B). If the interaction is a binding reaction, the rule gives the association and dissociation rates. If the interaction is enzymatic – such as a phosphorylation or dephosphorylation – then we assume a Michaelis-Menten mechanism, and the rules give the rate of formation of the product and the association and dissociation rates between the sites, which must be a catalytic site and a covalent modification site. The overall modelling process for a divalent adaptor protein and two ligands is illustrated in Figure 1C. Structures and rules are entered as text and saved to a file (section §2. 7 of Text S1). ANC reads the file, creates an initial set of seed structures, and launches an iterative compilation algorithm. At each iteration, the algorithm determines all inter- and intramolecular reactions, the products created, and their biochemical rates. In a subsequent iteration, the newly created products may in turn react to produce yet more species. Once a final biochemical reaction network has been obtained, it is simulated using deterministic or stochastic methods (see Methods). The deterministic simulation in Figure 1C shows how upon binding X, which could represent an activated receptor, the adaptor A recruits increasing amounts of Y, which could represent a downstream signalling protein. The generic model of a divalent allosteric protein shown in Figure 1C (full details in Figure 7 of Text S1) can be used to model proteins that play other roles than adaptors. For example, A could be a membrane-bound receptor, X an extra-cellular agonist, and Y an intracellular signalling protein which binds preferentially to the active conformation of the receptor. The usefulness and simplicity of this model motivates us to analyze it mathematically. In the model, the binding of X and Y to A is cooperative because binding of X to A changes the affinity of A for Y by a factor θ and likewise the binding of Y to A changes the affinity of A for X also by a factor θ. By coarse-graining over the conformations of A (Figure 2A, inset), we can express the cooperativity parameter θ as (section §2. 1 of Text S1): (1) where KRT is the allosteric equilibrium constant and ΓX (or ΓY) is the differential affinity of the X (or Y) to each conformation of A. The cooperativity increases as the degree of bias (ΓX and ΓY) that X and Y exert on the conformational transitions of A increase. We can also define the apparent affinity of X and Y to this coarse-grained A: (2) These equations relate the underlying parameters of the model to experimental observables: both the affinity KX of X to A and the affinity θKX of X to A when A is bound by Y are measurable. Counter-intuitively, an excess of some components of a macromolecular complex can inhibit formation of the complex [32], [33]. This phenomenon, called the prozone effect, is strongest for a protein that links two or more separable parts of a complex. It occurs because the linker protein competes with itself for the binding of the other components of the complex, and so if it is present in excessively high amounts, few of the linker proteins will succeed in simultaneously binding all the other components, resulting in partially formed complexes. Here, we show that allostery can mitigate the prozone effect, at least for a divalent allosteric protein. We consider the divalent structure A to represent a linking protein with X and Y being the remaining parts of a complex. In Figure 2A, we demonstrate how increasing the cooperativity increases the range of concentrations of A for which assembly of the complex XAY is efficacious (i. e. where the equilibrium amount of complex exceeds 50% of the maximum amount). This range increases from 1. 3 decades when θ = 1 to 4. 0 decades when θ = 104, and the maximal amount of complex formed increases by a factor of 5. 7 (Figure 11 of Text S1). Thus, an allosteric linker protein has a dual benefit in macromolecular assembly: it both increases the amount of complex when the components are present in their stochiometric amounts and makes complex formation more robust to the over-expression of the linker protein. Allosteric linker proteins could explain the low correlation observed between over-expression of a linker protein and lethality of that over-expression in budding yeast [34]. That the efficacy of macromolecular assembly depends strongly on the value of the cooperativity parameter θ suggests that assembly could be modulated by changing θ. Figure 2B shows the dependence of θ on the allosteric equilibrium constant and the differential affinity of the ligands X and Y. Cooperativity has a maximum at KRT = (ΓXΓY) −1/2 and thus assembly of the XAY trimer could in principle be controlled through the binding of a cofactor or a covalent modification that changes the allosteric equilibrium constant of A from a value far from its optimum to a value near the optimum (or vice versa). There are two well-known mechanisms for generating cooperative behaviour in proteins: concerted and sequential allostery. In their seminal paper, Monod, Wyman and Changeux introduced a two-state model to explain cooperative interactions in oligomeric enzymes and proteins [24]. They proposed that all subunits of such proteins undergo a concerted, reversible, and quaternary-level transition between two conformational states. Ligand-binding to each conformation is non-cooperative, but each conformation differs in its affinity for ligands and this difference gives rise to cooperative effects. Subsequently, Koshland, Nemethy and Filmer lifted the assumption of concerted transitions with their sequential model, in which each subunit transitions individually between two conformational states [25]. This model can explain negative cooperativity in oligomeric proteins. It assumes, however, that ligands cause an instantaneous conformational change, or an induced fit, in the structure of the subunit. Both the concerted and induced fit assumptions can be relaxed and are special cases of the general allosteric model of Herzfeld and Stanley [28]. ANC-structures can be used to implement these models of allosteric regulation. A concerted model of a generic, homotetrameric protein is shown in Figure 3A. The transition between the two conformations, labelled R and T, is concerted because a single allosteric component contains all subunits. Cooperativity will arise if a ligand has a higher affinity for one state, say the R state, and if the unligated protein is mostly in the alternate state T. Then, once bound by ligand, the protein spends more time in the R state and so favours the binding of additional ligands. In contrast, a sequential model has an ANC-structure with four allosteric components, each with r and t states. Figures 3B and 3C show how we implement the tetrahedral and square geometries described by Koshland et al. [25] through different configurations of the allosteric coupling between subunits. Two components are allosterically coupled if the conformation of one component biases the conformational equilibrium of the other component, and vice versa. For example, in Figures 3B and 3C, ligand binding favours the r state of an individual subunit, and this subunit when in its r state favours the r state in those subunits to which it is allosterically coupled and so generates a cooperative response. Finally, we illustrate the general approach with the tertiary two-state model shown in Figure 3D, which allows both quaternary, R↔T and tertiary, r↔t, allosteric transitions [29]. Here, ligand binding favours r at the bound subunit because of the ligand' s higher affinity for r. Cooperativity arises because the R state of the quaternary structure reciprocally favours the r state of the tertiary subunits. Thus, a subunit in the r state favours R at the quaternary level and so favours subunits not yet bound by ligand to also be in the r state, promoting binding of additional ligands. Our ANC implementation of the tertiary two-state model correctly reproduced the Henry et al. model, which has 252 molecular species [29]. An advantage of ANC is its ability to easily formulate and simulate mathematically complex models. For example, we will show that the cooperativity of an allosteric protein binding a ligand, such as a transcription factor binding an inducer, can be substantially changed through adding a competing ligand. Although a mathematical analysis of various allosteric models with two competing ligands exists [35], little is known about multiple ligands and the analysis is cumbersome for the sequential model despite simplifying assumptions. Using ANC, we characterized the binding cooperativity of a ligand L0 to a tetrameric allosteric protein in the presence of one of three different competing ligands for both the concerted and sequential models (Figure 4). In the absence of competitors, the Hill coefficients for binding L0 in the concerted and sequential models were ∼2. 8 and ∼2. 2 respectively. By increasing the concentration of the competitor ligand L1, which binds preferentially to the same conformation as L0, the Hill coefficient decreased progressively to 1 (i. e. no cooperativity). With ligand L3, which binds preferentially to the low affinity state for L0, the Hill coefficients increased to ∼3. 6 and ∼3. 4. With ligand L2, which binds with equal affinity to all conformations, the Hill coefficient did not change. With competitors L2 and L3, the EC50 of L0 binding increased but at low concentrations of L1, the EC50 of L0 was slightly lower (Figure 12 of Text S1). In addition to ligand binding, our methodology also describes other mechanisms for allosteric regulation that are ubiquitous in cellular signalling. Phosphorylation or other post-translational modifications, dimerization, receptor clustering and point mutations can also regulate or change protein function. Our thermodynamic framework (see Methods) unifies the treatment of such heterogeneous modifiers of protein activity. In section §2. 4 of Text S1, we discuss how dimerization and ligand binding jointly regulate the activity of the epidermal growth factor receptor, and how ligand binding combines with methylation to regulate a chemotaxis receptor (Figure 13 of Text S1). We can distinguish two types of parameters that affect modularity in different ways: intensive parameters and extensive parameters. Intensive parameters describe the conformational transitions and intramolecular interactions of a protein and, as such, are modular because they are inherent to the protein and independent of the protein' s interaction partners. Therefore, we associate intensive parameters with a protein' s ANC-structure. In contrast, extensive parameters describe the interactions of a protein with other biomolecules and increase in number as the number of these interactions increase. Extensive parameters, contained in rules for interaction, are the “wiring” between modules and are non-modular because they depend on the system in which the protein functions. Regulatory complexity occurs when the number of extensive parameters describing a system scales combinatorially with the number of interactions in the system. Our biomolecule-centric methodology minimizes regulatory complexity. For example, we analyzed a generic model of an N-valent, two-state protein where each of the N binding sites is unique and binds exactly L ligands (section §2. 5 and Table 6 of Text S1). In an “interaction-centric” modelling approach, there are no intensive parameters and the number of extensive parameters scales as LN. In the “biomolecule-centric” methodology of ANC, there are 2 intensive parameters and the number of extensive parameters scales linearly with the number of interactions NL. Using our methodology can therefore yield large savings; for instance if N = 6 and L = 5, we have 91 independent rate constants rather than over 233,000 (Table 7 of Text S1). Thus, by encoding the regulatory logic of proteins with intensive rather than extensive parameters, we reduce regulatory complexity. We therefore improve the model' s modularity because only extensive parameters change when a model is updated. Using our biomolecule-centric modelling framework, we can convert a non-modular model into a modular one. Such refactoring is also useful when a protein has more than two conformational states, unlike the core allosteric components in ANC-structures. To illustrate, we introduce a new model for the activation of G protein-coupled receptors (GPCRs). GPCRs are a common target for pharmaceutical drugs [36]. Such drugs include agonists that promote activation of the receptor, inverse agonists that promote deactivation of the receptor, and antagonists that by binding to the receptor block the action of agonists. Although several allosteric models have been proposed [37], [38], we will consider the cubic ternary complex model [39] because this model describes the constitutive activity of a GPCR, the action of inverse agonists and antagonists, and how some inverse agonists can cause a GPCR to recruit G proteins but remain inactive [40]. The model also explains the functional selectivity of receptors (also called agonist trafficking, ligand-biased agonism, or protean agonism) [41] through the notion of active states of the receptor that are specific to a ligand or a G protein [38], [42]. However, the model does not include receptor homo- or hetero-oligomerization [43]–[48], or the possibility that GPCRs form stable, pre-assembled complexes with downstream proteins [49], [50]. A naive implementation of the cubic ternary complex model in our framework uses a divalent ANC-structure with a single allosteric component (Figure 5A, C). This implementation captures Weiss et al.' s assumption that the receptor has only two conformational states. However, it does not capture the cooperative binding of a ligand to either the inactive or active states of the receptor because such binding is incompatible with the paradigm that each modifier contributes independently to the equilibrium constant of the allosteric transition [24]. This cooperative effect is described in Weiss et al.' s model through the cooperativity parameters γ and δ. However, these parameters are extensive and specific to each combination of ligand and G protein. They therefore introduce regulatory complexity. To resolve this difficulty, we propose a sequential allosteric model of the GPCR with two coupled allosteric components: an extracellular allosteric component, which binds a ligand, and an intracellular allosteric component, which binds a G protein (Figure 5B, D). The ligand and the G protein interact simultaneously with both allosteric components. They therefore “see” four possible conformations of the receptor instead of two. These conformations are implied in the cubic ternary complex model because each ligand has four distinct affinities to the receptor. However, none of the extensive parameters in our model are cooperativity parameters specific to a ligand-G protein pair, thus eliminating regulatory complexity. Our quartic ternary complex model can be projected onto the cubic model by defining coarse-grained variables that sum over the conformations of the extracellular allosteric component (Figure 5 and section §2. 6 of Text S1). The “inactive” and “active” states in the cubic model therefore correspond to a mixture of conformational states, providing a mechanism for how different ligands induce an apparently unique conformation of the activated GPCR with a distinct affinity for the G protein [51]. In our model, each ligand uniquely affects the allosteric equilibrium of the extracellular domain and therefore the fraction of time that the receptor is in the s and t states, which in turn uniquely modulates the affinity of the active GPCR for the G protein. Our quartic model for the GPCR is more modular and parsimonious than the cubic model because it includes a structurally and biophysically plausible mechanism for how ligands and G proteins interact cooperatively with the GPCR. We encode the logic of these regulatory interactions in the protein' s ANC-structure using intensive parameters, rather than in ad hoc rules with extensive parameters. Our “refactored” model has 11 parameters (3 of which are intensive) compared to the 7 parameters of the cubic model (1 of which is intensive) and double the number of states. This initial cost for increased modularity and “portability” becomes a benefit as the number of types of ligands and G proteins increases. The number of extensive parameters in our model scales linearly with the number of interactions; in the cubic model, the number of extensive parameters scales combinatorially. For example, suppose we wish to model 4 different ligands that activate the thyroid-stimulating hormone receptor. In human thyroid membranes, this GPCR can activate at least 10 different G proteins [22]. With 4 ligands and 10 G proteins, our quartic model is almost twice as parsimonious as the cubic model, requiring 59 rather than 109 parameters. The quartic model also has more predictive power than the cubic model and therefore can be more rigorously tested. For each pair of ligands and G proteins, the cubic model requires the specification of two cooperativity parameters, δ and γ, specific to that pair. It is therefore limited in the predictions it can make. For example, for each new G protein added to the system, new cooperativity parameters are needed for all previously characterized ligands to be able to predict the new G protein' s GPCR-mediated response to these ligands. In contrast, the quartic model is completely characterized for the new target pathway by measuring four extensive parameters – one for each conformation of the GPCR – and we can then predict the GPCR-mediated response to all ligands. In particular, we can predict the rank order of potency of the ligands to activate the new pathway, a standard means to compare agonists in pharmacology, and detect functional selectivity [51]. Like the cubic model, the quartic ternary complex model also explains functional selectivity, though this is not obvious considering that these models cannot be related through a simple projection when multiple ligands and G proteins interact with a single receptor. Indeed, in the quartic model δ and γ are not free parameters but are correlated because of their dependence on underlying rates. We therefore simulated the GPCR-mediated response to several ligands that cause (in) activation of two different G proteins (Figure 6A and 6B). Ligand L1 has the greatest ability to activate G protein G1 as measured by its potency (−log (EC50) of the response) and efficacy (maximal activation). For G1, ligand L2 has intermediate potency and efficacy and L3 has the lowest potency and efficacy. Also, L1 is better able to activate G1 than G2. If the receptor had only a single active state, we would therefore infer that this state must have a poorer ability to bind and activate G2 and would expect that L2 should also have a decreased ability to activate G2. The potency and efficacy of L2 on G2 is, however, greater than that of L1 indicating a reversal in the rank order of the potency and efficacy of L1 and L2. Also, ligand L3 is an agonist for G1 but an inverse agonist for G2. These observations of agonists selectively (in) activating distinct target pathways cannot be reconciled with a model comprising a single active state of the receptor. Functional selectivity can also be observed for the repression of activity by inverse agonists because there are also two inactive states of the receptor (Figure 6B). The quartic model is modular and therefore is easily extended to include additional signalling interactions such as the regulation of the receptor by allosteric ligands [52], [38]. Also, by adding dimerization sites, we could incorporate existing models of dimerization of GPCRs [47]. Some GPCRs may also oligomerize in vivo, for instance by forming tetramers [53]. We could model oligomerization by concatenating multiple receptor models within a larger ANC-structure through modular composition. A starting point could be one of the models of Figure 3, but substituting for each subunit the ANC-structure for a GPCR and, as appropriate, adding allosteric couplings to model inter-receptor interactions.
Biochemical networks are complex yet modular: networks exhibit both combinatorial and regulatory complexity, but individual proteins have intrinsic functional properties that determine how they detect and process information. Complexity is also reduced because similar proteins or similar protein domains appear in many signalling pathways and often interact with similar protein partners. We propose a modelling methodology, embodied by ANC, that exploits the modularity of proteins to reduce the complexity of modelling biochemical networks. Given modular ANC-structures, which encode a protein' s regulatory properties, adding new interactions to an ANC model usually requires substantially fewer parameters than with other rule-based models, particularly as the promiscuity of binding of proteins, and hence the complexity of the network, increases. ANC-structures are also portable because different signalling pathways are modeled by simply “re-wiring” proteins rather than through writing new ad hoc rules encoding the regulatory logic specific to each pathway. In our methodology, models are structured to minimize regulatory complexity both to avoid over-fitting data and because large numbers of biochemical parameters are difficult to measure in vivo. Indeed, our modelling framework reflects a natural division between two classes of parameters: “intensive” parameters describe the allosteric transitions and intramolecular interactions of a particular protein and are attributes of ANC-structures; “extensive” parameters describe the interactions of the protein with other biomolecules and are associated with rules. In different biochemical networks, only the extensive parameters of a protein change. Through its assumption that a ligand or substrate distinguishes only the conformation of a protein or enzyme and not its occupancy or state of covalent modification at distant sites, our modelling framework substantially reduces the number of extensive parameters required: their number scales linearly, rather than combinatorially, with the number of interactions. This reduction in extensive parameters allows a model for the interaction of a protein with individual ligands to also predict the response to mixtures of ligands with no additional parameters [54]. We have shown that interference with a competing ligand can either increase or decrease the cooperativity of the response to the original ligand (Figure 4) – a potentially useful mechanism of control. Our methodology reduces the regulatory complexity but increases the combinatorial complexity of a system because each conformation of an allosteric protein introduces a new state. Thus, a reduction in regulatory complexity incurs the computational cost of modelling additional species. Nevertheless, recent advances in rule-based modeling have introduced new methods that allow fast simulation of systems with large numbers of chemical species and reactions [17], [19], [20], [55]–[57]. By focusing on avoiding an exponential increase in extensive parameters in systems with promiscuous binding, our methodology both complements and potentially benefits from these innovations. We also make a first step at integrating free energy-based constraints into a rule-based modelling framework, adding to earlier work on imposing detailed balance in models of biochemical networks [58]–[61]. By automatically computing all dependent extensive parameters associated with allosteric transitions – the allosteric equilibrium and rate constants for each ligated and modified state – from the appropriate independent parameters, ANC prevents the modeller from incorrectly specifying these parameters. Thus, cycles comprising allosteric transitions are biophysically correct by construction. For complex models with a combinatorially large number of occupancy states and covalent modifications, this automation is essential. Two other advantages of our modelling framework are significant. First, ANC-structures enable a coarse-grained hierarchical description of physical structure by requiring the specification of protein domains and if desired tertiary and quaternary structure, including oligomeric receptor clusters. ANC-structures can also model the internal geometry of a protein by describing those domains of the protein that interact allosterically and those that do not (Figure 3). Second, the thermodynamic framework underpinning our method offers a systematic and unified way to model how proteins integrate heterogeneous inputs such as ligands, phosphorylations, or even small mutations to compute an output response. Our modelling framework encourages the modeller to develop a mechanism to explain the regulatory properties of a protein and hence to build models that have predictive power and so can be experimentally tested. For example, an ANC model of the activation of GPCRs suggests that the well-known cubic ternary complex model has implicitly coarse-grained some conformations of the GPCR. By including these conformations in an ANC-structure, our new quartic model prevents over-fitting and has the potential to predict the rank order of potency and efficacy of ligands acting through a GPCR. This model of the GPCR has two linked allosteric components, each with just two conformational states that interact independently with other molecules. These mechanistic assumptions do not, however, apply to the GPCR as a whole, which has four conformational states. Thus, while the two-state assumption may not hold for all proteins, other mechanistic models can be accommodated within our framework. Having allostery at its centre, our framework can suggest simple mechanisms through which the cell might regulate and increase the efficacy of cellular processes. For example, the assembly of macromolecular complexes can be considerably undermined through the prozone effect when linker proteins are over-expressed [32]. Consequently we might expect that expression of the components of macromolecular complexes is tightly regulated. Such regulation can be complex and expensive. Yet modelling with our framework suggests that if allosteric proteins are part of the macromolecular complex and if the linker proteins are allosteric then the prozone effect can be substantially reduced and without energy input (Figure 2A). A challenge in designing synthetic biological systems is to have predictive modelling tools. Here, ANC has several potential advantages. First, the modularity of ANC-structures allows models of synthetic systems to be straightforwardly extended: for example, as different synthetic subsystems are combined to generate more complex behaviour [7]. Second, through its rule-based modeling and the specification of rules of interaction between protein components rather than between proteins, ANC naturally models molecular cross-talk between synthetic sub-circuits in a larger synthetic circuit and between a synthetic circuit and the endogenous biochemistry (if a rule-based model of endogenous signalling is available). In both cases, ANC will find and model interactions if proteins are present that happen to have complementary binding domains. Such interactions could, for example, affect the formation of macromolecular signalling complexes (Figure 2) or change the Hill coefficient of the response of a crucial pathway (Figure 4). Finally, an ANC model includes background activity in all enzymes because control of each enzyme is described by an allosteric transition between inactive and active conformations. This transition will occur regardless of the presence of input signals to the system, although the probability of such occurrences can be small. Like molecular cross-talk, background activity can cause a synthetic circuit to deviate substantially from its designed behaviour. Faced with the complexity of cellular signalling and genetic networks, researchers are developing new computational methods to quantitatively model and predict cellular behaviour despite that complexity. In this spirit, we have identified and discussed a distinct form of complexity – regulatory complexity – which arises from the allosteric regulation of proteins. Combining and extending established biophysical principles with more recent rule-based methods, we propose a modular and scalable methodology, exemplified by our Allosteric Network Compiler, to describe the complexity of cellular signalling. By emphasizing the allosteric control of proteins, we capture the inherent modularity of protein structure and function exploited by cells themselves. Our method is a general, principled and simplifying addition to any modeling framework.
To compute how multiple modifiers collectively bias the conformational equilibrium of an allosteric component, we use thermodynamics [62]. We arbitrarily designate one conformation as the reference state R and the alternate conformation as T. If we let be the difference in free energy between the R and the T conformations in the absence of any modifiers, then the difference in free energy in the presence of N modifiers is, quite generally [63], (3) where we include contributions of free energy to the new allosteric equilibrium that are determined by each modifier alone, by pair-wise interactions between modifiers, and by all higher order interactions. We assume that all modifiers interact independently (non-cooperatively) with each conformational state of the protein component, with the energy of interaction to the R state of the component given by and to the T state by for the modifier indexed by i. Consequently, the free energy required to apply a modifier to each conformation does not depend on the presence or absence of other modifiers – a modifier can only distinguish the conformational state of the component. Therefore we need consider just the first order terms of equation (1) and ignore higher order interactions: (4) For each modifier, a reversible thermodynamic cycle exists around which the change in free energy must be zero. For example, equilibria exist between the R and the T states of the component, between the modifier i being applied to the R state (a free energy change of), the modifier being applied to the T state (a free energy change of), and between the R and the T states of the modified form of the component. To have no change in free energy around this cycle implies that. Hence, we have: (5) From statistical mechanics, we know that the equilibrium constant between any two states of a system, say A and B, is connected to the difference in their free energy through the expressionTherefore, we may exponentiate equation (3) to find the corresponding equilibrium constant: (6) with kT denoting the product of Boltzmann' s constant and temperature. Equation (6) describes the input-output function of an allosteric component, which may embody a domain, a subunit, or an entire protein. The output is the allosteric constant of the component under the effect of N modifiers. It is obtained by multiplying a baseline equilibrium constant with each “regulatory factor” Γi, which describes the effect of an input modifier i on the allosteric equilibrium. If the modifier is a ligand, then Γi is the ratio of the ligand' s affinity to each conformation. If the modifier is a covalent modification such as a phosphorylation, the regulatory factor is an independent parameter related to the free energies required to phosphorylate each conformation. If the modifier is another allosteric component to which the component is allosterically coupled (e. g. Figure 3B–D), then Γi is an independent parameter related to the free energy of interaction of the T form of the modifier with each conformation of the allosteric component, and gives the fold-change in the allosteric equilibrium constant induced by the T form of the modifier. When this modifier is in its reference state, which we label R, the output is by definition unchanged and the regulatory factor is not applied. To compute how the kinetics of a component' s allosteric transition are affected by the presence of modifiers, we first write the forward and backward rate constants for the unmodified component in terms of the free energy difference between the transition state (denoted †) and each conformational state [64]: (7a) (7b) To obtain the rate constants for a modified state, the simplest approach is to assume a constant pre-exponential factor C and that modifiers contribute independently to a change in the free energy of the transition state (section §1. 2 of Text S1). The assumption of independence is not arbitrary: at the core of the MWC paradigm of allostery is the assumption that modifiers contribute independently to the free energy of each conformational state. Here we extend this idea to the transition state of the allosteric transition. As a result, modifiers independently affect the kinetic rates, just as they do the equilibrium constant, and we can write: (8a) (8b) for parameters Φi. We choose this parameterization because Φi = Φj implies the existence of a linear free energy relationship for two modifiers i and j (section §1. 3 of Text S1). A linear free energy relationship [65] is a common, simplifying assumption in biophysical models: in a set of related reactions, the logarithm of a transition rate is assumed to be linearly related to the logarithm of the equilibrium constant [31], [66]. The parameter Φ denotes the proportionality constant. Assuming a linear free energy relationship to model simultaneous modifiers, for example in models of hemoglobin or of the nicotinic acetylcholine receptor [29], [66], [67], also implies that these modifiers independently affect the conformational transition, with each effect parameterized by the same value of Φ (section §1. 3 of Text S1). We validated our overall methodological flow (Figure 1C) and verified the output of ANC by implementing and simulating an allosteric model [68] of the signalling protein calmodulin (Figure 10A and Figure 10B of Text S1). Binding of calcium to calmodulin modulates its affinity for downstream effectors. We confirmed that ANC correctly generates the 352 biochemical equations of the model of Stefan et al. and that our simulation results were consistent with theirs, using their experimentally derived parameter values (Figure 10C of Text S1). ANC possesses a number of features which ease modelling and simulation of biochemical networks. First, ANC allows users to parameterize a model so that parameter values can be changed after compilation. Also, ANC supports stimuli, through which the user can apply input waveforms to specified nodes in the network, and probes – user-defined collections of molecules – to measure network output. Finally, ANC allows the creation of ad hoc regulatory conditions to support interaction-centric approaches. Such ad hoc conditions, however, reduce the modularity and scalability of a model and so do not play to the strength of our methodology. Using Facile [69], an application distributed with ANC, we can export an ANC-compiled network to standard tools such as Matlab, XPP, Maple or Mathematica for deterministic simulation or analysis, to EasyStoch for stochastic simulation [70], or to SBML [71]. The current implementation of ANC has three principle limitations. 1) The reaction network is enumerated, so ANC' s performance may degrade significantly if the compiled network is large. 2) Only rules for binding and Michealis-Menten interactions can be created. 3) While ANC supports unimolecular association and dissociation, detailed balance is enforced only for cycles comprising purely bimolecular associations. | The complexity of biochemical networks challenges our ability to create quantitative and predictive models of cellular responses to extracellular changes. In these networks, the regulation of allosteric receptors and proteins by multiple drugs or endogenous ligands introduces “regulatory complexity” because a large number of parameters is required to describe such interactions. Protein interactions also give rise to “combinatorial complexity” by generating large numbers of protein complexes and covalent modification states. To address these twin problems, we propose a modelling framework that combines a modular description of protein structure and function with a rule-based description of protein interactions. We define the input-output function of an allosteric protein through its thermodynamic properties and structural components. We show that our “biomolecule-centric” methodology, in contrast to ad hoc approaches that emphasize the regulatory logic of interactions, can reduce the number of parameters required to model experimental observations. We also demonstrate how the application of our framework gives insights into the assembly of macromolecular complexes and increases the predictive power of a standard model of G protein-coupled receptors. These benefits are possible in many systems, given the ubiquity of allostery in biochemical networks. Our research delineates a fundamental relationship between allostery, modularity, and complexity in biochemical networks. | Abstract
Introduction
Results
Discussion
Methods | biophysics/macromolecular assemblies and machines
biophysics/membrane proteins and energy transduction
computational biology/synthetic biology
computational biology/transcriptional regulation
biophysics/theory and simulation
biophysics/biomacromolecule-ligand interactions
pharmacology
computational biology/signaling networks
computational biology/systems biology | 2010 | Scalable Rule-Based Modelling of Allosteric Proteins and Biochemical Networks | 11,286 | 290 |
Disassembly of the nuclear lamina is essential in mitosis and apoptosis requiring multiple coordinated enzymatic activities in nucleus and cytoplasm. Activation and coordination of the different activities is poorly understood and moreover complicated as some factors translocate between cytoplasm and nucleus in preparatory phases. Here we used the ability of parvoviruses to induce nuclear membrane breakdown to understand the triggers of key mitotic enzymes. Nuclear envelope disintegration was shown upon infection, microinjection but also upon their application to permeabilized cells. The latter technique also showed that nuclear envelope disintegration was independent upon soluble cytoplasmic factors. Using time-lapse microscopy, we observed that nuclear disassembly exhibited mitosis-like kinetics and occurred suddenly, implying a catastrophic event irrespective of cell- or type of parvovirus used. Analyzing the order of the processes allowed us to propose a model starting with direct binding of parvoviruses to distinct proteins of the nuclear pore causing structural rearrangement of the parvoviruses. The resulting exposure of domains comprising amphipathic helices was required for nuclear envelope disintegration, which comprised disruption of inner and outer nuclear membrane as shown by electron microscopy. Consistent with Ca++ efflux from the lumen between inner and outer nuclear membrane we found that Ca++ was essential for nuclear disassembly by activating PKC. PKC activation then triggered activation of cdk-2, which became further activated by caspase-3. Collectively our study shows a unique interaction of a virus with the nuclear envelope, provides evidence that a nuclear pool of executing enzymes is sufficient for nuclear disassembly in quiescent cells, and demonstrates that nuclear disassembly can be uncoupled from initial phases of mitosis.
The nuclear envelope separates cytoplasm and nucleus requiring shuttling of cargos between the compartments. In non-dividing cells macromolecule exchange occurs via the nuclear pore complexes (NPC), which are composed of ∼30 different proteins (nucleoporins, Nups). NPCs allow the passage of macromolecules only in complex with soluble transport receptors as e. g. the nuclear import receptors of the importin (karyopherin) ß superfamily [1]. During transport the receptors interact with those nucleoporins comprising FxFG repeats, which are localized on unstructured domains [2]. At the end of nuclear import this complex becomes dissociated by the small GTPase Ran in its GTP-bound form. While the cargo diffuses deeper into the karyoplasm, the receptor-RanGTP complex is exported into the cytoplasm [3]. The nuclear envelope is composed of the double lipid bilayer of outer nuclear membrane (ONM) and inner nuclear membrane (INM) and a matrix of proteins separating INM and the chromatin. The matrix is composed of both peripheral and integral membrane proteins, including lamins and lamin-associated proteins. The nuclear lamina is required for proper cell cycle regulation, chromatin organization, DNA replication, cell differentiation, and apoptosis [4]. In contrast to closed mitosis in yeast open mitosis as it is the case in other eukaryotes but also apoptosis requires that the nuclear envelope (NE) disassembles (nuclear envelope breakdown, NEBD), involving depolymerization of the lamin network. In mitosis, NEBD starts at a single hole in the nuclear envelope, which expands within minutes over the nuclear surface [5]. As the space between ONM and INM in continuation with ER lumen is the space where free Ca++ is stored increased perinuclear Ca++ is observed directly before the NE disintegrates [6]. In contrast NEBD in apoptosis is characterized by dynamic nuclear membrane blebbing and fragmentation [7]. Several enzymatic activities participate in NEBD. In mitosis, lamin depolymerization is executed by hyper phosphorylation of lamin A/C, B1, B2 comprising different protein kinase C isoforms and cyclin-dependent kinases (cdks); their balanced activities control G1/S transition [8]. The role of caspase-3 in mitosis is controversial [9]–[11]. NEBD in apoptosis requires PKCδ and cdks but nuclear dismantling depends on caspase-3 [12]. NEBD is tightly controlled by the cdks and PKC isoenzymes activities. Their balanced activities controls G1/S and G2/M transitions and links signal transduction pathways to the cell cycle machinery [13]. Several reasons complicate research on NEBD: the regulations and interactions are complex and the executing enzymes - as it was described for PKC α/δ and caspase-3 - become imported into the nucleus during the initial phases of apoptosis or mitosis [14]–[16] where they fulfil other functions as for instance lamin phosphorylation and degradation than in the cytoplasm. Parvoviruses (PV) are well conserved viruses, comprising dependo-viruses as the adeno-associated virus (AAV), and autonomous PV as the canine parvovirus and H1. PVs are used in gene therapy trials and AAV-based vectors were recently licensed for gene therapy of lipoprotein lipase deficiency. Parvoviruses are composed of two (three in AAV) co-terminal structural proteins, VP1 and VP2, which form a capsid of 26 nm in diameter. The larger protein (VP1) has an additional/unique N terminal sequence (VP1u) comprising a potential nuclear localization signal (NLS) and a phospholipase A2 (PLA2) activity, which is essential for infection [17]. VP1u is hidden within the virion but is predicted to become exposed during infection [18], [19] while the capsid remains intact. PVs enter the cell by endocytosis and are subsequently transported in endosomes towards the nuclear periphery [20]. Acidification is required for infection and only a small proportion of PV escape the late endosomes [21], predicted to be mediated by the PLA2 domain on VP1u. PVs contain a single stranded DNA genome, which is replicated by cellular DNA polymerases inside the cell nucleus. DNA release remains poorly understood but occurs without capsid disassembly [22] and at least AAV2 enters the nucleus fully assembled according to the majority of studies (e. g. [23], [24]). The interaction between PV and nuclear envelope are not fully understood. After microinjection of canine PV into the cytosol, capsids appear in the nucleus after hours [25]. It remains open if these capsids were derived from nuclear import of the microinjected capsids or from progeny capsid formation. Microinjection of Minute Virus of Mice (MVM) into Xenopus laevis oocytes cause distinct breaks of the nuclear envelope [26], [27], which could be large enough to allow entry of the PV capsids into the nucleus [28]. We investigated the interaction between PV and the nucleus in more detail finding that PV attached directly – without the need of nuclear import receptors - to the NPC, which activated an intranuclear cascade leading to degradation of the nuclear envelope.
Upon infection of 1000 genome-containing PV H1 per cell we observed local NEBD at the site where PV accumulated. NEBD was indicated by the loss of NPC stain and – in some cases - chromatin escape into the cytoplasm (Fig. 1A). Nuclear envelope disintegrations were observed in 11% of the infected cells (102 out of 913) at a time before progeny viruses are made. In non-infected cells <1% of cells (2/294) showed such damage indicating that the phenomenon was parvovirus-dependent. This assumption was supported by the observation that disintegration only occurred at those sites of the nucleus were parvoviruses accumulated (Fig. 1B). We did neither observe chromatin condensation as it occurs in prophase of mitosis nor that we monitored the formation of chromatin patches closed to the NE as in apoptosis. Chromatin fragmentation, yet another characteristic of apoptosis was also not observed. Similar local disruptions were however observed recently upon egress of cytomegalovirus capsids [29] and also – as a temporary phenomenon - for MVM [27]. To investigate virus-induced nuclear disassembly in more detail we analysed the PV-mediated NEBD by digitonin-permeabilized HeLa cells using confocal laser scanning microscopy. Digitonin permeabilizes cholesterol-containing membranes leaving the nuclear and ER membrane intact [30]. Permeabilization was stopped by washing the cells at 4°C prior to destruction of the part of the plasma membrane, which connects the permeabilized cell with the cover slip. At 4°C microtubules depolymerise and the washes remove soluble cytoplasmic proteins including nuclear transport receptors. Accordingly we did not observe active nuclear import after addition of a karyophilic cargo (Supporting Information, Fig. S1A) and α tubulin was reduced to 2% compared to unpermeabilized cells (Supporting Information, Fig. S1B). We thus concluded that permeabilized, washed cells are devoid of significant amounts of soluble cytosolic proteins including nuclear import receptors. We added H1 to the permeabilized cells in the absence of cytosolic factors but in which cellular chromatin was stained by propidium iodide (PI), allowing to record nuclear integrity by time-lapse microscopy. Figure 1C shows that chromatin fluorescence disappeared upon addition of H1. As in infected cells no significant chromatin condensation was found, which is exemplified in figure 1D. Only little loss of fluorescence (4% in average) was observed in nuclei to which buffer was added, probably due to bleaching of the stain upon illumination with the laser beam. Loss of fluorescence was dye-independent as the same results were obtained upon chromatin stain using DAPI (see below), implying that the loss of stain was based on chromatin escape. Quantification of the chromatin escape in several assays was highly reproducible but varied between individual nuclei (Fig. 2A). The distribution of the fluorescence loss showed a Gaussian normal distribution (not shown), which allows presentation of the results as mean values with 95% confidence intervals (95% CI) shown in figure 2B. In fact non-overlapping CIs depict statistically significant differences. Figure 2B further shows that 50% of fluorescent chromatin was lost in mean within 4. 6 to 5. 4 min after PV addition, which is in the same range as the fenestration during mitosis [31] and much faster than nuclear degradation in apoptosis, which takes hours [32]. The kinetic of chromatin escape was dependent upon the number of H1 (Supporting Information, Fig. S2A) but independent upon the viral purification protocol (Supporting Information, Fig. S2B). A mock purification from non-infected cells did not show any chromatin escape (Fig. S2B) suggesting that chromatin escape was viral dose-dependent but not caused by cellular components co-sedimenting with H1. To further exclude that cellular proteins interacting with the parvoviruses were co-purified causing nuclear disintegration we analyzed the purified parvoviruses by silver stain after SDS PAGE. Figure S2C (Supporting Information) shows that all bands detectable by the silver stain also reacted with a polyclonal anti parvovirus H1 antibody, indicating that the preparation after iodixanol gradient centrifugation was free of contaminating cellular proteins. The parvovirus H1 preparation, which was purified via a CsCl gradient, was used for crystallization and the purity analysis was published elsewhere [33]. We observed that incubation of the permeabilized cells with H1 at room temperature instead of 37°C decelerated nuclear disintegration by 5fold (Supporting Information, Fig. S3A), which is in accordance with reduced enzymatic activities at lower temperatures. This observation makes it unlikely that parvovirus H1 has caused nuclear envelope damage by direct physical interaction. In accordance with this conclusion we observed an entire inhibition of H1-mediated nuclear disintegration when the permeabilized cells were energy-depleted (Supporting Information, Fig. S3B). The loss of chromatin requires not only pore formation of the nuclear membranes but also the disassembly of the nuclear lamina as it occurs upon NEBD. We analyzed the disintegration of the lamina in normal rat kidney cells (NRK cells), expressing the enhanced yellow fluorescent protein fused to the lamin B receptor (LBR). LBR localizes at the inner nuclear membrane and anchors the lamina and the heterochromatin to the membrane. Addition of H1 to permeabilized NRK cells – again in the absence of cytosolic factors – led to rapid loss of LBR, which was significantly faster than the loss of PI fluorescence (50% LBR: 5. 0 min, PI: 6. 6 min, Fig. 3A, B). Blebbing of the nuclear envelope, as it occurs in apoptosis, was not observed. Confocal laser scanning microscopy limits the observation to the equatorial section of the nuclei. Thus we next analyzed chromatin distribution after H1 exposure to permeabilized HeLa cells by 3D reconstruction. Due to bleaching, which occurs during multiple scans needed for the reconstruction we restricted our analysis to one time point 15 min after addition of H1. At this time little fluorescence rested in the equatorial sections. Figure 3C confirmed the absence of chromatin in most cells but also showed one cell in which some chromatin stayed attached to that area of nuclear membrane, which was directly in front of the plasma membrane attaching to the cover slip. As HeLa cells are extremely dedifferentiated a polarization-dependent effect was unlikely. Instead we hypothesize that the PV had restricted access to this area of the NE connecting the nucleus with the cover slip-bound region of plasma membrane. This idea is supported by the observation that digitonin permeabilization starts on the plasma membrane accessible to the exterior before progressing to the plasma membrane underneath the nucleus indicating an area of restricted access even for smaller molecules. Further this model is consistent with our observation that H1 caused local NEBD in infection during which a locally increased concentration of PV was observed (Fig. 1B). Having shown that NEBD is cell type-independent we next asked if this phenomenon is also conserved between different PV. We analyzed the canine parvovirus and AAV2 showing that all of them disintegrated the nuclear envelope (Fig. 4A) leading to chromatin release from the nuclei of permeabilized HeLa cells. The kinetic of NEBD were similar although AAV2 was somewhat more efficient than H1 while the canine parvovirus showed a delayed disintegration. As for H1 no soluble cytosolic factors were needed but AAV2-mediated NEBD was limited to capsids, which were exposed to pH 5. 2 (and subsequently neutralized). In fact pH 5. 2 is the characteristic pH of late endosomes from which PVs escape [20] but the acidification of H1 (not shown) had no effect indicating discrete differences between both viruses. To identify the trigger of NEBD we analyzed the contact site for PV at the nuclear envelope. Addition of wheat germ agglutinin, which blocks the attachment of nuclear import receptors to NPCs and which prevents active nuclear import at this concentration [34], did not prevent NEBD (Fig. 5A). This finding is in agreement with observations in Xenopus laevis oocytes in which wheat germ agglutinin has no effect on MVM-mediated pore formation [27]. Cohen et al. concluded that NPCs are not involved in nuclear envelope degradation we considered a direct interaction of the PV capsids with the proteins of the nuclear pore. In fact this hypothesis is consistent with our observation that soluble cytosolic proteins were not required for PV-mediated NEBD. We thus performed coprecipitations of H1 and AAV2 (after acidification) using a purified preparation of Nups (Supporting Information, Fig. S4). The preparation was devoid of importin α and importin β was reduced by 50fold compared to intact cells, which is consistent with the observation that purified nuclei - the first step of Nup preparation - are incompetent for active nuclear import. The changed abundance of the Nups in the preparation compared to intact cells further indicated that the NPCs were dissociated into Nups upon preparation. Despite of the absence of importin α both AAV2 after acidification and H1 precipitated Nups 358,153 and 62 (Fig. 5B). Nup62 gave the strongest band likely due to its higher abundance in NPCs than Nup358 and Nup153 [35]. In fact the strength of the signals corresponded well to those present in the preparation suggesting that there is no preferential interaction of the PV with one of the Nups. However as the antibody used for detection is limited to some FXFG repeat-containing Nups [36] we cannot exclude that other Nups were precipitated. Nonetheless the observation that the Nups were dissociated upon their purification argues against a co-precipitation of the Nups based on PV interaction to one Nup only, which than forms a complex with the other Nups. Furthermore the only known factor interacting with all PV-precipitated Nups would be the complex of importin α andβ, which could bind to the NLS on VP1u. The absence of importin α makes this scenario unlikely and supports direct interaction of the Nups with H1 and AAV2. For getting direct evidence that PV – NPC interactions are required for NEBD we next preloaded the NPCs of permeabilized cells with an excess of hepatitis B virus capsids, which attach to Nup153 without becoming released into the nucleus [37]. The capsids were bound in the presence of a cytosolic lysate, which contains the required nuclear transport factors importin α and β, needed to attach the HBV capsid to the NPC. At the NPC, the transport receptors are dissociated leading to direct attachment to Nup153. Consequently, the capsids localize at the cytoplasmic and nuclear phase of the NPC. After removal of the lysate by washing we added H1 observing that capsid-saturation blocked NEBD nearly entirely (Fig. 5C). Inhibition was specific as a capsid mutant, which fails to bind the transport receptors, thus not interacting with the nuclear pore, did not interfere with NEBD (Fig. 5C). To further exclude NPC-independent interaction with membranes we incubated PV at the same concentration of PV used in permeabilized cells with intact cells together with PI for 15 min, showing that the membrane impermeable PI remained excluded from the cells (Supporting Information Fig. S5). In summary we conclude that PVs need direct attachment to NPCs for membrane degradation. Considering that we added 300 H1 per nucleus, which is not sufficient to saturate the approx. 5000 NPCs per HeLa cells [37] is was surprising that PV cause discontinuities large enough to allow the escape of chromatin. We thus hypothesized that the holes spread as in mitosis [5]. Asking which PV domain causes NEBD, we added two mutants of AAV2 to permeabilized cells: AAV2-ΔVP1 is devoid of VP1 including VP1u and AAV2-ΔVP2, which is devoid of VP2 but which comprises VP1u [38]. NEBD was restricted to the VP1u-comprising mutant (Fig. 6A). The need of VP1u, which normally is hidden in the viral context for NEBD, suggests that the attachment to the nucleoporins causes structural changes including exposure of VP1u. We determined VP1u-exposure using the PLA2 domain, which becomes only active upon externalization for instance after heat treatment and acidification (PV B19; [39]). Incubating purified Nups with H1 increased the PLA2 activity by ∼2. 5 fold (Fig. 6B) indicating exposure of VP1u upon nucleoporins association. AAV2 needed acidification prior to Nup addition in order to exhibit PLA2 activity. This finding corresponds to the acidification-dependent NEBD capacity of AAV2 and supports the need of the structural change for the phenomenon. The PLA2 domain present on VP1u could not only be involved in endosomal escape as it was described recently [18] but also in NEBD. We thus added an AAV2 mutant, in which the catalytic centre of the PLA2 domain was inactivated [38], to permeabilized cells. As shown in figure 6C the mutant caused NEBD as wild type AAV2 only after exposure to acidic pH excluding that PLA2 activity is needed in NEBD. The effect was not as pronounced as for wt AAV2 as the mutant could not be obtained in the same concentration (only 40%). Additional investigations are currently performed for identifying the responsible domain on VP1u. H1 and acidified AAV2 bound at least to Nup358, which localises on the cytosolic face of the NPC, to Nup153 and probably Nup214, which both localize on both sides of the NPC [40], and to Nup62, which is part of the hydrophobic mesh filling nuclear pore [41]. Nup62 was not expected to be involved in NEBD as anti p62-antibodies do not block nuclear accumulation of AAV2 DNA or uptake of fluorescently labelled AAV2 into nuclei in vitro [42]. The external localization of Nups358 and 153 let us expect that VP1u causes local NEBD on the ONM first, followed by destruction of the INM. To test this idea we injected H1 into Xenopus laevis oocytes and analyzed local NEBD by EM. Figure 7A shows that PV H1-induced local membrane disruptions, which were not observed in buffer (MOCK) -injected oocytes. This observation is similar to the fenestration of the ONM recently observed in oocytes after microinjection of the minute virus of mice [27]. Number and size of the membrane breaks increased with the incubation time but their maximal size of ∼190 nm remained similar (Fig. 7B). We hypothesize that the difference between cell lines, which show larger breaks and the oocytes with restricted fenestrations by differences between somatic and germinal cells but we also considered that the nuclei of Xenopus laevis differ in their stability due to massive amounts of actin, which becomes exported upon fertilization [43]. Despite of differences in germinal vesicle breakdown and NEBD both systems are identical in terms of their nuclear pores and nuclear import [44]. We thus used the Xenopus laevis oocytes for analyzing the effect of nuclear H1 microinjection. Figure 7B showed few defects of the NE being in the same range than in MOCK-cytoplasmic injected oocytes. This observation is in agreement with infection during which progeny capsids accumulate inside the nucleus without nuclear disintegration. INM disruption was only observed at those sites where the ONM was disrupted. INM breaks were however much rarer than ONM breaks suggesting that the ONM occurred before the INM was disrupted. The occurrence of breaks in the oocytes took much longer than in permeabilized cells, which could be caused by/reflect the low temperature at which the oocytes have to be incubated. In addition to ONM and INM disruption, chromatin escape also requires lamina disintegration [45]. As shown in Supporting Information, figure S1A, Tx-100 mediated permeabilization of the nuclear membrane is not sufficient to cause even the release of a ∼100 kDa cargo. We thus analyzed the role of cellular enzymes essential for disassembly of the lamin network in mitosis and apoptosis. We inhibited PKC by the broad inhibitor H89 [46] and cdk-1/2 by roscovitine [47]. We used caspase-3 inhibition by zDEVD-fmk (not shown) and zVAD-fmk [48] as a control since it was shown that it is essential for nuclear disintegration by MVM infection [26] where it was demonstrated to be responsible for lamin degradation. The inhibitors we used blocked purified PKC, cdk-1/2 and caspase-3 specifically with regard to the other enzymes (Supporting Information, Fig. S6). Figure 8 shows that the caspase-inhibition in fact blocked the egress of a 100 kDa protein conjugate (M9-BSA), which was imported prior to H1 addition but also chromatin. As the specificity of caspase inhibitors is somewhat limited we further confirmed caspase-3 requirement using UV irradiated cells. Irradiation caused strong caspase-3 activation (Supporting Information, Fig. S7A), which significantly accelerated PV-mediated NEBD (Supporting Information, Fig. S7B). PKC inhibition blocked the release of chromatin as well as the escape of the 100 kDa cargo. The involvement of PKC show thus some homology to the egress of CMV, which causes local PKC-dependent disassembly of the nuclear lamina [29]. In the non-inhibited control cells both M9-BSA and chromatin escape appeared at the same time implying that after onset of fenestration nuclear disintegration proceeds rapidly. This finding further supports the homology with NEBD during mitosis during which such rapid progression was described [5]. Furthermore we observed that cdk-inhibition blocked NEBD as PKC inhibition did, supporting the similarity of PV-mediated NEBD with mitosis in which cdk-1/2 are essential throughout pro-, meta- and anaphase. The observation that cdks and PKC activities were required implies that nuclear disintegration needed a coordinated activity of both enzyme families and that the activity of one could not rescue the missing activity of the other. We next asked if there is a coordinated activation of the enzymes. We permeabilized cells, added H1 in the presence of the different inhibitors and analyzed the activity of i) PKC (α, β, γ, δ, ε, μ, θ, ζ), ii) cdk-2 and iii) caspase-3 in the lysates (Fig. 9A, B, C). We have chosen to analyze cdk-2 as it is needed during earlier steps of mitosis as G1/S phase transition. Its enzymatic activity could have been essential for NEBD as cdk-2 activates cdk-1 [49]. Cdk-1 in turn is required for lamin hyper phosphorylation, needed for lamin depolymerisation [50]. Permeabilization without H1 resulted in decreased activities of all enzymes to ∼50% due to the loss of the soluble cytoplasmic fractions (Fig. 9A, B, C). Adding H1 to the permeabilized cells doubled PKC and cdk-2 activities (Fig. 9A, B), while caspase-3 activity was not significantly altered (Fig. 9C). Consistently, inhibition of PKC by H89 or of cdk-2 by roscovitine did not affect caspase-3 activity. In contrast H89 inhibited not only PKC (Fig. 9A) but also cdk-2 (Fig. 9B) implying that PKC was activated first, which then activated cdk-2. This finding shows that there is also an indirect stimulating effect of PV on cdk-2, which might counteract the inhibitory function of PV as it was shown by biochemical assays using purified cdk-2 and AAV2 and 8 [51]. Roscovitine in turn did not change PKC activity (Fig. 9A) supporting that PKC activation is independent upon cdk-2 activity but is required for cdk-2 activation. Cdk-2 became as well inhibited by caspase-3 inhibition (Fig. 9B), which is in agreement with observations of others who showed a caspase-3 dependent activation of cdk-2 [52]. PKC activity was however not affected by caspase-3 inhibition (Fig. 9A), suggesting that no apoptosis-related PKCδ cleavage occurred. As the detection system for PKC comprised the detection of untypical Ca++-independent PKCs we asked for the impact of Ca++, assuming that Ca++ could have escaped from the lumen between ONM and INM upon membrane disruption. We pretreated HeLa cells with thapsigargin, which depletes Ca++ by inhibiting the endoplasmic reticulum Ca++ ATPase [53]. Figure 9A shows that Ca++ depletion restricted PKC activity in the permeabilized cells to the same extent as H89, supporting that the Ca++-independent proapoptotic PKCδ was not required. The finding is in agreement with observations of others showing that calcium chelator treatment prevents mitotic NEBD in sea urchins [54]. Thapsigargin-treatment also inhibited cdk-2 (Fig. 9B), which is in accordance with the observation that PKC has to become activated for subsequent activation of cdk-2. H89 is a relatively broad inhibitor of PKCs but the Ca++ dependence of the activations suggested that the PKC isoform involved in PV-mediated NEBD is also Ca++ dependent, thus involving PKCα, β and γ. PKCγ is neuron-specific, while PKCβ is involved in lamin phosphorylation [55]. PKCα also causes lamin phosphorylation [56] and its inhibition results in cell cycle arrest [57]. In order to decipher their function in PV-mediated NEBD we added H1 to permeabilized cells, which have been pretreated for 1 h with either LY333531 (an established inhibitor of PKCβ) or with a myristylated pseudo substrate peptide inhibiting PKCα. This inhibitor is considered to be highly specific as it acts as regulatory domain of PKCα suppressing activity of the catalytic domain [58]. Figure 9D showed that inhibition of PKCα strongly inhibited PV-mediated NEBD but that inhibition of PKCβ had only a marginal effect. In view of the results shown before we conclude that three enzymes which are either essential (PKCα and cdk-1/2) or discussed to be involved (caspase-3) in mitosis were also indispensable for PV-mediated NEBD. Further PKC and cdk-2 became activated in a Ca++-dependent manner; the activation of the latter – although Ca++-independent - as a consequence of PKC activation. Consistently, the Ca++-dependent PKCα but not PKCβ was identified as essential for PV-mediated NEBD. In order to link our in vivo infection data with the results obtained in permeabilized cells we used microinjection of H1. By time lapse microscopy we visualized the kinetics and extend of NEBD. For this approach we first co-injected differential fluorescently labelled marker molecules of increasing size alone or together with H1 into the cytoplasm of U2OS cells. Following injection in the absence of H1 a 10 kDa marker molecule (5 nm in diameter) equilibrated rapidly between nucleus and cytoplasm because its small size allows diffusion across the NPC barrier (Fig. 10A). A 40 kDa marker (10 nm in diameter) and a 150 kDa marker (IgG antibodies; 15 nm in diameter) stayed cytoplasmic, which is in agreement with the size exclusion limit of the NPC (Figure 10A and supplemental movie S1). Both 40 kDa and 150 kDa marker rapidly equilibrated in the cytoplasm. Given the linear relation between diameter and diffusion this observation indicates that also H1 (26 nm diameter) can reach the NE within less than a minute. Upon co-injection of ∼100 H1 particles we observed NEBD within minutes after injection indicated by the simultaneous entry of the 40 and 150 kDa marker (Figure 10B). Considering that 100 kDa markers were retained in the nuclei devoid of their nuclear membrane (Supporting Information, Fig. S1A) we conclude that the entry of at least the 150 kDa marker upon microinjection indicate lamin depolymerization as it occurred in permeabilized cells (Fig. 3A, B). The increase in NE permeability occurred at once without specific localization resembling a catastrophic event (supplemental movie S2), which corresponds to the isochronic escape of the 100 kDa cargo and the chromatin in permeabilized cells (Fig. 8). The sudden appearance of NEBD is a characteristic of cdk-1 activation (in the complex with cyclin A2) upon mitosis [59]. The onset of NE permeability was observed on average after 290 sec (95% CI from 128 to 395 sec. , 31 cells) and both markers were in equilibrium in cytoplasm and in nucleus supporting the idea of a massive increase of permeability. In fact this timescale was thus similar to that observed in permeabilized cells. Entry was simultaneous for the 40 kDa and 150 kDa markers, which is in accordance to the simultaneous egress of the 100 kDa protein and chromatin in permeabilized cells (Figure 8). Considering that only 11% of the infected cells showed NEBD while the nuclei of all permeabilized or microinjected cells were disintegrated we thus conclude that limited endosomal escape of PV accounted for the restricted number in infection as it was reported previously [60]. The importance of Ca++ shown by Thapsigargin-treatment (Fig. 9) and the need of the Ca++-dependent PKCα let us ask if we could trigger NEBD directly by Ca++ circumventing the Ca++ release from the nuclear envelope by PV. We microinjected Ca++ to a final concentration of 7 mM, which is the concentration between the inner and out leaflet of the NE [61]. The intracellular concentration upon microinjection was determined by comparing the fluorescence of the co-injected fluorescent markers (40 kDa Dextran and IgG antibodies) with a standard dilution series. Measuring the permeability by time-lapse microscopy as described before we observed a sudden onset of nuclear influx of both markers 200 sec after microinjection (Fig. 10C and supplemental movie S3). This finding is in agreement with observations of others showing that Ca++ is sufficient for driving mitosis in early mouse embryos [62]. The observation that NEBD occurred around the same time than upon PV microinjection supports that PV caused Ca++ release, which is the initial cellular trigger for starting the cascade of PKC and cdk activation leading to NEBD. In summary our data give evidence for a unique virus-mediated pathway that causes NEBD (see scheme Fig. 11). NEBD was observed for different PV and in cells ranging from human to Xenopus laevis implying an evolutionary well conserved phenomenon. Our model of PV host-interaction starts with the release of the viruses from the microtubules in the nuclear periphery, followed by (a) attachment to the NPC directly, which (b) subsequently causes exposure of VP1u. A yet undefined domain of VP1u then (c) permeabilizes the nuclear membranes leading to Ca++ efflux. (d) Ca++ activates nuclear PKCα, which phosphorylates lamins [56] and activates cdk-2, which becomes further activated by caspase-3 as described recently [63]. Activated cdk-2 – a key element of entry into mitosis - possibly leads to cdk-1 activation and (e) lamin A/C hyper phosphorylation [64]. (f) Lamin (hyper) phosphorylation than leads to lamin depolymerisation allowing entry and exit of large cargos from the nucleus by diffusion. At least in somatic cells the activation cascade must spread within the nucleus explaining why nuclear permeability increased suddenly, which is also a characteristic of mitosis. The PV-mediated NEBD apparently by-pass the early mechanisms of NEBD during mitosis entry, explaining the observed differences to mitosis. In consequence we have not observed chromatin condensation taking place during prophase. Although we have not investigated tubulin polymerization causing permeabilizing the nuclear membrane [5] and leading to Ca++ efflux, we assume that the pool of tubulin at least in permeabilized cells does not allow their formation. Instead PV cause Ca++ release directly from the space between INM and ONM as it occurs directly before NEBD [6]. The evasion of PV thus allows direct entry into the mitotic pathways at a later stage. Despite of significant differences to NEBD in mitosis we conclude that PV use the mitotic pathway for causing nuclear envelope disintegration. The homologies comprise not only the need of PKCα and cdk-1/2, which are implied in lamin depolymerization by phosphorylation [8] and cell cycle progression [65], [66] but also the sudden onset and rapid progression [5]. The need of caspase-3 for PV-mediated NEBD seems surprising in this context but its implication in mitosis is controversial [9]–[11]. Evidently there are striking differences to NEBD in apoptosis not only in that it takes muck longer [32] but also as apoptosis requires PKCδ [12] and not PKCα. Furthermore, we never observed DNA fragmentation or the characteristic DNA patches in the nuclear periphery. Assuming that PV use the mitotic pathways it was most surprising that virus-mediated NEBD was driven solely by nuclear factors. However although our study does not entirely explain the molecular mechanisms and activations for NEBD it reveals that pathways mediating NEBD can be uncoupled from the various checkpoints of mitosis. We thus assume that PV provide a new entry point to unravel the functions of various proteins upon the different stages of mitosis.
NBK-324, U2OS and HeLa cells were grown in DMEM/5% FCS at 37°C. EYFP lamin B receptor NRK cells were grown in DMEM/8% FCS. Viral stocks were generated by infecting NBK-324 cells with 0. 1 pfu viruses per cell and harvested after 2. 5 days. Virus stocks were extracted from infected cells by cell lyses in 50 mM Tris-HCl, 0. 5 mM EDTA and 5 cycles of freezing and thawing. Viruses were purified by iodixanol step gradient centrifugation [67]. To analyze the effect of acidification on parvoviruses, the viruses were incubated for 10 min in sodium acetate buffer pH 5. 2, followed by neutralisation with 0. 5 M Tris base. HeLa cells were seeded on collagenized cover slips and grown overnight prior to infection with 1000 H1/cell. Four h post infection the cells were washed with DMEM and fixed with 3% paraformaldehyde (Merck) /PBS at 4°C for 2 h. Cells were washed twice in PBS, permeabilized with 0. 1% Triton X-100/PBS for 10 min at RT, washed 3× with PBS and incubated in blocking solution (0. 5% BSA/5% goat serum/PBS) for 15 min at 37°C. Primary antibodies were diluted in blocking solution (rabbit polyclonal anti VP1/VP2 1∶100; mAb414 (Hiss Diagnostic) 1∶500) and added to the cells for 90 min at 37°C in a humidified chamber. Cells were washed in 4× in PBS, before 1∶200 diluted secondary antibodies (FITC-conjugated goat anti-rabbit 1∶200, Cy5-conjugated goat anti-mouse antibody (Jackson Immuno Research) with 0. 2 µg/ml propidium iodide was added for 45 min at 37°C. After 4 washes with PBS, the cover slips were mounted on glass slides using 50 mg/ml DABCO/Moviol. Real time microscopy was performed using a Leica SP-5 confocal microscope equipped with 3 internal PMT and 1 PMT trans, using a HCX Plan Apo CS 20X multi-immersion NA 0. 70 lens at 37°C in transport buffer (200 mM Hepes pH 7. 3,20 mM magnesium acetate, 100 mM potassium acetate, 50 mM sodium acetate, 10 mM EGTA) or life-cell imaging media (Invitogen). Images were acquired using the standard setting of the microscope and a pinhole size of 1. 0 and the LEICA acquisition software LAS AF. In the experiments with permeabilized cells images were taken at the indicated time points and quantification of the stained nuclei was done using Image J followed by analysis using excel data sheets. In microinjection experiments of somatic cells the images were taken at a frame rate of 1 frame per 15 seconds for a total of 40 frames and mounted in image J. Correction for photo bleaching was performed using Image J with a plugin described elsewhere (http: //fiji. sc/wiki/index. php/Bleach_Correction). Microscopy of fixed cells was done at RT using an HCX Plan Apo CS 40X oil NA 1. 25 lens. 3D reconstruction images were captured by confocal microscopy and reconstitute by Imaris software. HeLa cells were grown on collagenized cover slips overnight at 37°C. Cells were washed and 300 H1 viruses was added for 15 min at 37°C in medium followed by washing. Cells were stained with propidium iodide for 5 min at RT and mounted. 1×105 cells were grown on collagenized 12 mm cover slips, washed twice with serum free media, before serum-free medium/20 µg/ml digitonin/1 µg/ml propidium iodide was added. After incubation for 5 min at 37°C cells were washed with ice-cold transport buffer. The cover slips were placed in a heating device and the propidium iodide stain was used to focus the samples. The buffer was than replaced by 37°C pre-warmed virus in transport buffer (20 mM Hepes [pH 7. 3], 2 mM Mg-acetate, 110 mM K-acetate, 5 mM Na-acetate, 1 mM EGTA) /2 mM DTT (100 µl) or by new pre-warmed transport buffer/2 mM DTT (negative controls). Modifications are indicated in the individual experiments. When cargos were imported into the nuclei or when the nuclei were preincubated with hepatitis B virus capsids prior to addition of parvoviruses the washed, permeabilized cells were subjected to rabbit reticulocyte lysate (21 mg/ml, Promega) in transport buffer/2 mM DTT/20 U/ml creatine phosphokinase/5 mM creatine phosphate, containing 150 µg/µl M9-Alexa 647-BSA, or 150 µg/µl NLS-Alexa 594-BSA or 1200 ng capsids for 15 min at 37°C. After 3 washes with transport buffer, parvoviruses were added as described. Oocytes were microinjected and prepared for thin sectioning EM as previously described [27]. Oocytes were injected with about 100 nl of purified H1 (2. 13×109 pfu. /ml, or 2. 17×1012 genomes/ml) in the cytoplasm at the transitional zone between the animal and vegetal poles. As control experiments, oocytes were mock injected with 100 nl Tris-EDTA buffer (TE: 50 mM Tris, 0. 5 mM EDTA, pH 8. 7). Oocytes were then incubated at RT in modified Barth' s saline buffer (MBS: 88 mM NaCl, 1 mM KCl, 0. 82 mM MgSO4,0. 33 mM Ca (NO3) 2,0. 41 mM CaCl2,10 mM HEPES, pH 7. 5). After microinjection and incubation at RT, oocytes were fixed o. n. at 4°C with 2% glutaraldehyde in MBS. Oocytes were washed with MBS and their animal poles were dissected and fixed with 2% glutaraldehyde in low-salt buffer (LSB: 1 mM KCl, 0. 5 mM MgCl2,10 mM Hepes, pH 7. 5) for 1 h at RT. Dissected oocytes were washed with LSB, embedded in 2% low melting agarose and post-fixed with 1% OsO4. Fixed oocytes were sequentially dehydrated in ethanol and embedded in Epon 812 (Fluka) as described elsewhere [68]. Following embedding, 50-nm thin sections through the nuclear envelope (NE) were cut and placed on phalloidin/carbon coated copper EM grids, stained with 2% uranyl acetate for 30 min and 2% lead citrate for 5 min, and viewed with a Hitachi-7600 transmission electron microscope. Nuclear envelope disruption was quantified by measuring the length of outer nuclear membrane (ONM) and inner nuclear membrane (INM) disruptions from EM cross-sections of NE using Carnoy image analysis software (Biovolution). Bar graphs represent the average length of the ONM breaks, or the average proportion of NE damage calculated as the length of the ONM breaks divided by the total length of the ONM from electron micrographs. Microinjections into U2OS cells were performed using an Eppendorff FemtoJet microinjection device coupled to a LEICA SP5 confocal microscope. Injection solutions contained Texas Red coupled Dextran (10 kDa, 2 mg/ml final concentration), FITC coupled Dextran (40 kDa, 1 mg/ml final concentration) and Alexa647 coupled secondary mouse antibodies (150 kDa, 0. 4 mg/ml final concentration) all diluted in transport buffer. Seventy-five % dense HeLa cells from ten 16 cm dishes were treated with 4. 2 mM Cytochalasin B/DMEM for 30 min at 37°C. Cells were trypsinized and resuspended in 10 ml PBS. Cells were sedimented at 200× g for 10 min at 4°C. Washing and centrifugation steps were repeated 3 times. The pellet was resuspended in 5 ml nuclei buffer (10 mM PIPES (pH 7. 4), 10 mM KCl, 2 mM MgCl2,1 mM DTT) and the cells were sedimented at 200× g for 10 min at 4°C. The cells were resuspended in 10 volumes nuclei buffer/10 µM Cytochalasin B, incubate for 30 min on ice and homogenized by 30 strokes on ice. The sample was loaded on 4 volumes of 30% (w/w) sucrose/nuclei buffer and centrifuged at 800× g and 4°C for 10 min. The pellet was resuspended in 500 µl nuclei buffer and the centrifugation step was repeated. The nuclei-containing sediment was washed with 3 ml of nuclei buffer and the nucleoporins were isolated according to [69]. The nucleoporins were quantified by Bradford assay (Biorad, Germany). HeLa cells were seeded on collagenized cover slips o. n. at 37°C. Cells were washed 2 times with serum free medium followed by UV irradiation for 30 sec (9,000 µW/cm2, λ = 312 nm). One hundred µl of Phiphilux (OncoImmunin) and 10 µl of foetal calf serum were added to the cells for 1 h at 37°C. Cells were washed with PBS and images were taken by confocal laser scanning microscope. | Parvoviruses are small non-enveloped DNA viruses successfully used in gene therapy. Their nuclear replication requires transit of the nuclear envelope. Analyzing the interaction between parvoviruses and the nucleus, we showed that despite their small size, they did not traverse the nuclear pore, but attached directly to proteins of the nuclear pore complex. We observed that this binding induced structural changes of the parvoviruses and that the structural rearrangement was essential for triggering a signal cascade resulting in disintegration of the nuclear envelope. Physiologically such nuclear envelope breakdown occurs late during prophase of mitosis. Our finding that the parvovirus-mediated nuclear envelope breakdown also occurred in the absence of soluble cytosolic factors allowed us to decipher the intra nuclear pathways involved in nuclear envelope destabilization. Consistently with the physiological disintegration we found that key enzymes of mitosis were essential and we further identified Ca++ as the initial trigger. Thus our data not only show a unique pathway of how a DNA virus interacts with the nucleus but also describes a virus-based system allowing the first time to analyze selectively the intranuclear pathways leading to nuclear envelope disintegration. | Abstract
Introduction
Results and Discussion
Materials and Methods | 2013 | Parvoviruses Cause Nuclear Envelope Breakdown by Activating Key Enzymes of Mitosis | 11,721 | 276 |
|
Personal exome and genome sequencing provides access to loss-of-function and rare deleterious alleles whose interpretation is expected to provide insight into individual disease burden. However, for each allele, accurate interpretation of its effect will depend on both its penetrance and the trait' s expressivity. In this regard, an important factor that can modify the effect of a pathogenic coding allele is its level of expression; a factor which itself characteristically changes across tissues. To better inform the degree to which pathogenic alleles can be modified by expression level across multiple tissues, we have conducted exome, RNA and deep, targeted allele-specific expression (ASE) sequencing in ten tissues obtained from a single individual. By combining such data, we report the impact of rare and common loss-of-function variants on allelic expression exposing stronger allelic bias for rare stop-gain variants and informing the extent to which rare deleterious coding alleles are consistently expressed across tissues. This study demonstrates the potential importance of transcriptome data to the interpretation of pathogenic protein-coding variants.
Recent genome sequencing studies have highlighted that healthy individuals carry multiple loss-of-function and rare deleterious variants whose interpretation is expected to inform individual disease risk and facilitate precision medicine [1]–[3]. However, accurate interpretation of these variants remains a considerable challenge as phenotypic effects remain difficult to predict. Furthermore, even when a specific function can be ascribed to a genetic variant, the variable penetrance and trait expressivity of genetic variants may yield important differences. In this respect, an important modifier of a coding allele' s effect is its level of expression (Figure 1). This type of modification is likely to have considerable impact on interpretation of coding variant effects as genetic analyses of gene expression have reported that allele specific expression (ASE) influences at least 30% of genes for any given cell type [4], [5] and variability in allelic expression of pathogenic coding alleles has already been implicated in contributing to clinical variability for several diseases [6]–[10]. However, the degree to which deleterious and loss-of-function coding variants, routinely found through individual exome and genome sequencing, are allelically-expressed across multiple tissue types remains unexplored. In this study we investigated patterns of gene expression and ASE for rare deleterious and loss-of-function variants across multiple tissues using both RNA-Seq and mmPCR-Seq, a targeted and high-resolution sequencing assay for measuring allelic ratios [11]. A major advantage of mmPCR-Seq is that it uncouples a gene' s expression level, which can characteristically vary across tissues, from the power to measure allele-specific expression. Using this approach, we obtain 1000s of reads per heterozygous site per tissue to robustly quantify ASE. By comparing patterns of gene expression to allelic expression, we observed higher variability of allelic expression between tissues suggesting that expression level alone may be insufficient to predict the exposure of a damaging allele. Furthermore, we report patterns of ASE across tissues for both rare deleterious and loss-of-function protein-coding variants. These results demonstrate the extent to which regulatory variation can modify the functional impact of protein-coding variation across tissues, as well as the importance of using ASE for the interpretation of heterozygous variants in clinical sequencing analyses.
To map patterns of ASE for deleterious and loss-of-function coding variants, we sequenced the exome from two tissues (frontal lobe and small intestine) and RNA from ten tissues (cerebellum, frontal lobe, pancreas, stomach, small intestine, colon, heart, lungs, liver, and skeletal muscle) from a single individual. From the exome data, we identified 51,875 SNPs, of which 45,058 had consistent genotypes across tissues and were defined as “high-confidence” variants (Table S1). We identified 2,767 high-confidence variants that are private and not previously found in dbSNP [12], the 1000 Genomes Project [13], or the NHLBI Exome Sequencing Project (ESP) [14] (Table S2). Of these, 91 were heterozygous derived nonsynonymous variants classified by Sift [15] and Polyphen [16] as “damaging” and “deleterious”, respectively. Complementing these variants, we identified 106 SNPs that introduce premature stop-codons in exons, of which 75 SNPs were predicted to cause complete loss of function of all known transcripts using previously described prediction methods [1]. We performed RNA sequencing (RNA-Seq) for each tissue (Figures S1, S2, S3) and intersected this data with high-confidence heterozygous variants to identify ASE patterns (Figure S4). ASE was determined on a per-heterozygote per-tissue basis using a binomial test where p is the empirical probability that a reference allele maps to the genome compared to a non-reference allele across all sites (Figure S5). Quality control filtering (by depth, p-value, bi-allelic expression and intragenic location) was performed to identify high-confident ASE sites across all tissues (Figure S6 and Table S3). The detailed method is available at http: //montgomerylab. stanford. edu/resources. html. The measurements of ASE by RNA-Seq are influenced by the depth of coverage of a gene in the assayed tissue [17], introducing challenges for ASE comparisons across tissues where genes are characteristically differentially expressed. To more accurately quantify ASE, we also applied our recently developed method that couples microfluidics-based multiplex PCR and next generation sequencing (mmPCR-Seq) [11]. We applied this technique to 74 deleterious, 50 nonsense and 205 control variants (Figure S7). Seventeen deleterious and 25 nonsense sites were excluded because they showed no evidence of expression in any of the ten tissues. For each tissue, we performed two technical replicates and mapped the merged sequence reads since we target-sequenced specific loci (Figure S8, S9). We applied the same pipeline and filters to detect ASE as those used for RNA-Seq. We further evaluated the correlation of effect size between technical replicates and observed high technical reproducibility (Figure S10). The small intestine and skeletal muscle have the greatest reproducibility (Pearson Correlation, R>0. 93). The tissue with the lowest reproducibility is the pancreas (R>0. 70), which contains a high concentration of nucleases and other enzymes that can degrade RNA. The variability of effect size between the replicates was also quantified for each tissue at varying read depths (Figure S11). As expected, sites with higher read depths have less variability between replicates. With the exception of the pancreas and frontal lobe, which are two tissues known to have low RNA quality post-mortem. Regardless, the variability of allelic ratios between replicates was well below 0. 2 across all samples read depths. For tested sites, mmPCR-Seq provided greater depth and power to detect ASE and in many cases facilitated estimates for sites immeasurable without extreme RNA-seq coverage (Figures S12, S13). For instance, for 598 measurements which had no reads with RNA-Seq, we obtained an average of 2639 reads for mmPCR-Seq. Furthermore, only 73 measurements had greater than 100 reads for RNA-Seq compared to 817 for mmPCR-Seq. We next examined the sharing of gene expression and allelic effects across different tissues. Shared patterns of gene expression are detectable for tissues with shared functional roles or embryonic origins (Figure 2A, inset). For instance, the small intestine and colon, which are both digestive system organs derived from the endoderm layers, have a high degree of correlation (Spearman Correlation, R = 0. 92). Likewise, the frontal lobe and cerebellum, which are both neural tissues derived from the ectoderm, have a high degree of shared gene expression (R = 0. 91). To test the degree of correlation of allelic expression across tissues, we measured concordance of allelic ratios between pairwise tissues using the high-depth mmPCR-Seq data. Here, allelic ratios are defined as the ratio of the non-reference allele to the sum of the non-reference allele and the reference allele. We observed that the concordance of ASE between tissues does not as strongly reflect the relationships seen for shared gene expression or shared embryonic origin (Figure 2B). The range of pairwise tissue correlation for allelic effects ranges between 0. 46 and 0. 80, with the small intestine and colon having the most similarity (R = 0. 80). We also compared in detail the pairwise correlation coefficients for expression and allelic ratios for tissue pairs of highly similar embryonic origin (Figure S14). We compared two neural tissues (frontal lobe and cerebellum) both derived the ectoderm and two intestinal tissues (small intestine and colon) both derived from the mesoderm. Irrespective of read depth and sequencing technology, the correlation of expression for tissues is consistently greater than the correlation of allelic effects across tissues. This observation suggests that allelic effects exhibit more variability than gene expression across tissues. We also investigated the sharing of monoallelic expression across tissues (Figure 1). We identified five genes (NDN, MAP2K3, FRG1B, IGSF3, and DUSP22) that showed monoallelic expression across all testable tissues (N≥5) in the RNA-Seq data. Two of these genes were mono-allelically expressed across all ten tissues: NDN, which is a known maternally imprinted gene [18], and MAP2K3, which has known allele-specific expression bias [19]. For all five genes, the same allele was mono-allelically expressed in all testable tissues suggesting that these genes are not imprinted in a tissue-specific manner. The majority of sites tested by mmPCR-Seq have equal expression of both alleles, as expected. However, many sites exhibit consistent or variable allelic patterns across different tissues (Figure S15). By comparing the mean and variance of allelic ratios as quantified through mmPCR-Seq across tissues, we stratified sites into those that exhibited no ASE, shared ASE and variable ASE across tissues. Due to the inherent nature of the binomial test, minor deviations from equal allelic expression will appear significant with high read coverage and therefore p-value significance alone is not sufficient for distinguishing between these classes. Therefore, we also took effect size (allelic ratios) into account when classifying sites as ASE. However, the definition of what constitutes a biologically important allelic effect is not easily discernable; therefore, to distinguish between each group, we accounted for previously reported definitions of ASE [20], [21] and applied cutoffs based on the reproducibility of both the allelic ratio and its variance across replicates (Figure S16). Variants were classified as non-ASE sites if the allelic expression was balanced (mean allelic ratio = 0. 5+/−0. 15) and if there was low variance (σ2<0. 2) of the allelic ratios for all tissues tested. Variants were classified as shared ASE sites if they had a significant p-value (p<0. 01), an imbalanced mean allelic ratio (0. 35<mean allelic ratio <0. 65), and non-variable allelic ratios (σ2<0. 2) across all tissues. Lastly, variants were classified as variable (tissue-specific) ASE sites if they had a significant p-value (p<0. 01) and variable allelic ratios (σ2>0. 2) across tissues. The reproducibility of the groups between replicates was tested at varying allelic ratio and variance cut-offs (Figure S16) and was also assessed when the pancreas and frontal lobe, two tissues that had high variability between replicates, were removed (Figure S11). The concordance between replicates increases as the variance cut-off increases and reaches a plateau of ∼95% at a variance of 0. 2. Since the greatest reproducibility is observed when the ASE cutoff is <0. 35 or >0. 65, the variance cutoff is 0. 2, and the pancreas is removed, these cut-offs were chosen for Figure 3. Using these cut-offs, the reproducibility between replicates for the three groups (non-ASE, shared ASE and variable ASE) is 93. 3%. The reproducibility between replicates for the classification of non-ASE and ASE (shared ASE plus variable ASE) is 95. 7%. In total, for sites tested with mmPCR-Seq, 172 showed no ASE across tissues, 52 showed shared ASE, and 8 showed variable ASE (Figure 3A). These proportions are similar to those obtained with RNA-Seq (Figure S17). We then tested if sites exhibiting shared or variable ASE are more likely to be deleterious sites compared to sites exhibiting no ASE. Of the sites exhibiting no ASE, only 25. 0% are deleterious. Comparatively, we find no significant enrichment in deleteriousness among sites which exhibit variable ASE compared to non-ASE sites (p = 0. 423, Fisher' s exact test; not significant); however, a significantly higher proportion of shared ASE sites (42. 3%) are deleterious compared to non-ASE sites (p = 0. 022; Fisher' s exact test). Next, we investigated the relationship between ASE effect sizes and direction of effect across tissues. Figure S15 highlights the range of effect sizes and directions of effect seen across tissues. By focusing on the range of allelic ratios for variants tested in three or more tissues, we further reviewed the distribution of minimum and maximum allelic ratios observed across all tested tissues (Figure S18). As expected, most sites have an allelic ratio around 0. 5, and imbalanced loci show similar direction of effect. Interestingly, several sites exhibit opposing directions of effect in different tissues. For example, heterozygous sites in genes PCDHA13, SCRIB, and PDE4DIP have a major flip in direction of effect from an alternate allele ratio less than 0. 2 to an alternate allele ratio greater than 0. 8 across tissues. Four additional heterozygous sites have a large directional flip from an alternate allele ratio less than 0. 4 to greater than 0. 8, and five more heterozygous sites have a directional flip from an alternate allele ratio less than 0. 2 to greater than 0. 6. To determine if gene expression level informed allelic expression level, we investigated the relationship between gene expression and allelic expression level as measured by mmPCR-Seq (Figure S19). As expected, due to the nature of mmPCR-Seq, no general pattern between absolute expression levels and ASE was observed. Four sites (circled in Figure S19) did have noticeably lower non-reference allele ratios and lower gene expression levels in the pancreas, stomach and lung; however these outliers were not enriched in any variant class and did not influence distinction of variable versus shared ASE. By focusing on patterns of ASE for rare deleterious variants in this individual, we identified 40 sites corresponding to 40 unique genes which were quantified by mmPCR-Seq across three or more tissues. Of these genes 28 exhibited no ASE across tissues, 11 exhibited shared patterns of ASE across tissues and 1 exhibited variable ASE (Figure 3B; Figure S20). We next investigated if genes with different patterns of ASE have relevant disease associations using the Online Mendelian Inheritance in Man (OMIM) database of heritable diseases (Table S4) [22]. Although the OMIM database a limited catalog of genomic variants, OMIM variants serve as examples of the pathogenic consequences of deleterious alleles. Among those that exhibit shared ASE is the FMO3 gene, which encodes a monooxygenase enzyme responsible for hepatic metabolism and whose deficiency causes the rare Mendelian disorder trimethylaminuria that is manifested in a range of phenotypes (OMIM 602079; Figure S20) [23], [24]. Here, the shared allelic effect is detectable in all tissues, but the strongest effect against the deleterious allele is detected in the liver (non-ref to ref allelic ratio = 0. 16; Figure S20). In contrast, no ASE patterns are observed for a deleterious SNP located in the gene encoding a cryopyrin (NLRP3), which is associated with the Mendelian disease Muckle-Wells Syndrome (OMIM 191900) and associated with inflammasome function and immune responses. The single deleterious site that demonstrates variable ASE is a gene encoding a protocadherin (PCDHA13). In the skeletal muscle and heart, the deleterious allele exhibits greater expression than the normal allele, but in the liver and colon the deleterious alleles exhibits less expression. Of interest, PCDHA13, which is known to play a critical role in establishing specific cell-cell connections in the brain, shows no strong patterns of ASE in the two disease-relevant neural tissues, frontal lobe and cerebellum. While the consequences of allelic expression of this individual' s deleterious alleles are unknown, different patterns of allelic expression across tissues highlight the potential importance of testing multiple tissues to better elucidate the functional context of rare, deleterious alleles. Loss-of-function alleles that introduce premature stop codons have been identified to exhibit patterns of allelic expression indicating nonsense-mediated decay (NMD) [1]. We sought to test the extent of this impact across different tissues. Indeed, comparison of ASE data using mmPCR-Seq for nonsense (stop-gained) and control sites indicates considerable reduction in the expression of the nonsense allele across all tissues (Figure 4A and Table S5). We also observed lowered expression of rare, deleterious alleles at heterozygous sites compared to control sites (p<0. 05, student' s t-test). This observation has been previously reported in a single cell-type, with a possible explanation for this phenomenon being that lowly-expressed alleles can better tolerate the fitness impact of deleterious protein-coding alleles [25], [26]. Furthermore, we identified that rare (MAF<5%) nonsense alleles exhibited even stronger evidence of nonsense-mediated decay than common alleles (Figure 4B). To ensure that genotype errors and mappability did not affect this observation, we compared RNA allelic bias to DNA allelic bias from exome-sequencing. Nonsense variants were removed from the analysis if the alternative allelic ratio was below 0. 2 in both tissues. This filtration step ensures that genotyping and mappability of non-reference variants did not influence our observation that rare nonsense variants have decreased allelic expression compared to common nonsense variants. This observation suggests that haplotypes that harbor rare nonsense variants are either considerably unlikely to be expressed or altered transcripts are being efficiently degraded by the NMD machinery. In conclusion, despite the feasibility of sequencing individual genomes, the functional impact of potentially pathogenic protein-coding variants remains difficult to ascertain by DNA sequencing or computational prediction methods alone. The incorporation of transcriptome data can enhance the interpretation of such variants by providing insight into their patterns of ASE. We demonstrate the advantage of ASE for interpretation of pathogenic protein-coding allele by generated high resolution measurements of ASE for these variants across multiple tissues. Such data enables us to identify the extent to which these alleles are modified by regulatory effects and the extent to which this effect is detectable across tissues. We highlight as many as a 1/3 of all deleterious alleles are imbalanced and that nonsense alleles show characteristic and consistently lower expression across multiple tissues. Ultimately, by coupling interpretation of personal genomes with their corresponding transcriptomes, these results highlight that it may be possible to better understand the impact of pathogenic protein-coding variants within different tissues of an individual.
In order to investigate the differential allelic effects of divergent tissues in a single individual, we obtained the genomic DNA and RNA for ten somatic tissues (cerebellum, frontal lobe, pancreas, stomach, small intestine, colon, heart, lungs, liver, and skeletal muscle) from Biochain Institute, Inc (Newark, CA, USA). The samples were collected post-mortem from a healthy 25-year-old male with no significant medical history. Genomic DNA from the frontal lobe and small intestine were prepared for exome sequencing. The enrichment of targeted regions (consensus coding sequence definition of exons and flanking introns, ∼50 Mb) was performed using the Agilent SureSelect Human All Exon 50 Mb Kit (Agilent Technologies, Santa Clara, CA, USA) following the manufacturer' s recommended protocol. Paired-end libraries were constructed using the Illumina Paired End Sample Prep Kit following the manufacturer' s instructions and sequencing was carried out using the Illumina HiSeq 2000 platform (Illumina, San Diego, CA, USA). Exome sequence data was processed through a pipeline based on Picard (http: //picard. sourceforge. net/) with base quality score recalibration and local realignment at known indels and BWA [27], for mapping reads to the human reference genome (build hg19). GATK version v2. 3-13 [28] was used for SNP calling, with the default filters, and the additional parameters: -T UnifiedGenotyper; –downsample_to_coverage 75; –genotype_likelihoods_model BOTH; -contamination 0. 0; -nct 1. For ASE detection (described below), we filtered for heterozygous variants that were present in both the frontal lobe and small intestine. Paired-end RNA-Seq libraries were prepared using the Illumina TruSeq RNA Sample Preparation kit. PolyA+ RNA was isolated using Sera-Mag oligo (dT) beads (Thermo) and fragmented with the Ambion Fragmentation Reagents kit. Complementary DNA (cDNA) synthesis, end repair, A-base addition and ligation of the Illumina-indexed adaptors were performed according to Illumina' s protocol. Each sample was barcoded and all samples were sequenced on one lane of the Illumina HiSeq 2000 platform (2×100-nt read length). In total, we obtained 13. 3±3. 7 (mean ± SD) million paired end reads per sample. We assessed the sequence quality using the publicly available software FastQC. For each sample, we examined per-base quality scores across the length of the reads to ensure that >95% of the reads had >Q60 for bases 1–100. Reads were mapped by TopHat (version 2. 0. 0) to the known transcriptome (-G option; Gencode version 7 annotations) the human reference genome (hg19) using default parameters [29]. Cufflinks (version 2. 0. 2) was used to quantify gene expression for known transcripts (-G option; Gencode version 7 annotations) using the default parameters [30]. To quantify allele-specific expression at lowly expressed site, we applied a high-throughput method that couples microfluidics-based multiplex PCR and deep sequencing (mmPCR-Seq) [11]. We designed primers and applied this technique to 74 deleterious nonsynonymous variants, 50 nonsense variants, and 205 control variants. The control sites are common (MAF>0. 05), non-deleterious variants. First, multiplexed PCR reactions were carried out using the Fluidigm Access Array for each sample. Then, the PCR products were indexed using barcoded adaptor primers via a single PCR reaction for each tissue sample. All indexed samples were pooled and purified using a Qiagen RNeasy. Six picomoles were loaded into one lane of an Illumina MiSeq for deep sequencing. The sequence reads were mapped to the human reference genome (hg19) using the Spliced Transcripts Alignment to a Reference (STAR, version 3. 2) aligner [31]. Since we targeted together specific heterozygous sites in the genome, the default parameters were modified (minimum score and match filters lowered from 0. 66 to 0. 3) to increase the number of mapped reads. Allele-specific expression was determined on a per-heterozygote-site per-tissue basis using the pipeline depicted in Figure S4 and available online (http: //montgomerylab. stanford. edu/resources. html). First, mapped reads were sorted using the Samtools (version 0. 1. 7) [32]. Next, Samtools mpileup was used to call variants from the aligned reads using a list of known heterozygous sites from the individual. Heterozygous sites with a base quality score (MAQ) below 10, individual allele read depth below 5 and a total (both alleles) read depth below 20 were filtered out. Next, we calculated the reference to non-reference allele mapping ratio for each tissue. To test for ASE, we performed a binomial statistical test for each heterozygous site in each tissue modifying p to be the empirical probability of observing a reference versus non-reference allele across all sites. A significance cut-off of 0. 05 and 0. 01 were used for the RNA-Seq and mmPCR-Seq data, respectively. The raw mmPCR-seq data has been submitted to the NCBI Gene Expression Omnibus (GEO) (http: //www. ncbi. nlm. nih. gov/geo/) under accession number GSE51769. The code for ASE detection pipeline can be found online (http: //montgomerylab. stanford. edu/resources. html). | Gene expression is a fundamental cellular process that contributes to phenotypic diversity. Gene expression can vary between alleles of an individual through differences in genomic imprinting or cis-acting regulatory variation. Distinguishing allelic activity is important for informing the abundance of altered mRNA and protein products. Advances in sequencing technologies allow us to quantify patterns of allele-specific expression (ASE) in different individuals and cell-types. Previous studies have identified patterns of ASE across human populations for single cell-types; however the degree of tissue-specificity of ASE has not been deeply characterized. In this study, we compare patterns of ASE across multiple tissues from a single individual using whole transcriptome sequencing (RNA-Seq) and a targeted, high-resolution assay (mmPCR-Seq). We detect patterns of ASE for rare deleterious and loss-of-function protein-coding variants, informing the frequency at which allelic expression could modify the functional impact of personal deleterious protein-coding across tissues. We demonstrate that these interactions occur for one third of such variants however large direction flips in allelic expression are infrequent. | Abstract
Introduction
Results and Discussion
Materials and Methods | sequencing techniques
genome sequencing
genome expression analysis
genomics
functional genomics
genome analysis
gene expression
transcriptome analysis
genetics
biology and life sciences
molecular biology techniques
computational biology
molecular biology | 2014 | Allelic Expression of Deleterious Protein-Coding Variants across Human Tissues | 6,116 | 260 |
Atrial fibrillation (AF) is the most frequent form of arrhythmia occurring in the industrialized world. Because of its complex nature, each identified form of AF requires specialized treatment. Thus, an in-depth understanding of the bases of these arrhythmias is essential for therapeutic development. A variety of experimental studies aimed at understanding the mechanisms of AF are performed using primary cultures of neonatal rat atrial cardiomyocytes (NRAMs). Previously, we have shown that the distinct advantage of NRAM cultures is that they allow standardized, systematic, robust re-entry induction in the presence of a constitutively-active acetylcholine-mediated K+ current (IKACh-c). Experimental studies dedicated to mechanistic explorations of AF, using these cultures, often use computer models for detailed electrophysiological investigations. However, currently, no mathematical model for NRAMs is available. Therefore, in the present study we propose the first model for the action potential (AP) of a NRAM with constitutively-active acetylcholine-mediated K+ current (IKACh-c). The descriptions of the ionic currents were based on patch-clamp data obtained from neonatal rats. Our monolayer model closely mimics the action potential duration (APD) restitution and conduction velocity (CV) restitution curves presented in our previous in vitro studies. In addition, the model reproduces the experimentally observed dynamics of spiral wave rotation, in the absence and in the presence of drug interventions, and in the presence of localized myofibroblast heterogeneities.
Atrial fibrillation (AF) is the most frequent form of arrhythmia occurring in patients with or without the risk for heart failure [1–8]. Each identified form of AF (paroxysmal, persistent, or chronic) requires specialized treatment. Therefore, an in-depth understanding of the underlying mechanisms is required for optimal management in the treatment of such arrhythmias. Chronic AF is known to induce functional tissue remodeling through changes in ion channel expression and activity in the cardiomyocyte membranes. Of the various ion channels that have been identified in atrial cardiomyocytes, one family of particular interest is the Kir3. x. These channels produce an inwardly-rectifying K+ current (IKACh) that is normally controlled by vagal nerve activity through its transmitter acetylcholine (ACh). There is growing consensus that during chronic AF, IKACh becomes constitutively active (IKACh-c) [9–11] and hyperpolarizes the cell membrane. This leads to abbreviation of the action potential duration (APD), and stabilization of reentrant electrical circuits that characterize AF [1]. However, despite decades of research on a variety of mammalian hearts, the role of ion-channel remodeling in AF, remains incompletely understood. Development of a better biophysical understanding of ion channel remodeling in AF requires thorough in vivo investigations in the human heart. However, a more practical approach involves studying AF in cell cultures in vitro, prior to ex vivo and in vivo trials. Primary cultures of neonatal rat atrial cardiomyocytes (NRAMs) have been used by researchers interested in cell signaling [11–13]. Recently, we have demonstrated that NRAMs can generate IKACh-c, in primary cultures, thereby making them attractive sources for arrhythmogenic substrates that can be used to study chronic AF [11]. Although in vitro and in vivo studies have often provided significant insights into the mechanisms underlying reentry initiation in both healthy and diseased cardiac tissue, they have a few limitations. One cannot easily address complex questions related to electrophysiology, that would involve challenging techniques like patch- clamp in parallel with optical mapping measurements, or sub-surface measurements of electrical signals in intact cardiac tissues. Furthermore, independent experimental interventions involving gradual modifications of cell properties, such as gap junctional conductance, coupling between different cells types, localized changes in conductivities of specific ionic channels etc. are very difficult to obtain in an experiment. This is where computer models find wide-scale applicability, from basic science [14–23] to direct clinical applications [24–33]. The basis for any computational study in cardiac tissue is a model for the cardiac cell. However, a model for the NRAM is still lacking. Development of this model is important, because it would allow researchers to perform specific in silico studies next to in vitro experiments, thereby increasing the possibilities for the development of biological insight into arrhythmias like AF. In this study, we formulate the first complex ionic model for NRAM monolayers, based on patch-clamp data from recent literature. To remain consistent with in vitro experiments, we incorporate natural cellular heterogeneity, and a baseline 15–20% randomly distributed myofibroblasts into these monolayers. We then generate spiral waves in these monolayers by burst pacing (BP) and reproduce spiral wave dynamics as seen in experiments, both under normal conditions and under pharmacological inhibition. In particular, we simulate the effect of the atrial-specific drug tertiapin-Q, a blocker of IKACh, on BP-induced arrhythmias, as reported by Bingen et al. [11] In addition, we reproduce spiral wave dynamics in a monolayer with localized myofibroblast heterogeneity.
In the present study, we introduced a mathematical model of neonatal rat atrial tissue, based on available experimental electrophysiological data from NRAMs, and our additional patch-clamp data of IKACh in these cells. If data from NRAMs was unavailable, we used parameters suited to NRVMs. We demonstrated the feasibility of our model to successfully reproduce AP characteristics such as dV/dtmax, AP amplitude, APD50, APD80 and RMP. Using our model, we successfully reproduced the APDR and CVR curves generated using optical mapping experiments on monolayers of neonatal rat atrial cardiomyocytes. These curves are essential determinants of stability of arrhythmias in cardiac tissue. Furthermore, we demonstrated the feasibility of simulations in spatially extended media with our monolayer model by successfully initiating spiral waves by burst pacing. In agreement with the experiments by Bingen et al. , [11] our model demonstrated the possibility of removing reentrant circuits upon inhibition of IKACh-c by a specific blocker tertiapin-Q. In addition, our model showed a meandering spiral wave with similar average period (80 ± 4. 8 ms, versus 81. 3 ± 11. 3 ms in experiments), and a rounded Z-shaped core of size 5 mm. Using our monolayer model, we also reproduced spiral wave dynamics from our in vitro data, in the presence of myofibroblast heterogeneity, thus establishing the robustness of the model, for use alongside experiments. However, in spite of our best efforts to design a state-of-the-art mathematical model, certain inherent flaws remain. A model is only as good as the data that is used to build it. Since the formulations of most of our currents were derived from patch-clamp data from available literature, we faced the problem of dealing with yet incomplete data extracted using different cell-isolation, cell-culture conditions and the patch-clamp protocols and procedures for obtaining these crucial data. Regardless of the variability, these factors may influence the experimental outcome to a great extent. Therefore, instead of relying on absolute numbers for the maximal currents from the I-V curves, we used normalized currents to preserve the trend, while compromising in some cases on the absolute values of the currents produced by the cell. Discrepancies between results obtained from literature and those derived using our model show up in a few important areas. A major limitation of the model is the lack of evidence on Ca2+ dynamics in atrial cardiomyocytes. For simplicity, we adapted the dynamics for intracellular Ca2+ from the ventricular model with modifications as described in the S1 Appendix of the supplementary material. However, this lack of key information about Ca2+ levels in atrial cardiomyocytes can also perhaps account for the slight alteration in model AP morphology. Such difference may impinge on modeling phenomena where accurate representation of the depolarized state is critical. Furthermore, our experimental records of IKACh-c-behavior in single cell versus monolayers raises a fundamental question, that of compatibility between single cell and tissue properties. In this respect, it is important to clarify that being' constitutively active' is not the same as being' maximally active' . Even a small, but stimulant-independent activity of a current can be regarded as constitutive activity. In our single cell patch-clamp experiments, we observed a high-conductance stimulant-independent current at hyperpolarizing potentials, which gave way to a low-conductance current at physiological depolarized voltages. However, monolayers constructed using the same cells showed a pronounced response to antagonist tertiapin-Q, which translated into 2-fold prolongation of the action potential, indicating the activity of a high-conductance current in monolayers. In 2006, Cha et al, [62] performed dedicated studies to measure IKACh-c in single and multicellular preparations of canine atrial cardiomyocytes and observed similar anomaly in the behavior of IKACh-c (see Figs 1C, 1D, 2 and 3 of Cha et al. [62]) Our findings are in line with these observations, i. e. , modest versus pronounced effect of tertiapin-Q on APD in single and multicellular level, respectively. One approach to quantify IKACh-c involves performing patch-clamp studies on neonatal rat atrial cardiomyocytes in the presence of agonist Carbachol. This technique has been used before for that purpose by other researchers such as Dobrev et al. [9] However, addition of Carbachol to observe substantial difference between the IV curves for basal current and IKACh-c would require (i) an extracellular environment for the cell that contains abnormally high levels of K+, resulting in (ii) a condition that will not reflect the actual situation in monolayers, which we aim to model and study in silico. To this day, this apparent discrepancy between the behavior of IKACh-c at single vs multi-cell level remains unresolved. In fact, this matter may not only involve this one particular current, but is likely to concern a more fundamental aspect of single cell vs tissue vs whole organ behavior regarding ion channel dynamics. Therefore, in this manuscript, we employ a reverse-engineering approach in which we use our single cell patch-clamp data for IKACh-c to build a basal formulation for the I-V characteristic curve in our single cell model. When this model is extended to simulate tissue, we combine the concept of Fig 1A from Dobrev et al. [9] about split-I-V curves at physiologically relevant depolarized voltages and further adjust this formulation on the basis of control and tertiapin-Q-treated APD restitution curves from monolayer studies, to include the strength of the current at varying pacing cycle lengths. Nevertheless, the translation from a single to multi-cell level in terms of ion channel dynamics certainly warrants further investigation. Until new insights become available, we believe, however, that our study introduces a new and valuable model that is robust enough to facilitate simulation studies into key aspects of atrial fibrillation. A principal limitation of our model is that an experimentally-validated, detailed Q10-compensation was not factored in for usability of this model at all temperatures. Because of the limited availability of experimental data on NRAMs, we used a Q10 compensation similar to Hou et al. , [51] with certain other adjustments based on the idea that temperature affects channel kinetics of Na+, K+ and Ca2+ currents as well as their maximal conductances, though to a lesser degree. Because no information was available regarding any of the time constants for the gating variables that controlled the different ionic currents, except fast Na+ activation, all the time constants were either taken from the neonatal rat ventricular cardiomyocyte model by Korhonen et al. , [41] or adapted to suit the AP characteristic and APDR data fromNRAMs. Inactivation kinetics of the fast Na+ channel is an important determinant of cardiac tissue excitability. In our model, we fitted the steady-state inactivation curve with data derived from Voitychuk et al. [36] The V½ for this data was ~15 mV more positive than in the steady-state inactivation data of Ramos-Mondragón et al. [35] However, we chose to use the data derived from Voitychuk et al. , [36] instead of the data derived from Ramos-Mondragón et al. , [35] because for a fixed dV/dtmax the kinetics from Ramos-Mondragón et al, did not allow wave propagation in a monolayer, as opposed to the kinetics from Voitychuk et al. [36] Taken together, we present in this paper, a quantitatively-robust, computationally-efficient, experimentally-validated mathematical model of the NRAM expressing the IKACh-c. We use this model to build simulation domains that mimic cardiac monolayers in vitro and successfully demonstrate the usability of the monolayer model to study initiation, maintenance and termination of spiral wave arrhythmias by reproducing key results from our previous in vitro experiments.
NRAMs were isolated from hearts of 2-day-old Wistar rat pups as previously described [11,63]. Isolated cells were plated on round glass coverslips (15-mm diameter) coated with fibronectin (Sigma-Aldrich, St. Louis, MO, USA) in 24-well plates (Corning Life Sciences, Amsterdam, the Netherlands). Depending on the assay, cell densities of 0. 1–8 x 105 cells/well were used and culture as described in our commonly used neonatal rat cardiac myocyte preparation [57,58]. NRAMs were patch-clamped using the whole-cell technique at 20–23°C on day 2–4 of culture. The patch pipettes were pulled from borosilicate glass capillaries (1. 5 mm outer diameter and 1. 17 mm inner diameter; Harvard Apparatus, Kent, UK) with a vertical puller (P-30; Sutter Instrument Company, Novato, CA, USA). Pipettes had typical electrical resistances of 2–3 MΩ in the extracellular solution when filled with the internal solution (for composition, see later). Solitary beating cells were selected under an inverted microscope Zeiss Axiovert 35 (Carl Zeiss AG, Oberkochen, Germany) for the measurements by a MultiClamp 700B amplifier connected to a Digidata 1440A A/D converter (Axon CNS, Molecular Devices, Sunnyvale, CA, USA). After reaching the giga-ohm seal, the holding potential was set to -50 mV, and the patch membrane was ruptured by gentle suction through the pipette. Whole-cell capacitance (Cm) was calculated from capacitive transient currents evoked during 5 mV steps from a holding potential of -50 mV. The capacitance transients were cancelled with the amplifier. To minimize voltage error and improve the adequacy of the voltage-clamp, pipette series resistance was routinely electrically compensated by >75%. The voltage and current command pulses were applied to the amplifier through its external command input from the A/D converter connected to a personal computer. The instruments were controlled and driven by commercially available MultiClamp 700B Commander and Clampex V10. 3 softwares (Molecular Devices) for Windows. Throughout experiments, the current and voltage outputs of amplifier were continuously sampled at intervals of 100 μs and recorded onto the personal computer after low-pass filtering at 2–4 kHz with a four-pole Bessel filter. Action potentials and total membrane currents were recorded in the standard extracellular solution containing (in mM): 126 NaCl, 11 glucose, 10 HEPES, 5. 4 KCl, 1 MgCl2, and 1. 8 CaCl2 (adjusted to pH 7. 40 with NaOH). The standard internal pipette solution used for recording of action potentials and total membrane currents contained (in mM): 80 potassium DL-aspartate, 40 KCl, 8 NaCl, 5. 5 glucose, 5 HEPES, 5 EGTA, 1 MgCl 2,4 Mg-ATP, and 0. 1 Na 3 -GTP (adjusted to pH 7. 20 with KOH). To study the effect of IKACh on AP, 2 μM carbachol (Sigma-Aldrich, St Louis, MO, USA) and 100 nM tertiapin-Q (Alomone Labs, Jerusalem, Israel) were added to standard extracellular solution to activate and block this current in NRAMs, respectively A correction of approximately 11 mV, because of liquid junction potential was applied in the analysis of action potential recordings. Time-dependent hyperpolarization-activated currents (IKH) were recorded in a modified extracellular solution. In addition to the standard extracellular solution, tetrodotoxin (30 μM) (Alomonne Labs), nitrendipine (10 μM), 4-Aminopyridine (2mM) and atropine (100 nM) all from Sigma-Aldrich were added in this solution to suppress, INa, ICaL, Ito and muscarinic receptor-activated currents, respectively. The highly selective Kir3. X channel blocker tertiapin-Q (100 nM) was used to isolate IK1 from the total IKH. The standard internal pipette solution used for the experiments contained (in mM): 80 potassium DL-aspartate, 40 KCl, 8 NaCl, 5. 5 glucose, 5 HEPES, 5 EGTA, 1 MgCl 2,4 Mg-ATP, and 0. 1 Na 3 -GTP (adjusted to pH 7. 20 with KOH) was used to fill the pipettes. For off-line analysis, the data stored on the personal computer were analyzed by pClamp V10. 3 (Molecular Devices) and GraphPad Prism software version 6 (GraphPad Software, Inc. , La Jolla, CA, USA). All data are expressed as means ± standard error of mean (SEM) for a specified number (n) of observations. Significant differences were determined at the P <0. 05 level, unless specified. Unpaired Student’s t test or the one-way ANOVA test followed by Tukey’s test were used for comparing different experimental groups. Statistical significance was expressed as follows: *: P<0. 05, **: P<0. 01, ***: P<0. 001 NS: not significant. Virtual monolayers were constructed by allocating cells on a circular selection in a simulation domain containing 256 x 256 grid points. In order to achieve a random distribution of myofibroblasts within the monolayers, N specific sites were selected from within the array, by a random-number generator. For consistency with experiments, N was chosen such that the myofibroblasts constituted 15–20% of the total number of cells comprising the monolayers. The myofibroblasts were coupled to the cardiomyocytes via a gap junctional conductance Ggap = 0. 5 nS/pF [59]. This coupling is ~6 times smaller than myocyte-myocyte coupling. Whole-cell patch-clamp measurements of APs from different samples of NRAMs showed a slight variation in AP characteristics (S1 Fig). When cardiomyocytes as heterogeneous as these are coupled together to form a monolayer, it would therefore be reasonable to expect a certain degree of natural intercellular variability within the monolayer, although most of the heterogeneity in electrophysiological properties would be compensated by electrotonic coupling. Please refer to Fig 7D for proof of variability in the conductance values of currents IK1 and IKACh. It is possible to adopt a very basic approach to accomplish the same in computer models. To incorporate X1-X2% intercellular variability within the monolayers the following protocol was used: (i) A random-number generator was used to generate 19 numbers from within the range 0. 01 x (X1-X2), for each cardiomyocyte occupying a site on the cellular grid. (ii) At each cell site, the maximal conductance, associated with an ionic current/flux, was multiplied by one of these 19 random numbers (one random number for each individual ionic current/flux). This ensured a spatially-random distribution between the electrophysiological properties of the individual uncoupled cells. Next, these heterogeneous cells were coupled electrotonically. Electrotonic coupling averages out the electrophysiological properties so that the culture exhibits uniform APD distributions. Simulated monolayers were discarded if the APD80 maps showed non-uniform distribution of APD or APD dispersion larger than 20 ms. Finally, APD80 maps from our in vitro cultures were compared with those from the simulated monolayers with varying levels of intercellular heterogeneity. Because the APD80 maps at 50–150% intercellular variability showed closest resemblance with in vitro data in terms of mean value and spatial distribution, we chose to use this range for our numerical experiments. | A fundamentally important element in cardiac in silico research is a model for the cardiac cell. It provides a link between measurable characteristics at the subcellular level and biological processes at the whole cell level, thereby allowing the researcher to study mechanisms of cardiac arrhythmias from a molecular cell biological perspective. Such studies are of vast importance for the advancement of understanding of living systems from cells to patient populations. This paper is a joint in silico-experimental study in which we propose the first model for the action potential of an NRAM. To develop this model, we fitted patch-clamp data from recent literature, while additionally performing specific measurements of IKACh-c in NRAMs. IKACh-c is an important factor in atrial arrhythmogenesis and a promising target for pharmacological AF-management. The model reproduces in vitro results such as standard characteristics of AP morphology, restitution, and spiral wave dynamics in monolayers, with effects of a subsequent drug-intervention and in the presence of localized myofibroblast heterogeneities. Thus it can be used as a tool to provide computational support to a variety of systematic experimental studies that investigate the mechanisms underlying atrial fibrillation (AF) in NRAM cultures. | Abstract
Introduction
Discussion
Methods | carbachol
medicine and health sciences
action potentials
neurochemistry
acetylcholine
drugs
membrane potential
electrophysiology
neuroscience
fibroblasts
connective tissue cells
cardiac pacing
pharmacology
neurotransmitters
cardiology
arrhythmia
atrial fibrillation
animal cells
connective tissue
biological tissue
biochemistry
cell biology
anatomy
physiology
biology and life sciences
cellular types
neurophysiology | 2016 | A Mathematical Model of Neonatal Rat Atrial Monolayers with Constitutively Active Acetylcholine-Mediated K+ Current | 5,119 | 282 |
The taxonomic distinctiveness of Ascaris lumbricoides and A. suum, two of the world' s most significant nematodes, still represents a much-debated scientific issue. Previous studies have described two different scenarios in transmission patterns, explained by two hypotheses: (1) separated host-specific transmission cycles in highly endemic regions, (2) a single pool of infection shared by humans and pigs in non-endemic regions. Recently, A. suum has been suggested as an important cause of human ascariasis in endemic areas such as China, where cross-infections and hybridization have also been reported. The main aims of the present study were to investigate the molecular epidemiology of human and pig Ascaris from non-endemic regions and, with reference to existing data, to infer the phylogenetic and phylogeographic relationships among the samples. 151 Ascaris worms from pigs and humans were characterized using PCR-RFLP on nuclear ITS rDNA. Representative geographical sub-samples were also analysed by sequencing a portion of the mitochondrial cox1 gene, to infer the extent of variability at population level. Sequence data were compared to GenBank sequences from endemic and non-endemic regions. No fixed differences between human and pig Ascaris were evident, with the exception of the Slovak population, which displays significant genetic differentiation. The RFLP analysis confirmed pig as a source of human infection in non-endemic regions and as a corridor for the promulgation of hybrid genotypes. Epidemiology and host-affiliation seem not to be relevant in shaping molecular variance. Phylogenetic and phylogeographical analyses described a complex scenario, involving multiple hosts, sporadic contact between forms and an ancestral taxon referable to A. suum. These results suggest the existence of homogenizing gene flow between the two taxa, which appear to be variants of a single polytypic species. This conclusion has implications on the systematics, transmission and control programs relating to ascariasis.
Ascariasis in pigs and in humans is caused by two of the most socio-economically important nematodes: Ascaris suum Goeze, 1782 and Ascaris lumbricoides Linneaus, 1758, respectively. Human ascariasis is a soil-transmitted helminthiasis (STH), included in the WHO list of neglected tropical diseases (NTD), infecting more than one billion people [1]. Even if the majority of infections are asymptomatic, clinical manifestations of human ascariasis typically involve acute and chronic symptoms (lung inflammation and fever due to larval migration; abdominal pain, nausea, retarded growth in children and intestinal obstruction due to the massive presence of adult worms) [1]. Ascariasis in pigs is frequent in both intensive and extensive breeding systems, being a source of substantial economic losses [2]. Due to their morphological and biological similarities, the taxonomic distinctiveness of A. lumbricoides and A. suum still represents a debated scientific issue. Importantly, this issue is of great relevance for both systematists and epidemiologists alike, given its implications on parasite transmission, zoonotic potential, and the establishment of control programs [3], [4], [5]. Several hypotheses have been proposed to explain the origin of the two ascarid taxa in their respective hosts and their taxonomic status [3], namely: a) A. suum and A. lumbricoides are two valid species; b) A. suum is the ancestor of A. lumbricoides, originated by an allopatric event of host-switching; c) A. lumbricoides is the ancestor of A. suum; d) A. suum and A. lumbricoides are conspecific and therefore occur as variants of a single polytypic species. Previous molecular epidemiological studies have described two different scenarios in transmission patterns that could be explained by two different hypotheses. First, distinct, host-specific transmission cycles have been observed in highly endemic regions as Guatemala and China [4], [5], [6], [7]. Second, a single pool of infection, shared by humans and pigs, has been observed in non-endemic regions, as Denmark and North America [8], [9]. Conversely, recent results strongly suggest that A. suum acts as an important source of human ascariasis in endemic area such as China, where both Ascaris spp. co-occur. Here, the authors observed cross-infections and hybridization of human and pig Ascaris, thus supporting the second hypothesis on transmission cycles [10]. Considering the uncertain epidemiological picture, the main aim of the present study was to investigate genetic variation in two nuclear and mitochondrial target regions (ITS and cox1, respectively) within and among Ascaris populations of human and pig origin, collected from a range of non-endemic regions. These molecular data, along with other published sequences available at both local and global scales, were then used to infer the evolutionary, phylogenetic and phylogeographic relationships among samples. The nuclear ribosomal marker (ITS) was chosen to distinguish A. suum, A. lumbricoides and the hybrid form of the two taxa. Meanwhile, mitochondrial DNA is the most frequently used molecular marker in this kind of studies, due to desirable biological features such as maternal inheritance, high mutation rate, very low recombination rate, haploidy, and putative selective neutrality, making mtDNA markers particularly suitable as barcoding tools to identify sibling and cryptic species [11], [12]. Studies aimed at investigating the molecular epidemiology of ascariasis are important not only to clarify the transmission patterns of the two roundworms, but also to better quantify the level of gene introgression between host-associated populations [10]. Such knowledge is important, given that introgression often results in the selection of novel genes, the promotion of rapid adaptive diversification, and homogenization across the genomes of the interbreeding populations [13], [14]. Additional sources of information are now available from the recently published draft genome of A. suum [15].
A total of 151 adult nematodes belonging to Ascaris spp. were collected from pig (n = 143) and human (n = 8) hosts. Nematodes collected were repeatedly washed in saline and stored in 70% ethanol. Collection data including collecting sites, hosts, number of parasites specimens analysed and identification codes are summarised in Table 1. DNA was isolated using the Wizard Genomic DNA purification kit (Promega) according to the manufacturer' s protocol. All samples, from human and animal origin, were obtained from existing collections. Samples from human origin were obtained from existing collections at Tor Vergata and Sant' Andrea Polyclinics in Rome. Data collection includes only the geographical origin of patients and no reference to personal data was recorded, thus guaranteeing the absolute anonymity of these specimens. Sample collection at the Polyclinics that provided the nematodes from humans was performed in concordance with the WMA Helsinki Declaration (Edinburgh 2000) and its subsequent modification, as well as with the Italian National Law n. 675/1996 on the protection of personal data. The entire ITS nuclear region (ITS1,5. 8S, ITS2) was amplified using 5. 0 µl of template DNA (20–40 ng), 10 mM Tris-HCl (pH 8. 3), 1. 5 mM MgCl2 (Bioline), 40 mM of a nucleotide mix (Bioline), 50 pmol/µl each of the forward primer NC5 (5′-GTAGGTGAACCTGCGGAAGGATCAT-3′) and the reverse primer NC2 (5′-TTAGTTTCTTCCTCCGCT-3′) described by Zhu et al. [16] and 1. 0 U of BIOTAQ DNA Polymerase (Bioline) in a final volume of 50 µl. The PCR was performed in a GenePro Eurocycler Dual Block (Bioer) under the following conditions: 10 min at 95°C (initial denaturation), 30 cycles of 30 sec at 95°C (denaturation), 40 sec at 52°C (annealing) and 75 sec at 72°C (extension), and a final elongation step of 7 min at 72°C. A negative control (without genomic DNA) was included in each set of amplification reactions. A representative subset of specimens (Table 2) was also analysed by sequencing a portion of the mitochondrial cytochrome oxidase I gene (cox1), after amplification using the forward primer As-Co1F (5′-TTTTTTGGTCATCCTGAGGTTTAT- 3′) and the reverse primer As-Co1R (5′-ACATAATGAAAATGACTAACAAC- 3′), as described by Peng et al. [6], under the following conditions: 5 min at 94°C, followed by 35 cycles of 94°C for 30 s; 45 s at 55°C; 90 s at 72°C, followed by 5 min at 72°C. Aliquots (5 µl) of individual PCR products were separated by electrophoresis using agarose gels (1%), stained with ethidium bromide (0. 4 µg/ml) and detected using ultraviolet trans-illumination. Positive ITS amplicons were digested with the restriction endonuclease HaeIII, as the resulting patterns have been previously proved useful for the identification of human and pig Ascaris species [8]. Digests were resolved by electrophoresis in 2% agarose gels, stained with ethidium bromide (0. 4 µg/ml), detected under UV trans-illumination, and the fragments sizes determined by comparison with a 100 bp DNA ladder (Promega). Information on geographical origin, hosts, codes, number of parasites successfully genotyped, and genotypes recovered using PCR-RFLP are available in Table 1. Positive amplicons were purified by SureClean (Bioline), following the manufacturer' s instructions, and then sequenced by MWG Eurofins DNA. Two different datasets were created, each representing different partial cox1 alignments: the first including only samples analysed in the present paper (Dataset1), with the exclusion of two human nematodes due to small sample size (single specimens from Pakistan and Romanian human patients), and the second including all GenBank retrieved sequences of specimens collected from endemic and non-endemic regions (Dataset2). Information about specimens sequenced for cox1, identification codes and accession numbers, also of GenBank retrieved sequences are available in Table 2. Nucleotide sequences were aligned using Clustal X implemented in MEGA 5 [17] and then analysed using DnaSP v5 [18] to infer haplotype composition. In addition, sequences were analysed using Arlequin 3. 11 [19] to estimate several variability indexes: the relative frequencies of haplotypes; population differentiation (FST) among samples for Dataset1; hierarchical analyses of molecular variance (AMOVA) to evaluate the amount of population genetic structure for Dataset2, using information on the allelic content of haplotypes, as well as their frequencies. The significance of the covariance components associated with the different levels of genetic structure (within individuals of populations, among populations and among groups) was tested using non-parametric permutation procedures [20]. The AMOVA was undertaken twice, using two different criteria to define groups and population structure: geographical origin (endemic and non-endemic regions) and host affiliation (pig and human). Both Dataset1 and 2 were also analysed using a phylogenetic approach based on Bayesian reconstruction method. The program JModeltest [21] was used to compare the fit of nucleotide substitution models using the Akaike Information Criterion (AIC), under a total of 83 models, corresponding to 11 different schemes; the best-fit model and parameters determined for both cox1 datasets were then used for the Bayesian analyses. The Bayesian analyses were performed using the HKY+I model for both datasets (as selected by ModelTest), using BEAST software [22]; datasets were run twice for 106 generations. Posterior probability values (BPP) shown in the Bayesian consensus trees were determined after discarding trees from the burn-in period. For each dataset, burn-in was estimated to include the first 104 generations. A second phylogenetic method was performed only on Dataset 2 using MEGA5 [23]: the evolutionary distances were computed using the Tamura-Nei [24] with Neighbor joining method (NJ) and statistical support at nodes was evaluated using 1000 pseudoreplication bootstrap [25]. Phylogenetic trees included Anisakis Dujardin 1845 as outgroup (GenBank accession number: JN102304). Moreover, statistic parsimony networks [26] using TCS software [27] were inferred for both datasets in order to determine the phylogeographic distribution and genealogy of the Ascaris specimens analysed, running the network at a 95% connection limit, which is the maximum number of mutational connections between pairs of sequences justified by the parsimony criterion.
A PCR product of around 1000 bp was obtained for 137 of the 151 specimens analysed. Amplicons were subsequently digested using the HaeIII restriction enzyme. This approach yielded the identification of three genetically distinct banding patterns belonging to the genus Ascaris: the “lumbricoides” genotype displays two bands of about 610 bp and 370 bp, the “suum” genotype shows three bands of about 610 bp, 230 bp and 140 bp, and the “hybrid” genotype displays all the four bands mentioned above (Figure 1). While the proportion of each genotype varied somewhat across the various localities sampled, all regions revealed instances of discordance between the expected genotype and host of origin (Table 1). For Italy, although 49 of 60 positive samples from pigs displayed the expected “suum” genotype, nine displayed the “hybrid” genotype and two displayed the “lumbricoides” genotype. In contrast, neither of the two positive human isolates displayed the expected “lumbricoides” pattern, instead revealing one “suum” and one “hybrid” genotype. Positive samples obtained from nematodes collected in other countries included four specimens from humans and 71 from pigs. Of the human nematodes, three specimens (Syrian, Pakistan and Romanian patients) showed the typical “lumbricoides” genotype and one (another Romanian patient) displayed the “suum” genotype. Among Slovak pigs (n = 44), 36 showed the “suum” genotype, four the “lumbricoides” genotype, and four the “hybrid” pattern, while Hungarian pigs (n = 27) included 19 specimens and eight specimens displaying the “suum” genotype and “hybrid” genotypes, respectively. Overall, the “hybrid” genotype was encountered in specimens from both pig and human hosts, at a frequency of 16%. A PCR product of around 400 bp was obtained for 62 specimens amplified. The alignments of Dataset1 (62 sequences) and Dataset2 (120 sequences) yielded a usable alignment of 327 bp. Representative sequences for each haplotype recovered in the course of the present study are available in GenBank under the following Accession Numbers: Hap1: KC455923, Hap2: KC455924, Hap3: KC455925, Hap4: KC455926, Hap5: KC455927, Hap6: KC455928, Hap7: KC455929, Hap8: KC455930, Hap9: KC455931, Hap10: KC455932, Hap11: KC455933, Hap12: KC455934, Hap46: KC455935. Twelve haplotypes were identified in Dataset1 (Hap1-12), with a total haplotype diversity (Hd) of 0. 70 (haplotypes recovered were deposited in GenBank, see Table 2 for accession numbers). Five haplotypes were observed in Slovak sample, with Hd = 0. 71; three haplotypes were observed in Hungarian sample, with Hd = 0. 24 and seven haplotypes were observed in Italian sample, with Hd = 0. 62. The most frequent haplotype was Hap5, shared among the Italian (frequency of 61. 5%), Hungarian (87. 5%) and Slovak samples (5. 5%). Hap1 was the most frequent haplotype in the Slovak population (44. 4%) and it has been less frequently reported also in Italian specimens (7. 7%). Results from FST analysis showed significant differences between Slovak sample and the Italian (0. 29) and Hungarian samples (0. 49), and little differentiation between Italian and Hungarian samples (0. 05). Considering Dataset2, forty-five haplotypes were identified, with Hd = 0. 89; Hap5 was observed also in the Chinese pig sample. The Italian and Slovak samples showed haplotype Hap7 in common with endemic (Brazil, Zanzibar and China) and non-endemic regions (Japan); the Italian sample showed also haplotype Hap12 in common with endemic regions. Information about haplotypes recovered in the partial cox1 sequences analyses, haplotype affiliation to phylogenetic clusters A (A1, A2) -B-C, GenBank accession numbers, codes, correspondences to genotypes identified using RFLP approach on ITS, hosts, endemic and non-endemic origin of samples and haplotypes relative frequencies for populations of Dataset1 are available in Table 2. AMOVA analysis suggested a higher influence of the epidemiological (endemic/non-endemic origin) criterion in modulating the accumulation of variability with respect to host affiliation, even if the percentage of variation at group level was not significant (3. 83% and 0. 10%; p = 0. 38 and 0. 61, respectively). Significant values (p≤0. 05) were obtained for the variation observed among populations within groups and among individuals within populations in both analyses, but with an opposite trend: percentage of variation within population was higher than among populations of the same group if the endemic/non-endemic criterion is considered as feature to group samples. Bayesian and NJ phylogenetic analyses, based on Dataset1 and Dataset2, described similar topologies, with three main clusters (Figure 2), analogous to the clusters named A, B and C in Anderson and Jaenike [28] and Snabel et al. [29] studies. Clusters A and B have been recently reported also by Iniguez et al. [30]. Cluster A includes samples from both pigs and humans collected from endemic and non-endemic zones; it showed further slight internal subdivision according to host affiliation and epidemiological features, although no statistical support for this partitioning was found. Sub-cluster A1 contains mainly specimens from pigs and few from humans, collected from non-endemic zones. It is important to underline that the specimens of human origin (ASR_H and ASI12_H) included in this group showed the typical “suum” genotype for PCR-RFLP analysis of the ITS region. Sub-cluster A2 includes mainly specimens from humans collected from endemic areas, except for one human sample (ASI13 corresponding to Hap12) collected from non-endemic regions, although the country origin of the patient is unknown. Cluster B is also characterized by the presence of specimens from both pigs and humans collected from endemic (Brazil, China, Zanzibar, Pakistan) and non-endemic zones (Japan, Italy). Cluster C comprises only specimens from pig collected from non-endemic regions (Italy and Slovakia). It appears to be well separated from clusters A and B that are more closely related to each other. The existence of the three clusters is well supported by very high posterior probability values (BPP ranging from 92 to 98 for Dataset1 and from 90 to 100 for Dataset2); NJ tree bootstrap values show high statistical support for cluster C (93) and lower values for cluster A (51) and B (38), nevertheless the value supporting the distinctiveness of cluster C from A and B together is fairly high (77). Results obtained from parsimony network analysis on Dataset2 (Figure 3) describes a very complex scenario where the three clusters observed in phylogenetic analysis are recognized and the slight subdivision inside cluster A is still evident. The main haplogroup, where Hap5 is the more frequent and typically associated to A. suum, corresponds to cluster A1 with several haplotypes branching around. The star-like distribution of haplotypes is also evident in the other haplogroups, represented by Hap12 for cluster A2 and Hap7 for cluster A. Cluster A2 is mainly represented by haplotypes from endemic regions, typically associated to A. lumbricoides, with the exception of Italian and Japanese human cases; while cluster B includes both pig and human specimens from endemic and non-endemic regions. The Slovak haplogroup appears completely separated from the other haplotypes. These results confirm the relationships observed in the Bayesian phylogenetic trees.
Human and pig Ascaris spp. are two of the world' s most common soil-transmitted parasites and together cause serious health and socio-economic problems. Ascariasis is considered a NTD as it occurs commonly in rural and poor urban areas and promotes poverty due to its high impact on child health and development, pregnancy and worker productivity. Similarities in the morphology and biology of these two nematodes entail ongoing ambiguity concerning their taxonomic status and argue for the need to delve deeper into their comparative molecular epidemiology. The present paper provides additional information on the molecular epidemiology of ascariasis in non-endemic regions, such as Italy and Eastern Europe. Molecular characterization using a PCR-RFLP approach on a nuclear marker has confirmed that most pig nematodes sampled herein displayed the typical A. suum pattern, corresponding to the genotype G3, while the two human nematodes from endemic regions such as Pakistan and Syria showed the typical A. lumbricoides pattern, corresponding to the genotype G1 [31]. Cross-infection is confirmed in both hosts by instances of A. suum genotypes in human nematodes and A. lumbricoides in pigs. Moreover, a significant percentage of nematodes displaying the “hybrid” pattern, corresponding to the G2 genotype [31], has been observed in both human and pig nematodes, strongly inferring the presence of gene flow between the two taxa. This combined evidence suggests that Ascaris suum can function as a relevant agent of human infection in non-endemic areas. These data are in agreement with recent results described firstly by Betson et al. in Zanzibar [32] and then by Zhou et al. in China [10], where zoonotic transmission of A. suum is suggested to occur also in these endemic areas. The zoonotic potential of A. suum therefore needs to be reevaluated in order to plan more efficient control programs. Phylogenetic analyses revealed the homology to the clusters previously observed in Anderson and Jaenike [28] and in Snabel et al. [29], confirming that geographical origin plays an important role in structure of cluster A, where endemic and non-endemic samples split in two sub-clades, but not in cluster B, which contains specimens from both epidemiologically classified regions. Finally, significant values on population differentiation analysis and high haplotype diversity confirm the genuine separation of cluster C. As these parameters are important indexes for evaluating genetic diversity and differentiation, further analysis will be required to understand the significance of this pronounced genetic dissimilarity. Phylogeographic analyses are helpful in understanding population differentiation, species formation and ecological adaptation [33]. Results obtained from the haplotype network analysis have revealed a very complex scenario: the typical A. suum haplotype is the most frequent among samples from non-endemic regions plus is observed also in human patients (circle A1); moreover, this haplogroup is closely related to the haplogroup including the distinctive A. lumbricoides haplotypes found in endemic regions (circle A2), which is related in turn to a mixed group homologous to cluster B obtained in phylogenetic inferences (circle B). The picture described a cross-linked relationships among haplotypes, where no clear geographical or host-affiliation criteria seem to be relevant in shaping haplogroups. Shared haplotypes between pig and human Ascaris spp. could be explained by evolutionary processes such as introgression and/or retention of ancestral polymorphisms, as suggested previously [9], [34]. In addition, molecular variance analysis underlined that accumulation of genetic variability is observed at the individual and population level rather than at the level of groups defined on geography or host-affiliation. The overall results showed no fixed differences between human and pig Ascaris, describing two taxonomic entities intimately interconnected and therefore likely to experience gene flow. These data strongly infer the absence of a major genetic barrier between the two taxa and therefore suggest that A. suum and A. lumbricoides may be variants of the same species, as suggested by Leles et al. [3] and Liu et al. [35], and more recently by Iniguez et al. [30]. Together all four studies have found no evidence of diagnostic genetic heterogeneity between human and pig Ascaris, plus an absence of genetic clusters discriminating each host. | Ascaris lumbricoides, the world' s most common human nematode, and A. suum, the pig roundworm, are two of the most important soil-transmitted helminthes of public health and socio-economic concern. However, previously documented similarities at the morphological and genetic level, coupled with evidence for hybridization and gene flow, have clouded the taxonomic distinctiveness of these two nematodes. To date, molecular epidemiological studies have been carried out, mostly in highly endemic regions, where two different transmission cycles have been described. Recently, pigs have been recognized as an important source of human ascariasis in China, opening questions about the zoonotic potential and the efficiency of control programs. Here, samples from non-endemic regions have been analysed using a nuclear marker to identify nematodes to species level plus a mitochondrial marker to investigate the phylogeographic relationships among individuals of the two species from both endemic and non-endemic regions. Results obtained suggested that A. suum and A. lumbricoides may be variants of the same species, with the lack of fixed genetic differences and considerable phylogeographic admixture confirming an extremely close evolutionary relationship among these nematodes. This study highlights the need to further explore the evolutionary affinities of the two taxa to help shed light on the epidemiology of ascariasis. | Abstract
Introduction
Methods
Results
Discussion | taxonomy
medicine
public health and epidemiology
population genetics
parasitic diseases
phylogenetics
molecular systematics
neglected tropical diseases
ascariasis
infectious diseases
soil-transmitted helminths
genetic polymorphism
molecular epidemiology
epidemiology
biology
evolutionary systematics
haplotypes
evolutionary biology | 2013 | Phylogeographical Studies of Ascaris spp. Based on Ribosomal and Mitochondrial DNA Sequences | 6,135 | 326 |
Vpx is a small virion-associated adaptor protein encoded by viruses of the HIV-2/SIVsm lineage of primate lentiviruses that enables these viruses to transduce monocyte-derived cells. This probably reflects the ability of Vpx to overcome an as yet uncharacterized block to an early event in the virus life cycle in these cells, but the underlying mechanism has remained elusive. Using biochemical and proteomic approaches, we have found that Vpx protein of the pathogenic SIVmac 239 strain associates with a ternary protein complex comprising DDB1 and VprBP subunits of Cullin 4–based E3 ubiquitin ligase, and DDA1, which has been implicated in the regulation of E3 catalytic activity, and that Vpx participates in the Cullin 4 E3 complex comprising VprBP. We further demonstrate that the ability of SIVmac as well as HIV-2 Vpx to interact with VprBP and its associated Cullin 4 complex is required for efficient reverse transcription of SIVmac RNA genome in primary macrophages. Strikingly, macrophages in which VprBP levels are depleted by RNA interference resist SIVmac infection. Thus, our observations reveal that Vpx interacts with both catalytic and regulatory components of the ubiquitin proteasome system and demonstrate that these interactions are critical for Vpx ability to enable efficient SIVmac replication in primary macrophages. Furthermore, they identify VprBP/DCAF1 substrate receptor for Cullin 4 E3 ubiquitin ligase and its associated protein complex as immediate downstream effector of Vpx for this function. Together, our findings suggest a model in which Vpx usurps VprBP-associated Cullin 4 ubiquitin ligase to enable efficient reverse transcription and thereby overcome a block to lentivirus replication in monocyte-derived cells, and thus provide novel insights into the underlying molecular mechanism.
Vpx accessory proteins are virulence factors encoded by viruses of the HIV-2/SIVsm/SIVmac lineage of primate lentiviruses. vpx gene disruption results in greatly reduced rates of virus replication in monocyte-derived cells, such as differentiated macrophages, but has no overt effect in primary T lymphocytes, as well as T and monocytic cell lines [1], [2], [3]. Intact vpx gene is required for optimal replication of these viruses in the infected host [4], [5]. Thus, it is thought that the role of Vpx in natural infection is to enable the establishment of virus reservoirs in macrophages. Vpx is recruited into viral particles through the interaction with the p6 component of Gag [6], [7], and thus is available to facilitate an early event in the virus life cycle upon virion entry into the target cell. Indeed, an early study revealed that Vpx is required for efficient transport of preintegration complexes to the nuclei of infected macrophages [3]. In more recent studies HIV-2 and SIVsm Vpx proteins were found to promote accumulation of reverse transcribed viral genomes upon infection of dendritic cells (DCs) and this effect may reflect the ability of Vpx to overcome a proteasome dependent mechanism that inhibits an as of yet unidentified early event in the viral replication cycle [8]. How Vpx intersects this ubiquitin-dependent proteasomal protein degradation mechanism is unclear. Vpx is a paralogue of Vpr accessory factor encoded by all known lineages of primate lentiviruses [9]. Although their amino acid sequences are closely related, the two proteins have different roles along the viral life cycle. For example, Vpr has the ability to activate DNA damage checkpoint and thereby arrest cells in the G2 phase of the cell cycle, while Vpx does not possess this function (reviewed in [10]). Results from recent proteomic studies revealed that lentiviral Vpr proteins associate with components of the ubiquitin proteasome system (UPS), such as Vpr Binding Protein (VprBP, GenBank NM014703) termed also DDB1 and CUL4-associated factor 1 (DCAF1), damaged DNA-binding protein 1 (DDB1, GenBank U18299), DET1 and DDB1 associated 1 (DDA1, GenBank DQ090952) and Cullin 4 (GenBank NM001008895, NM003588) ([11], [12], [13] reviewed in [14]). Cullin 4 is a scaffold protein that assembles a family of E3 ubiquitin ligase complexes. DDB1 is an obligatory subunit of all Cullin 4 E3' s that bridges the catalytic cores organized on the Cullin 4 scaffold to a substrate-recruiting subunit, and VprBP/DCAF1 is a putative substrate adaptor for Cullin 4-based E3 ubiquitin ligases ([15] reviewed in [16], [17]). Evidence has been obtained showing that these interactions provide Vpr with the ability to modulate specifically the intrinsic catalytic activity of the Cullin 4 E3 containing VprBP and with a potential to influence the recruitment of substrate proteins for ubiquitination by Cullin 4, which in turn leads to the activation of DNA damage checkpoint [13], [18]. Since Vpx, similarly to Vpr, probably functions as an adaptor protein, we have used a combination of biochemical and proteomic methods to identify downstream effectors of Vpx encoded by the pathogenic SIVmac 239 strain. Here we show that SIVmac Vpx also binds DDA1-DDB1-VprBP complex, which links Vpx to Cullin 4, thus extending the previous observation that another SIV Vpx variant can bind VprBP [11]. Importantly, we demonstrate that VprBP, and its interaction with Vpx, are required for efficient macrophage transduction by SIVmac. Surprisingly, in the absence of Vpx, the incoming RNA genome is reverse transcribed very inefficiently. These findings indicate that Vpx facilitates macrophage infection by acting prior to and/or during reverse transcription, rather than by facilitating nuclear transport of the fully reverse transcribed preintegration complex, as has been thought previously ([3], reviewed in [10]). Together, our findings identify the UPS system and the VprBP associated protein complex as cellular machinery and immediate downstream effector that Vpx uses to promote replication of cognate primate lentiviruses in cells of monocyte/macrophage lineage, and provide novel insights into the underlying mechanisms.
Two complementary strategies were used to identify cellular proteins that are bound by SIVmac 239 Vpx. As one approach, U937 monocytic cell populations were transduced with BABE-puro retroviral vectors stably expressing Vpx tagged at its N-terminus with a triple HA-FLAG-AU1 epitope tag (hfa-Vpx). The population of positively transduced cells was then selected with puromycin and expanded in spinner cultures for biochemical experiments. Surprisingly, we observed that U937 cells that stably expressed Vpx grew more slowly than the control U937 population transduced with an empty BABE-puro vector (data not shown), suggesting that Vpx is toxic and/or cytostatic to these cells. This in turn raised the possibility that chronic Vpx expression could lead to selection of escape variants where Vpx interaction with cellular proteins is not faithfully reproduced. Therefore, as an additional approach hfa-Vpx was expressed transiently in human embryonic kidney 293T (HEK 293T) cells by calcium phosphate co-precipitation. Next, Vpx and its associated proteins were purified from U937 and HEK 293T detergent extracts by sequential immunoprecipitations with anti-HA- and anti-FLAG- epitope antibodies, each followed by elution with the respective peptide epitope. The immunoprecipitates were proteolyzed without prior separation of protein bands by SDS-PAGE and peptide mixtures analyzed by multidimensional protein identification technology (MudPIT, [19]). Interestingly, the most abundant cellular polypeptides we found associated with Vpx both in U937 and HEK 293T cells, but were absent from control purifications from cells that did not express Vpx were DDA1, DDB1 and VprBP/DCAF1 (see Table 1). Significantly, DDB1, an obligatory subunit of all known Cullin 4 based E3 ubiquitin ligases [16], VprBP, a known Vpr-binding cellular protein that has been recently shown to bind DDB1 and postulated to function as a substrate receptor for Cullin 4 E3 ubiquitin ligase [15], [20] and DDA1, a DDB1-binding protein that links to a negative regulator of Cullin4 E3 ubiquitin ligases [21], were thus identified as relatively abundant Vpx-associated proteins. Notably, DDB1, VprBP and DDA1 were recently shown to assemble a ternary complex that associates with Vpr proteins of HIV-1 and SIVmac and mediates activation of DNA damage checkpoint by these accessory factors [13]. Thus, our observations suggested that Vpx and its Vpr paralog both act through the DDA1-DDB1-VprBP complex, even though the two proteins execute distinct functions. To verify the data from MudPIT analyses, experiments were performed to confirm that Vpx associates specifically with the endogenous DDB1-VprBP-DDA1 complex. hfa-Vpx was transiently expressed in HEK 293T cells. Then, hfa-Vpx and its associated proteins were immunoprecipitated from detergent extracts prepared from the transfected cells with anti-FLAG-affinity resin, separated by SDS-PAGE and analyzed by western blotting with antibodies specific for VprBP, DDB1 and DDA1. As shown in Figure 1A, VprBP, DDB1 and DDA1 were readily detected in immune complexes isolated from hfa-Vpx expressing, but not from control, HEK 293T cells. Thus these data confirm that Vpx associates with DDB1, VprBP and DDA1 (compare lane 2 with 1). The finding that Vpx associates with DDA1, VprBP and DDB1 was not entirely surprising because SIVmac Vpx amino acid sequence is approximately 25% and 50% identical to those of HIV-1 and SIVmac Vpr proteins, respectively, and because some of the previously tested SIVmac/HIV-2 Vpx variants were reported to bind VprBP [11], [22]. Previous studies have demonstrated that Vpr binds DDA1-DDB1-VprBP complex via its C-terminal α-helical region [11], [13]. Given the high degree of sequence identity between Vpx and Vpr proteins, this raised the possibility that Vpx binds the above complex in a manner similar to that seen with Vpr [11]. To test this and to develop mutant Vpx proteins defective for the interaction with DDA1-DDB1-VprBP, we substituted amino acid residues located in the C-terminal α-helical region of Vpx that are conserved in Vpr proteins (see Figure 1B). Mutant Vpx proteins were then transiently expressed in HEK 293T cells, immunoprecipitated via their FLAG tags, and immune complexes analyzed by Western blotting. As shown in Figure 1A, alanine substitution for the conserved glutamine Q76 (Q76A) disrupted Vpx ability to associate with DDA1, VprBP and DDB1. Also, alanine substitution for the conserved phenylalanine F80 (F80A) and arginine substitution for histidine 82 (H82R) had similar effects. Of note, the corresponding mutations in HIV-1 Vpr were previously shown to disrupt the binding to the VprBP-associated protein complex ([11], data not shown). Finally, mutating the conserved glycine G86 and cysteine C87 residues (GC86NG) did not have a detectable effect. We conclude that Vpx binds the DDA1-DDB1-VprBP complex via its C-terminal domain, probably using an interaction surface that is also conserved in the Vpr protein. The VprBP-DDB1 module was found to bind Cullin 4 and to participate in a functional Cul4-DDB1[VprBP] E3 ubiquitin ligase complex [13]. Therefore, we tested whether Vpx can associate with Cullin 4 and, if so, whether VprBP mediates this association. hfa-Vpx was transiently expressed together with myc-tagged Cullin 4A isoform (m-Cul4) and/or myc-tagged VprBP (m-VprBP) in HEK 293T cells, and anti-FLAG immunoprecipitates were analyzed for Cullin 4 by immunoblotting. As shown in Figure 2, VprBP co-expression dramatically elevated the levels of Cullin 4 associated with wild type Vpx, but not with the Vpx (Q76A) variant that is unable to interact with VprBP and DDB1 subunits of the E3 complex (compare lane 4 with 2 and 5). We conclude that VprBP links Vpx to the Cullin 4-based E3 complex. Notably, we observed that the Vpx-associated Cullin 4 migrated as a doublet (see lane 4). The slower migrating form of Cullin 4 was much less abundant in immune complexes assembled with VprBP in the absence of Vpx (lane 9), and co-migrated with the neddylated form of Cullin 4 shown previously to be induced by HIV-1 Vpr (see Figure S1, and ref. [13]). As expected, the upshifted Cullin 4 isoform was not detected in the E3 complex containing the DDB2 substrate receptor, which is catalytically repressed in the absence of damaged DNA (lane 7, see ref. [23]). Notably, in contrast to Vpr, the Vpx-induced modification was much less pronounced, and did not lead to a robust increase in catalytic activity of the associated E3 (Figure S1). We conclude that SIVmac Vpx is a much less potent inducer of Cullin 4 neddylation and E3 catalytic activity, than HIV-1 NL43 Vpr. The ability of Vpx to enable infection of primary macrophages is well documented, yet the immediate downstream mediator (s) of Vpx remains unknown [1], [2], [3]. Therefore, experiments were performed to assess whether the interaction with VprBP and its associated E3 complex is important for Vpx' s ability to facilitate macrophage transduction by SIVmac 239. Since this function is probably mediated by the virion-bound Vpx molecules, our initial experiments assessed the ability of the mutant Vpx proteins to be incorporated in SIVmac 239 virions. VSV-G pseudotyped single cycle SIVmac 239 (GFP) viruses encoding wild type or mutated Vpx variants that do not bind VprBP, or possessing an inactive vpx coding sequence due to termination codon substitutions for methione codons M1 and M62 were produced from HEK 293T cells. All viruses contained a frameshift mutation in the env gene which prevented expression of a functional Env glycoprotein, and expressed GFP marker protein from an IRES element positioned immediately downstream of the nef gene (SIVmac 239 (GFP) ). A reference panel of virions containing decreasing amounts of wild type Vpx were also produced from HEK 293T cells transiently co-expressing SIVmac 239 (GFP) proviral construct possessing wild type vpx gene mixed with an isogenic construct containing the M1- and M62- mutated vpx, at 1∶3,1∶7, or 1∶15 ratio. Virions were partially purified and concentrated by pelleting through 20% sucrose cushion and then analyzed by immunobloting for p27 Capsid and for Vpx. As shown in Figure 3A, the Q76A and F80A substitutions had only minor effects on the abilities of the mutant proteins to be incorporated into the virions (compare lanes 5 and 6 with 1–4). The H82R substituted Vpx was incorporated into viral particles very poorly (data not shown), and therefore was not studied further. Next we measured the abilities of the VSV-G pseudotyped single cycle virions to transduce human monocyte derived adherent macrophages. Monocytes obtained from human peripheral blood mononuclear cells (PBMC) by negative selection for CD3, CD7, CD16, CD19, CD56, CD123 and Glycophorin were differentiated into macrophages in the presence M-CSF. Macrophage cultures were then infected with normalized virion preparations and transduction efficiencies of the wild type and mutant viruses were quantified by flow cytometric analysis of GFP expression in the infected cell populations. As controls, CD4+ T lymphocytes purified from PBMC by positive selection for CD4 and activated by phytohemagglutinin in the presence of IL-2, and Jurkat T cells, were also infected and analyzed in parallel. As shown in Figure 3B, panels 2–6, wild type Vpx stimulated macrophage transduction by up to 100-fold, in a dose-dependent manner (5. 1% vs 0. 06% GFP-positive cells). Significantly, Vpx (Q76A), or Vpx (F80A), substituted Vpx failed to support macrophage infection (panels 7 and 8), even though the mutant Vpx molecules were efficiently incorporated into the virions. In contrast, all viruses displayed similar infectivities to primary CD4+ T lymphocytes and Jurkat T cells (panels 9–16 and 17–24), indicating that Vpx is not required for transduction of primary T cells and established T cell lines, consistent with previous observations [2]. Thus, the Q76A and F80A changes link SIVmac Vpx ability to enhance macrophage transduction to its interaction with VprBP and its associated E3 ubiquitin ligase complex. Vpx proteins encoded by HIV-2 viruses also enhance transduction of monocyte-derived cells, but a previous report suggested that they may be unable to bind VprBP [22], [24]. This in turn raised a question whether HIV-2 Vpx uses a VprBP-independent mechanism to enable macrophage infection. To address this issue we asked whether Vpx variant encoded by HIV-2 Rod proviral clone binds VprBP. We chose this particular Vpx variant because it is required for the ability of HIV-2 Rod to transduce primary macrophages and, therefore, is functional [24]. Of note, it is evident from phylogenetic analyses that both the Rod Vpx and SIVmac 239 Vpx are representative of two major groups of HIV-2 Vpx variants (see Figure S2). As shown in Figure 4A, wild type, but not Q76A-substituted, Rod Vpx protein readily bound VprBP in a transient expression assay in HEK 293T cells (compare lanes 1 and 2). Next we assessed the abilities of both proteins to enhance macrophage transduction by a single cycle SIVmac 239 (GFP) reporter virus. We found that only wild type Rod Vpx rescued the infectivity of single cycle SIVmac 293 (GFP) reporter virions that were devoid of SIVmac Vpx, even though both the wild type and Q76A substituted Rod Vpx variants were incorporated into the virions to similar extents (Figure 4B and 4C). We conclude that the interaction with VprBP is a conserved function of SIVmac and HIV-2 Vpx proteins, and that both use VprBP to enable macrophage infection. Vpx was reported to be essential for efficient reverse transcription and/or nuclear import of lentiviral genomes in monocyte-derived cells [3], [8]. Hence we examined the effect of Q76A and F80A substitutions in Vpx on reverse transcription (RT) of the incoming SIVmac genomes by real-time quantitative fluorescent PCR. Macrophages were transduced with a reference panel of VSV-G pseudotyped single cycle SIVmac 239 (GFP) virions containing decreasing amounts of wild type Vpx, or Vpx (Q76A) and Vpx (F80A) variants, characterized in Figure 3A. DNA was isolated from the transduced cells 18 hours and 72 hours later and RT intermediates were quantified by real time PCR with four sets of primers shown in Figure 5A. The primers were designed to amplify strong-stop DNA (early), RT products synthesized immediately following minus strand transfer (U3), or a region of the gag gene located approximately 8000 nucleotides distal from U3 (gag), as well as late RT products synthesized following successful plus strand transfer (late). As shown in Figure 5B, these analyses revealed that reverse transcription was defective following infection with virions lacking, or containing suboptimal amounts of Vpx. First, the steady state levels of the early strong-stop RTs were approximately 10-fold lower in the absence of Vpx and the magnitude of the decrease was inversely correlated with the Vpx virion content. Second, the levels of U3, gag and late RTs were progressively lower upon infection with Vpx-deficient virions (approx 100-fold, 300-fold and 1000-fold, respectively) at the 18 hour time point. These differences were less pronounced at the 72 hour time point. Importantly, Vpx was not required for efficient reverse transcription following infection of Jurkat T cells. Together, these observations indicate that Vpx is required for events that lead to an efficient initiation and progression of reverse transcription of SIVmac genome in macrophages. Next we tested the effects of Q76A and F80A substitutions in SIVmac Vpx for its ability to enable efficient reverse transcription of the SIVmac genome in macrophages. As shown in Figure 5C, both Vpx variants conferred a Vpx-deficient virion phenotype. Since Q76A and F80A each disrupts Vpx binding to VprBP, these findings link the interaction with VprBP to Vpx ability to facilitate reverse transcription of lentiviral genome in macrophages. To obtain further insight into the role of VprBP, we knocked down its expression in macrophages by RNA interference (RNAi, [25]). As illustrated in Figure 6A, a pool of small interfering RNAs (siRNA) targeting VprBP, but not the control non-targeting siRNAs, severely diminished VprBP expression (compare lane 3 with 1 and 2). Two days following initiation of RNAi macrophages were infected with VSV-G pseudotyped single cycle SIVmac 239 (GFP) reporter virus and the transduction efficiency was assessed 3 days later. Flow cytometry analysis of GFP expression revealed that nontargeting siRNA decreased transduction efficiencies by only approximately 30% and the magnitude of this effect was constant over a wide range of siRNA concentrations (Figure 6B, compare panels 4 and 6 with 2). A similar result was observed with another non-targeting siRNA pool (data not shown). These observations indicate that non-specific engagement of RNAi machinery had only a minor negative effect on macrophage transduction by SIVmac 239. In contrast, RNAi to VprBP decreased transduction efficiency by approximately 10-fold at a lower dose, and 30-fold at a higher dose of the targeting siRNA (compare panel 3 with 4 and 5 with 6). These experiments were repeated 4 times and we consistently observed a decrease in transduction efficiency following RNAi to VprBP, ranging between 6-fold and >100-fold. To further exclude the possibility that the observed resistance of VprBP-depleted macrophages to SIVmac infection is caused by the off target effects of the siRNA pool targeting VprBP, additional experiments were performed using individual VprBP-specific siRNAs (Figure S3). We observed good correlation between the abilities of the four siRNAs to knock down VprBP expression and to disrupt macrophage transduction by SIVmac 239 (GFP) reporter virus. Of note, VprBP-depletion in U2OS cells did not compromise transduction of these cells by SIVmac 239 (GFP) reporter virus regardless of the presence or absence of Vpx (Figure S4). Together these data indicate that VprBP is required for efficient macrophage transduction by SIVmac. If VprBP indeed facilitates macrophage transduction through the interaction with Vpx, we expected the arrest of SIVmac replication in the absence of VprBP and that in the absence of Vpx to be similar in nature. To test this prediction we examined the steady state levels of SIVmac 239 (GFP) late reverse transcription products 72 hours post infection of VprBP-depleted and control macrophage populations, by real time PCR. As expected, the levels of late reverse transcripts were approximately 100-fold lower in VprBP-depleted versus non-targeting siRNA treated, or untreated macrophages (Figure 6C, compare VprBP to scr, or none). Together our data indicate that VprBP has an important role in macrophage transduction by SIVmac and that this function requires Vpx.
Vpx enables efficient transduction of monocyte-derived cells, such as macrophages and DC' s by SIVsm/mac and HIV-2 viruses; however the mechanism that mediates this effect has not been identified. Our findings link this Vpx function to its ability to interact with components of the ubiquitin proteasome system and identify a ternary protein complex - comprising DDA1, DDB1 and VprBP/DCAF1, a putative substrate receptor for Cullin 4-based E3 ubiquitin ligase - as the immediate downstream effector that Vpx uses to promote macrophage transduction. Importantly, the DDB1-VprBP/DCAF1 module was previously shown to participate in a functional Cullin 4 E3 ubiquitin ligase complex [13]. Together, these findings support a model in which Vpx usurps the Cullin 4 E3 ubiquitin ligase utilizing the VprBP/DCAF1 to overcome a block to lentivirus replication upon its entry into monocyte-derived cells. Our data indicate that Vpx acts early following virion entry into macrophages to allow efficient initiation as well as completion of reverse transcription of the incoming SIVmac RNA genomes. This can be clearly seen from a >10-fold decrease in steady state levels of early reverse transcription products and 103-fold decrease in late reverse transcripts upon challenge with Vpx-deficient virions. These phenotypes could result from defects in virion uncoating and/or in its transit into a permissive cytoplasmic compartment. A similar, albeit less dramatic, loss of vpx function phenotypes were previously reported for other SIVsm/mac viral isolates and/or vpx alleles upon infection of monocyte-derived DCs [8]. The finding that Vpx is required for efficient reverse transcription in macrophages was somewhat surprising, because it has been thought that this factor acts at a later stage in the replication cycle by enabling the import of the fully reverse transcribed preintegration complex into the nucleus [3]. Our findings, taken together with these previous observations, indicate that SIVmac replication is restricted by the same mechanism in DCs and in macrophages. Thus, it is important to refocus future studies towards post entry events that precede reverse transcription in these monocyte derived cells. The phenotype of Vpx-deficient virions is reminiscent of that resulting from a block to retrovirus replication imposed by tripartite motif protein 5α (TRIM5α) restriction factors. TRIM5α is a E3 ubiquitin ligase that inactivates the incoming virions, probably by deregulating their uncoating so rapidly that the late reverse transcripts fail to accumulate [26], [27], [28]. Also, the observation that proteasome inhibitors partially rescue reverse transcription of Vpx-deficient viruses in DCs is consistent with the idea that SIVmac virions may be targeted by a TRIM5α-like restriction, or by another E3 ubiquitin ligase in monocyte-derived cells [8]. Whereas these observations raise the possibility that Vpx could act by counteracting TRIM5α, we note that this is not likely, because TRIM5α is expressed in Jurkat T cells [29], which we found not to restrict Vpx-deficient SIVmac virions. How does Vpx facilitate reverse transcription in macrophages via its interaction with VprBP? As mentioned above, a recent study suggested that the replication of SIVmac cells could be restricted, at least in part, by an as yet unidentified E3 ubiquitin ligase [8]. We initially considered that VprBP-linked Cullin 4 E3 complex could be that enzyme and that Vpx counters the restriction by inhibiting its activity. However, our data from RNAi experiments revealed that VprBP is not required for the restriction to occur and, therefore, do not support this possibility. Furthermore, the incoming virions probably contain at most only several hundred Vpx molecules, similar to Vpr, which also is virion recruited through its interaction with Gag p6 [6], [30], [31]. Therefore, it is difficult to envision that the limited amounts of virion-bound Vpx would be able to saturate and inhibit the cellular pool of VprBP-associated Cullin 4 E3 complexes, even by a noncompetitive mechanism. Instead of blocking SIVmac replication, our evidence indicates that VprBP is required for Vpx to overcome the block, implying that Vpx uses VprBP-associated E3 to enable reverse transcription in macrophages. Notably, the same VprBP-associated ubiquitin ligase was shown previously to be targeted by a Vpx paralogue, Vpr, which stimulates the intrinsic catalytic activity ofthis E3 [13]. The findings that both Vpx and Vpr interact with VprBP in a similar manner via their C-terminal regions, and that both interactions lead to post-translational modification of their associated Cullin 4 subunits suggest that Vpx also usurps the VprBP-associated E3, probably to inactivate a cellular factor that inhibits lentivirus replication in macrophages and DC' s. Indeed, viral accessory proteins are known to utilize E3 ubiquitin ligases to direct ubiquitination and proteasomal degradation of cellular proteins that mediate innate immunity to viral infection [32]. Both Vpx and Vpr bind VprBP through similar molecular interactions, yet the functional outcomes are different. Vpr uses VprBP-associated E3 to activate DNA damage checkpoint controlled by the Ataxia-telangiectasia and Rad3-related (ATR) kinase, while Vpx does not have this function and, instead, enables efficient reverse transcription of SIVmac genome in monocyte-derived cells [33]. These different outcomes likely reflect that Vpr and Vpx recruit different sets of substrates for ubiquitination by the same E3 enzyme [34], [35], and that they affect differently the activities of their associated Cullin 4 E3s (see Figure S1). It will be important in the future to identify cellular proteins whose ubiquitination is altered by Vpx and Vpr in order to advance the understanding of these important virulence factors. In summary, our findings provide novel insights into the mechanism by which Vpx enables macrophage infection, as they link this function to Vpx interaction with VprBP and its associated Cullin 4 E3 ubiquitin ligase complex. Further studies of how Vpx manipulates protein ubiquitination through its interaction with VprBP should lead to detailed understanding of the biochemical mechanism that limit replication of primate lentiviruses in monocyte-derived cells, and how it is countered by viruses of the HIV-2/SIVmac/sm lineages. This knowledge in turn will likely lead to the conception of new strategies aimed to prevent the virus from establishing reservoirs in these cells.
pCG expression vectors expressing epitope tagged VprBP/DCAF1, DDB1, DDA1, Cullin 4A, and Vpr proteins of HIV-1 NL43 and SIVmac 239 viruses were described previously [13]. SIVmac 239 Vpx was tagged with hfa- triple epitope tag and subcloned into BABE (puro) and pCG vectors [13]. HIV-2 Rod vpx gene was amplified by PCR from Rod proviral clone [24] kindly provided by Michael Emerman (Fred Hutchinson Cancer Research Center, Seatte). Mutations were introduced using QuikChange XL II kit (Stratagene, La Jolla, CA, United States) and confirmed by DNA sequencing. HEK 293T cells were transfected by calcium phosphate co-precipitation method. Detergent extracts and anti-FLAG immune complexes were analyzed by immunobltting as described previously [36]. FLAG-, HA- and myc- epitope tagged proteins were detected with anti-FLAG M2 (Sigma-Aldrich, St. Louis, MO, United States), 12CA5, and 9E10 monoclonal antibodies (mAb), respectively. The following antibodies were also used: anti-DDB1 (37-6200) from Zymed (Invitrogen, Carlsbad, CA, United States), anti-α-adaptin (AC1-M11) from Alexis Corp (San Diego, CA, United States), anti-Gag SIVmac 251 (13-112-100) from Advanced Biotechnologies Inc. (Columbia, MD, United States) and anti-Vpx 6D2. 6 Vpx hybridoma supernatant. DDA1 and VprBP were detected with rabbit sera raised to recombinant proteins [13]. SIVmac 239 Vpx and its associated proteins were purified from U937 cells stably expressing hfa-tagged Vpx, or HEK 293T cells transiently expressing hfa-Vpx by two sequential immunoprecipitations via FLAG and HA epitope tags, each followed by competitive elution with the appropriate peptide epitope. MudPIT analysis was performed as described previously [13]. Tandem mass (MS/MS) spectra were interpreted using SEQUEST [37] against a database of 82242 sequences, consisting of hfa-tagged SIVmac Vpx, usual contaminants, and 40873 human proteins, as well as, to estimate false discovery rates, randomized amino acid sequences derived from each non-redundant protein entry. Peptide hits from multiple runs were compared using CONTRAST [38]. Vpx and Vpr mutations were introduced into a single cycle SIVmac 239 (GFP) reporter proviral clone containing a frameshift mutation in the env gene constructed by filling in a unique ClaI site [39]. A proviral clone deficient for vpx was constructed by substituting methionine and serine codons at positions 1 and 2 in vpx with threonine and termination codons, respectively, such as not to alter the overlapping vif gene. The second consecutive methionine codon (M62) in vpx was also changed to a termination codon to prevent the possibility that a truncated C-terminal fragment of Vpx protein will be expressed. Mutagenesis was performed with QuikChange XLII kit (Stratagene) using 1. 3 kb PacI-SphI fragment of SIVmac 239 (GFP) provirus [39] that comprises vif and vpx open reading frames, subcloned into pCR 2. 1 vector, as a template. All mutations were confirmed by DNA sequencing and reintroduced into SIVmac 239 (GFP) proviral clones containing a frameshift mutation in env, by exchanging the 1. 3 kb PacI-Sph1 restriction fragment. VSV-G pseudotyped single cycle viruses were produced from HEK 293T cells transiently transfected with proviral clones and a VSV-G expression plasmid. In some experiments vpx-defective SIV were complemented in trans with wild-type or mutant Vpx proteins expressed from cotransfected pCG vectors. Culture medium was harvested 24 hours after transfection, cell debris removed by centrifugation at 7,000 rpm for 10 minutes and virus containing supernatants were then treated with DNAse I (Roche) for 60 minutes at 30°C. Viral particles were partially purified and concentrated by pelleting through 20% sucrose in 10 mM Tris-HCl [pH 7. 4], 100 mM NaCl, 1 mM EDTA cushion at 27,000 rpm for 3 hours. Virion preparations were normalized based on reverse transcriptase assays and/or infectivity to Jurkat T cells, and stored at −70°C. Monocytes obtained from human PBMCs by negative selection for CD3, CD7, CD16, CD19, CD56, CD123 and Glycophorin using Monocyte Isolation Kit II (Miltenyi Biotec Inc. , Auburn, CA, United States), were plated in 24 well plates at 4–7×105 cells/well and differentiated into macrophages by culturing in DMEM supplemented with 10% fetal bovine serum (FBS), Macrophage-Colony Stimulating Factor (M-CSF, 50 ng/ml, R&D Systems, Minneapolis, MN, United States) for 6 days. Cells were fed every alternate day by replacing one half of the cell culture medium with fresh medium. Purity of CD14+ cells obtained by negative selection with the Monocyte Isolation Kit II from Miltenyi usually ranged between 95% and 99% while the final purity of the adherent macrophage population was typically greater than 99. 9%. RNA interference was initiated at day 6 and followed by infections with SIVmac on day 8. Cells were harvested for QPCR analysis of reverse transcription products 18 hours to 72 hours post infection. Flow cytometry analysis of GFP expression was performed 4 days post infection. Macrophages were detached from wells by trypsin treatment, resuspended in 1% paraformaldehyde and GFP expression analyzed by flow cytometry. CD4+ T cells were purified from PBMC using CD4+ T cell isolation kit (Miltenyi Biotec) and stocks were frozen in 107 cell aliquots. Stocks were plated in 5 ml of RPMI 1640 supplemented with 10% FBS, 2 mM glutamine, 10 mM HEPES, pH = 7. 4,50 µM β-mercaptoethanol, and containing phytohemagglutinin (PHA, 10 µg/ml) and recombinant human IL-2 (10 u/ml, Roche) in single wells of a 6 well plate. After 48 hours cultures were diluted into the same medium but without PHA and 5×105 cell aliquots were infected in the total volume of 2 ml in wells of a 24 well plate. Expression of GFP marker protein was quantified 48 hours post infection by flow cytometry. SIV reverse transcription products were quantified by real time PCR on ABI PRISM 7700 SDS. A typical reaction contained 50 ng of DNA isolated with DNAeasy Kit (Qiagen, Valencia, CA, United States) from infected or control cells and SYBR Green PCR master mix in a total volume of 25 µl (Applied Biosystems, Foster City, CA, United States). Early reverse transcription products were amplified with ERT. 2. s (5′-CTTGCTTGCTTAAAGCCCTCTT-3′) and S. ERT. as (5′-CAGGGTCTTCTTATTATTGAGTACC-3′) primers, U3 with U3. SIV. s (5′-ATCATACCAGATTGGCAGGATT-3′) and U3. SIV. as (5′-GAAGTTTGAGCTGGATGCATTA-3′), gag with SIV. GAG. s (5′-ATTAGTGCCAACAGGCTCAGA-3′) and SIV. GAG. as (5′-GCATAGTTTCTGTTGTTCCTGTTT-3′), late with SIV. 1 (5′-AGCTAGTGTGTGTTCCCATCTC-3′) and SIV. 3 (5′-TACTCAGGAGTCTCTCACTCTCCT-3′). Serial dilutions of known amounts of a plasmid containing SIVmac293 provirus, served as a copy number standard to generate standard curves. Macrophages cultured in 24 well plates (Becton & Dickinson, San Jose, CA, United States) were fed with antibiotic free 10% serum containing DMEM 24 hrs before transfections. Cells were transfected with 1–200 pmol aliquots of a control nontargeting pool of siRNA (D-001206-14-05, Dharmacon, Lafayette, CO, United States) or ON-TARGET plus SMARTpool siRNA targeting human VprBP (L-021119-01), or individual VprBP-specific siRNAs (VprBP1 sense: GAUGGCGGAUGCUUUGAUAUU, antisense: UAUCAAAGCAUCCGCCAUCUU; VprBP2 sense: GGAGGGAAUUGUCGAGAAUUU, antisense: AUUCUCGACAAUUCCCUCCUU; VprBP3 sense: ACACAGAGUAUCUUAGAGAUU, antisense: UCUCUAAGAUACUCUGUGAUU; VprBP10 sense: CCACAGAAUUUGUUGCGCAUU, antisense: UGCGCAACAAAUUCUGUGGUU) using Lipofectamine 2000 (Invitrogen) according to manufacturer' s instructions. Briefly, siRNA stocks were prepared in phosphate buffered saline (PBS). Liposomes were formed using 4 µl of Lipofectamine 2000 per well, as recommended by manufacturer (Invitrogen). 4–6 hours post transfection culture medium was replaced with fresh antibiotic-free DMEM supplemented with 10% FBS. U2OS cells were plated at 5×104/well of 12 well plate in 1 ml of DMEM supplemented with 10% FBS and 2 mM glutamine in the absence of antibiotics 24 hours before initiation of RNAi. Cells were transfected with lipofectamine 2000 (4 µl/well) containing indicated amounts of siRNA duplexes in 0. 5 ml of medium, and the medium was replaced with 1 ml of fresh medium 5 hours later. 48 hours following initiation of RNAi cells were harvested for immunoblot analysis of VprBP expression, or infected with SIVmac or HIV-1 derived vectors. Transduction efficiencies were quantified by flow cytometric analysis of GFP expresssion. | Monocyte-derived tissue macrophages play crucial roles in infection by primate lentiviruses. Human and simian lentiviruses of the HIV-2 and SIVsm/mac lineages encode a virion-bound virulence factor termed Vpx. Vpx is required to establish infection specifically of monocyte-derived cells, but the underlying molecular mechanism is unclear. In this study we characterize how the replication of SIVmac is blocked in the absence of Vpx and how Vpx overcomes this block. We find that Vpx is required for efficient reverse transcription of the incoming RNA genome, suggesting that Vpx acts early following virion entry into the macrophage, probably on events linked to virion uncoating and/or reverse transcription. We also identified a Vpx-associated ternary protein complex that is the key mediator of Vpx function specifically in macrophages. This complex links Vpx to the cellular machinery that mediates protein ubiquitination and degradation. Together, we describe the immediate downstream effector, the molecular machinery and a tentative mechanism that lentiviral Vpx uses to enable reverse transcription in macrophages. Our findings should lead to the conception of new strategies to control macrophage infection by human and simian lentiviruses. | Abstract
Introduction
Results
Discussion
Materials and Methods | virology/virulence factors and mechanisms
virology/immunodeficiency viruses
infectious diseases/hiv infection and aids
biochemistry/macromolecular assemblies and machines
chemical biology/protein chemistry and proteomics
cell biology/chemical biology of the cell | 2008 | Lentiviral Vpx Accessory Factor Targets VprBP/DCAF1 Substrate Adaptor for Cullin 4 E3 Ubiquitin Ligase to Enable Macrophage Infection | 10,499 | 304 |
The dynamics of dengue virus (DENV) circulation depends on serotype, genotype and lineage replacement and turnover. In São José do Rio Preto, Brazil, we observed that the L6 lineage of DENV-1 (genotype V) remained the dominant circulating lineage even after the introduction of the L1 lineage. We investigated viral fitness and immunogenicity of the L1 and L6 lineages and which factors interfered with the dynamics of DENV epidemics. The results showed a more efficient replicative fitness of L1 over L6 in mosquitoes and in human and non-human primate cell lines. Infections by the L6 lineage were associated with reduced antigenicity, weak B and T cell stimulation and weak host immune system interactions, which were associated with higher viremia. Our data, therefore, demonstrate that reduced viral immunogenicity and consequent greater viremia determined the increased epidemiological fitness of DENV-1 L6 lineage in São José do Rio Preto.
The spread of dengue (DENV) over the past several decades has made this arbovirus infection a major public health concern of global impact [1,2]. The disease has a complex epidemiological pattern and a high economic impact globally and is considered hyperendemic, i. e. dengue fever has a high incidence and/or prevalence rate affecting all groups equally [3,4]. Worldwide, it is estimated that 390 million new DENV infections occur annually [1], and this number will likely increase with the creation of new vector habitats due to climate change [5]. Because of their wide distribution, particularly in urban and peri-urban environments in tropical and subtropical regions, mosquitoes of the Aedes genus are the main vectors of this disease [6,7]. DENV can lead to a wide spectrum of clinical manifestations that are classified by the World Health Organization (WHO) as dengue without warning signs, dengue with warning signs and severe dengue [8]. Previous infections with a heterologous type are usually, but not exclusively, associated with progression to more severe disease [1]. There are four genetically distinct serotypes of DENV (DENV-1 to -4) that share similar epidemiological features [6]. Each serotype is divided into distinct groups defined as genotypes, which in turn subdivided in different lineages [9,10]. The circulation of the virus is characterized by frequent lineage turnover, in which a lineage of circulating viruses is usually replaced by a new lineage, a well-documented phenomenon known as clade replacement (CR). CR can lead to an increase in the number of cases and in the severity of the disease [10–17]. In previous CR events, an established lineage circulating for a number of years in a given population is replaced when a new lineage is introduced. This replacement is usually followed by the extinction of the previous lineage after a period of co-circulation of both lineages [10–17]. Since the mid-1980s, DENV-1 has been circulating in Brazil. All Brazilian DENV-1 isolates described to date belong to genotype V, which is subdivided into three distinct clades (lineages 1,3 and 6) [18–20]. These lineages were introduced into Brazil at different times, and CR or the co-circulation of different lineages has been observed in the country [18–20]. This pattern of co-circulation of different lineages was also observed in São José do Rio Preto (SJRP), São Paulo, Brazil. Here, we combine phylogenetic, molecular and immunological analyses to describe the epidemiological dynamics of two Brazilian DENV-1 lineages (L1 and L6) circulating in SJRP from 2008 to 2015 to provide a more precise understanding of the role of fitness in lineage dynamics with the persistence of L6 even after the introduction of L1 without CR. The reduced immunogenicity of L6, which contributed to B and T cell-specific immune response evasion, appears to have played a prominent role in its dominance over time.
We used 20 envelope sequences (1,485 nucleotides) of DENV-1 isolates obtained from patients in SJRP, from 2008 to 2012, for the phylogenetic analyses. The results indicated that the isolates were subdivided into two lineages within genotype V. Isolates 59/2011,287/2011,354/2011,395/2011,422/2011,430/2011,437/2011,442/2011,492/2012 and 516/2012 were grouped within one lineage previously called L1 or lineage II [13,18] the most recent common ancestor (MRCA) for those species dates back to approximately 2008 (95% BCI = 2006–2009). Ten isolates from SJRP were grouped in another lineage, previously called L6 or lineage I [13,18]: 365/2008,09/2009,88/2010,64/2011,205/2011,384/2011,387/2011,484/2012,531/2012 and 552/2012. These isolates share an MRCA from approximately 2007 (2006–2008) (Fig 1A). When deduced protein envelope sequences were analyzed, a total of four amino acid (aa) substitutions were observed between the two groups of isolates from SJRP. Those aa substitutions were observed at positions 338,394 (located in domain III), 428 and 436 (located in the stem loop region). The aa observed in SJRP isolate sequences within L1 were as follows: 338, serine; 394, arginine; 428, valine; and 436, valine; whereas SJRP isolates within L6 presented leucine, lysine, leucine and isoleucine residues, respectively. A comparison between the complete genome sequences of the L1 and L6 lineages from SJRP (287/2011 and 484/2012, respectively) revealed 56 aa differences (Fig 1B). The L6 lineage was first identified in SJRP in 2008, and the L1 lineage was detected only in 2010. These two lineages of DENV-1, genotype V, co-circulated in SJRP from 2010 until 2012. Based on sequencing or genotyping analysis by Taqman-based qPCR, 64 serum samples were identified as infected by L1 from 2010 to 2012, whereas 102 samples obtained from 2008 to 2015 were identified as L6, resulting in a total of 166 discriminated samples. In SJRP, L6 became the dominant circulating lineage after 2013 (Fig 1C).
To understand the forces that drive DENV epidemics, it is crucial to develop efficient control methods for this disease. The introduction of new lineages and CR is important for maintenance of the disease in a given population. CR is a well-documented event that occurs in many epidemics when a new DENV strain emerges and displaces the endemic strain, even between viruses of the same serotypes and genotypes. This process usually causes an increase in both the number of cases and the severity of the disease [15,16]. Although relatively common, the forces that dictate whether CR will occur are not completely understood. Genetic variations may alter viral fitness; however, the interplay of this factor with epidemiological factors has not been defined. Since 2008, the L6 lineage of DENV-1 (genotype V) has circulated in SJRP. The duration of the circulation of this lineage in the population allowed us to consider it an endemic lineage. In mid-2010, a different Brazilian lineage arose: L1. This newly introduced lineage appeared to have improved viral fitness compared to L6; thus, L1 was expected to replace L6 as the dominant strain. From the emergence of the L1 lineage until 2012, the epidemics in SJRP appeared to follow the path of other epidemics, and the CR event would occur. However, the L1 lineage did not displace L6, but rather both co-circulated without a clear predominance. L1 started to decline until its complete disappearance in 2013. This finding raised questions about differences in the fitness between the lineages, which were investigated by our group. Lineages of the same serotype may present different characteristics in transmissibility, virulence and antigenic properties, some of which may lead to increased fitness and could be related to a greater epidemic potential [15,16,23]. The viral fitness is defined by the capacity of a virus to produce infectious progeny in a given environment, whereas the capacity of a virus to become dominant in a given region has been called epidemiological fitness [15]. The latter is determined by a combination of factors, including the genetics of the viral strain, transmission potential by the vector and human system interactions [24]. Investigations of the factors involved in lineage replacement have revealed important implications for advancing our understanding of DENV epidemiology, evolutionary dynamics and control [11]. Viral fitness is an important contributing factor to CR and, on some occasions, it can account for the persistence of the strain. In other DENV epidemics, improved replicative fitness of the emerging strain has been responsible for the dominance of the newer lineage. For example, the replicative fitness advantage of the NI-2B lineage over the NI-1 lineage in Ae. aegypti could have contributed to the CR event, resulting in the dominance of NI-2B in Managua [16]. Replacement of the NI-1 lineage by the NI-2B lineage was also associated with an increase in disease severity. In the SJRP epidemic, differences in the replicative fitness could have accounted for the persistence of L6 even after the introduction of the new lineage (L1). However, the L1 had higher replication rates than L6 in mosquito, human and non-human primate cell lines. It also had an increased fitness in two different populations of Ae. aegypti mosquitoes and thus could have a higher potential for transmission than L6. Unfortunately, it was not possible to analyze viral levels in mosquitos collected at that time. In theory, the coinfection of vectors with both lineages can result in competition and can accurately represent the natural transmission cycles. However, the superior viral fitness of L1 in mosquito cell culture indicates that coinfections do not have any effect on viral fitness. In fact, all tests conducted by our group revealed superior viral fitness of L1, in vitro or in vivo. Therefore, the better replicative fitness L1 would induce its dominance and not its replacement by L6. Some studies theorize that in vitro infections can sometimes fail to predict the actual dynamics of DENV epidemics [16]. Therefore, differences in replicative fitness did not explain the observed pattern of L6 persistence. Evasion of interferon responses by the virus could be a factor underlying the maintenance of the L6 lineage in the population. This has been identified as a determinant of epidemiological fitness in the lineage dominance of DENV-2 (PR-2B) in Puerto Rico [15]. During flavivirus infections, short flavivirus RNAs (sfRNAs) are produced by the incomplete degradation of viral RNA by the host-cell exonuclease Xrn1. These sfRNAs are associated with the pathogenesis of flaviviruses [25]. The high expression of sfRNAs could have inhibited type I interferon responses and facilitated evasion of the immune response. In a study conducted in Puerto Rico, the PR-2B lineage produced elevated expression of subgenomic flavivirus RNA (sfRNA) relative to genomic RNA (gRNA) during replication. Notably, sfRNA could bind TRIM25 and inhibit the IFN response. Thus, higher sfRNA: gRNA ratios are associated with superior epidemiological fitness [15]. Although L6 had higher sfRNA: gRNA ratios than L1, which in turn could increase its epidemic potential, it did not appear to be related to IFN inhibition, as previously reported [15]. Both lineages appeared to be incapable of inhibiting the IFN response and produced similar levels of type I IFN in vitro and in vivo, suggesting this was not the mechanism underlying the differences observed. Previous antibody immunity to dengue could also play a role in epidemic dynamics. For example, the immunological status of the population could interfere with L1 propagation through an antibody-dependent enhancement (ADE) phenomenon. Waning immunity due to a prior DENV infection can alter the outcome of the present infection, as observed in Nicaragua [2]. High titers of pre-infection cross-reactive neutralizing antibodies drastically reduce the probability of a severe disease in a second infection, although this protective effect is not observed against all DENV serotypes [26]. However, no significant differences in the frequency of anti-dengue IgG antibodies were found in our patients. Because previously reported mechanisms were unable to explain the persistence of L6, we evaluated various aspects of the immune response in order to provide an alternative explanation to our findings. To assess possible differences in immunological responses, a panel of 29 cytokines, chemokines, adhesion molecules and growth factors were measured in plasma of patients. Dengue infection was associated with the increase of several cytokines in serum of patients. Overall infection with the L6 lineage was associated with greater increases of cytokines with anti-inflammatory (IL-1RA and IL-13) or Th2-like activity (IL-13 and CCL11). In contrast, infection with the L1 lineage was associated with greater increase of cytokines with Th1/Th17–like activity (IL-12 and IL-17) and IL-7, a cytokine that promotes lymphocyte development in the thymus and maintains survival of naive and memory T cell homeostasis in the periphery. The exact role of these cytokines in the context of dengue infection is not precisely know but these results clearly show that the L1 lineage had a tendency to generate immune responses usually associated with resistance to dengue infection whereas the L6 lineage the tendency to generating cytokines with anti-inflammatory activity or that block Th1/Th17 responses. The latter studies suggested that immune responses to L1 were enhanced and more pro-inflammatory and led us to investigate in greater detail adaptive immune responses in patients infected with the L1 or L6 lineages. In silico analysis suggested that the L1 lineage was potentially more immunogenic than the L6 lineage. These predictions were confirmed by studies in mice that showed increased immunogenicity of L1, as assessed by greater B and T cell activation. Unfortunately, we did not have access to PBMCs derived from L1 or L6 infected patients, and only the sera were available. Therefore, it was not possible to assess directly T cell recall responses to L1 and L6 viruses. However, experiments in human PMBCs infected with either lineage showed that the L1 lineage induced stronger responses than those infected with L6 viruses. Altogether, he data suggest that different DENV isolates might induce distinct levels of B and T cell activation. L1-induced response was mostly associated to IFN-γ production, whereas L6 induced activation was driven to inflammatory IL-8 secretion. Distinct T cell phenotype and function, with increased T cells activation and IFN-γ production might then be associated to disease control [27–30]. Altogether, these studies suggest that the adaptive immune responses in infected individuals triggered by L1 were stronger than those triggered by L6. Our prediction would be that differences in immune response would necessarily have to associate with altered viral loads, if these responses were to affect the likelihood of one strain to override the other. Indeed, patients infected with the L6 lineage, the one with significantly decreased immune activation, were those with the higher viral loads. We would expect that the high viral L6 loads would enhance the probability of transmission of the L6 lineage to the vector and, consequently, to another host. The current dogma indicates that, after infection with a certain DENV serotype, only a heterologous DENV serotype can cause infection in the same individuals. However, studies using non-human primates have indicated that new inoculations with either the same or different genotypes of DENV-2 can cause a persistent boost in neutralizing antibodies [31]. Because L6 only weakly stimulates B and T cells, it may not increase immunological memory, nor may it induce the development of neutralizing antibodies in sufficient titers to protect against a new exposure. Although further work with specific experiments are needed to strengthen this evidence, it appears that neutralizing antibodies may quickly control infections, preventing more severe disease; however, they may not avoid future infections. A similar mechanism has been proposed for studies in Nicaragua [31] and may provide an explanation for the maintenance of L6 for such a long period in the population. High viral loads of L6 enhance the probability of vector transmission of the L6 lineage. However, despite the better replicative fitness, the L1 lineage appears to elicit a stronger immune response, preventing broader propagation of this lineage due to the extinction of the susceptible population. Based on our data, the absence of CR together with the superior epidemiological fitness of L6 in SJRP was a result of the human immune system functioning as a bottleneck that favored the L6 lineage to achieve a broader distribution, even with lower viral fitness. The model is summarized in Fig 8. The inability of the L1 lineage to replace the endemic L6 lineage in this city shows that the interplay between replicative, immunological and epidemiological features that affect the dynamics of viral propagation is far more complex than previously suspected. In this case, the lower viral immunogenicity for B and T cells associated with host immunological factors counteracts the superior viral fitness, contributing to the lineage dominance.
The panel of 186 DENV-1 viruses used in the present study was collected from symptomatic patients from 2008 to 2015 in a public healthcare facility in SJRP (São Paulo, Brazil) as part of the flavivirus and hantavirus surveillance program in the city. This study was reviewed and approved by the Human Research Ethics Committee of Faculdade de Medicina de São José do Rio Preto (CAAE: 02078812. 8. 00005414). All the samples are obtained from an existing collection in the laboratory (LPV-Dengue 2008–2015) and all of them were already anonymized. Blood samples (buffy coats) from healthy donors were obtained anonymously from the Hemotherapy Service at the Hospital Universitário Clementino Fraga Filho (HUCFF) of Universidade Federal do Rio de Janeiro (UFRJ). The study protocol was approved by the Experimental Ethics Committee of UFRJ (Permit Number: IMPPG 025), and the review board waived the need for informed patient consent. All animal work was performed in accordance with the Fiocruz Animal Use Committee (protocol P-60/14-4; license number LW-30/15). Fiocruz personnel are required to adhere to applicable federal, state, local and institutional laws and policies governing animal research, including the regulations from the Brazilian Council of Animal Use Control (CONCEA - 3rd Edition, published at Sep, 26,2016), Federal Law 11794/08 and Protocols for Animal Use—Oswaldo Cruz Research Foundation (ISBN: 85-7541-015-6). We observed predominantly dengue without warning signals. All the samples used in this study were collected untill 5 days of the symptoms onset. The serum samples were subjected to molecular biological and serological diagnostic tests for dengue. Initially, DENV infection was confirmed using reverse transcription-PCR (RT-PCR) and multiplex nested PCR (M-N-PCR) assays as described previously [32], which distinguishes the four serotypes. Ninety-six DENV-1-positive samples were tested for anti-dengue IgG antibodies, as recommended by the SERION ELISA classic Dengue IgG test kit (Virion Serion). Twenty DENV-1-positive samples were used to amplify the envelope gene sequence, followed by nucleotide sequencing using the Sanger-based method with a previously described primer [33]. Viral RNA was extracted using the QIAamp Viral RNA Mini kit (Qiagen) as recommended by the manufacturer. First-strand cDNA was synthesized using the Superscript III First Strand Synthesis System (Invitrogen) following the manufacturer’s instructions with primer d1a16. PCR was performed to amplify a 1,855-bp fragment, of which 1,485 bp corresponded to the entire DENV-1 E gene. The reaction consisted of 2 μL of cDNA, 5 μL of 10X Accutaq LA buffer, 2. 5 μL of dNTPs (10 mM/μL), 1 μL of DMSO 2%, 1 μL of primers d1s3 and d1a17 (10 μM), 0. 5 μL of Accutaq LA DNA polymerase (5 U/μL; Sigma-Aldrich) and DEPC-treated water. The reactions were submitted to the following cycle conditions: 98°C for 30 sec, followed by 30 cycles of 94°C for 15 sec, 50°C for 20 sec and 68°C for 1 min and 30 sec. A final extension step was performed at 68°C for 10 min. An analysis of the amplicons was performed by electrophoresis on a 1% agarose gel. The PCR product (40 μL) was purified using 2. 8 μL of 3 M sodium acetate and 1. 2 μL of cold absolute ethanol. The samples were stored overnight at -20°C or for 1 h at -80°C and subsequently centrifuged at 16,100 x g for 20 min. The pellet was washed with 200 μL of 70% ethanol and centrifuged at 16,100 x g for 10 min. The dried pellet was resuspended in 20 μL of Milli-Q water. Twenty nanograms of purified PCR products were used as templates in 20 μL of cycle sequencing reactions using 2 μL of 1X Sequencing buffer, 2 μL of BigDye Terminator v. 3. 0 (Applied Biosystem), 1 μL of forward and reverse primers (3. 2 μM) (see S1 Table) and DEPC-treated water; these samples were submitted to 96°C for 1 min followed by 25 cycles of 96°C for 10 sec, 50°C for 5 sec and 60°C for 4 min. Precipitation of sequencing reactions was performed using ethanol/EDTA as recommended by the BigDye Terminator kit v. 3. 0 (Applied Biosystems) protocol. The samples were resuspended in 10 μL of Hi-Di Formamide (Applied Biosystems) and analyzed with ABI PRISM 3130 equipment (Applied Biosystems). The quality of the sequences was analyzed using Sequencing Analysis 5. 2 software (Applied Biosystems). The consensus sequence was edited using Accelrys Gene v. 2. 5 (Accelrys). Nucleotide sequences were then aligned with the previously published E gene sequence from GenBank (GU131863. 1) using MEGA 6. 0. 6 Molecular Evolutionary Genetics Analysis (http: //en. bio-soft. net/tree/MEGA. html). To obtain insight into the genetic relationship among DENV-1 strains, envelope sequences obtained from different DENV-1 isolates from SJRP and other geographic sites were aligned using ClustalW [34], taking into account the codon sequences. Amino acid sequences were predicted, and substitution patterns were analyzed. Phylogenetic and coalescent analyses were conducted using BEAST package v. 1. 8 with Markov Chain Monte Carlo (MCMC) algorithms [35]. Input files for BEAST were generated using BEAUTI v. 1. 8. 1 [35], and the year each strain was isolated/obtained was used as a calibration point. Analyses were performed using the General Time Reversible nucleotide substitution model with four gamma categories (GTR + 4G), the Bayesian Skyline method [36] and a relaxed (uncorrelated lognormal) molecular clock. Two independent runs (100 million chains, discarding the first 10 million steps) were run, and parameters and trees were sampled every 10,000 steps. The convergence of parameters was checked with Tracer v1. 6. 0 [37], and uncertainties were addressed as 95% Bayesian Credible Intervals (BCI). Using Tree Annotator v. 1. 8. 1 [38], a maximum clade credibility (MCC) tree was generated and then visualized in Figure Tree v. 1. 4. 2 [38]. In addition, full-length sequences of viral RNA genomes from SJRP were sequenced using next-generation sequencing with Illumina MiSeq System (Illumina) as described previously [39]. Briefly, isolates were subjected to RNA extraction using the QIAamp Viral RNA Mini kit (Qiagen), followed by quantification using a PicoDrop (Picodrop Limited). RNA was treated with DNAse I (Sigma-Aldrich), and reverse transcription-PCR was performed with random primers (50 ng; Invitrogen) using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems), according to the manufacturer' s instructions. After sequencing, the obtained sequences were assembled and edited using Geneious v. 7. 1. 4 (Biomatters Ltda), and polyprotein sequences were aligned and translated to compare aa substitutions. Two sequences (287/2011 and 484/2012) were selected to represent the L1 and L6 lineages, respectively. When it was not possible to obtain an amplicon of the envelope gene region or complete genomic sequences, lineage discrimination of DENV-1-positive samples was performed using the TaqMan Real-Time PCR genotyping assay and the AgPath-ID One-Step RT-PCR kit (Applied Biosystems) with two primers and probe sets for the envelope (2021_F, 2021_R and 2021_P1) and NS5 (8587_F, 8587_R and 8587_P2) regions, as shown in S2 and S3 Tables. For the genotyping assay validation, twenty samples previous grouped into L1 or L6 lineages by phylogenetic analyses were selected and the experimental validation was performed under blind conditions. All of the samples used were correctly discriminate. The One-Step qRT-PCR reaction was performed using two different master mixes and consisted of 7 μL of RNA sample, 12. 5 μL of 2X RT-PCR buffer, 1 μL of forward and reverse primers (20 μM), 0. 75 μL of probe (10 μM), 1 μL of 25X RT-PCR enzyme mix and nuclease-free water in a total volume of 25 μL per reaction. To identify the L1 lineage, primers 2021_F and 2021_R and probe 2021_P1 were used; 8587_F, 8587_R and 8587_P2 were used for lineage L6 identification. The reactions were subjected to the following cycle conditions, with data collections at 60°C: 50°C for 10 min, 95°C for 10 min, 60°C for 30 sec and 95°C for 10 min, followed by 50 cycles of 95°C for 15 sec and 60°C for 1 min. A final extension step was conducted at 60°C for 30 sec. Genotyping was performed using human sera and standardized samples (to provide standard curves) to allow the relative quantification of virus levels. All qRT-PCR reactions were performed using a StepOne Real-Time PCR System (Applied Biosystems). Peripheral blood mononuclear cells (PBMCs) were obtained after centrifugation of buffy coat samples over a ficoll-hypaque gradient. Moreover, human brain microvascular endothelial cells (HBMECs) were stably transfected with the reporter vector pISRE-Luc-Hygro containing an NdeI-Bst1107 site of pISRE-Luc (Stratagene) and cloned into vector pCEP4 (Invitrogen). Cells then referred to as HBMEC-ISRE-Luc were kindly provided by Dr. Laura Helena Vega Gonzales Gil, CPqAM, FIOCRUZ, Recife-PE, Brazil. Both human primary cells were cultured in RPMI-1640 medium (Cultilab) supplemented with 10% fetal calf serum (FCS; Gibco) (complete medium) at 37°C in a 5% CO2 atmosphere. Mosquito, human and non-human primate cell lines were also used for the in vitro assays. C6/36 cells (ATCC) were cultured in Leibovitz' s medium (L-15; Cultilab) and Aag-2 (kindly provided by Dr. João Trindade Marques, UFMG, Brazil) in Schneider’s insect medium (Sigma-Aldrich) at 28°C. Vero E6 and LLC-MK2 cells (ATCC) were cultured in Minimum Essential Medium (MEM; Cultilab) and HepG2 cells (ATCC) in Dulbecco' s Modified Eagle' s Medium (DMEM; Cultilab) at 37°C in a 5% CO2 atmosphere. All culture mediums were supplemented with 1% fetal bovine serum (FBS; Gibco) for maintenance or 10% for expansion, excluding Aag-2, for which 8% FBS, 10 U/mL of penicillin, 10 g/mL of streptomycin and 250 μg/mL of amphotericin B were used (Gibco). Initially, samples were selected to represent each lineage, and viral isolation was performed based on previous investigations [40]. Briefly, viruses selected from L1 or L6 DENV-1 human sera were diluted 1: 10 in L-15 and used to inoculate C6/36 cells, which were then incubated at 28°C for 7–10 days. Successful isolation was confirmed by RT-PCR of the culture supernatant, as previously described for the sequencing reaction, followed by PCR. PCR was performed to amplify an 1,855-bp fragment using 2 μL of cDNA, 5 μL of 10X buffer, 2 μL of dNTPs (10 mM/μL), 2 μL of primers d1s3 and d1a17 (10 μM), 1 μL MgCl2,0. 25 of Taq DNA polymerase (5 U/μL; Sigma-Aldrich) and DEPC-treated water. The reactions were subjected to the following cycle conditions: 94°C for 2 min, followed by 30 cycles of 94°C for 45 sec, 56°C for 45 sec and 72°C for 45 sec. A final extension step was performed at 72°C for 10 min. Amplification was confirmed by electrophoresis on a 1. 5% agarose gel. Titration was determined by flow cytometry to calculate the number of infectious particles/mL (IP/mL), as described previously [41] with modifications, using FACSCalibur (BD Biosciences) equipment. The adaptations were cells fixed in 4% paraformaldehyde and permeabilized with 0. 1% triton X-100. Viral stocks from the third passage were used for the experiments. Approximately 0. 05x106 cells (C6/36, Aag-2, Vero E6, LLC-MK2 and HepG2) were plated in each well of a 24-well plate 24 h prior to infection. Five cell lines were infected with L1 or L6 isolates at a multiplicity of infection (MOI) of 0. 1 for 1: 30 h in triplicates. The cells were then washed with 1X phosphate-buffered saline (PBS) to remove unabsorbed virus and then incubated in 1 mL of maintenance medium. The supernatants were harvested at 24,48 and 72 hpi for relative quantification using the SYBR Green Real-Time PCR assay with the GoTaq qPCR Master Mix kit (Promega) and primers Den_F (5’-TTAGAGGAGACCCCTCCC-3’) and Den_R2 (5’-GAGACAGCAGGATCTCTGG-3’), as previously described [42]. Total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s protocol. First-strand cDNA was synthesized using the Superscript III First Strand Synthesis System (Invitrogen) following the manufacturer’s instructions with primer Den_R2. The qRT-PCR reaction was also performed according to the manufacturer’s protocol. The results were obtained using a standard curve and analyzing the melting curve (~ 85°C) to approximately CT 35, according to the minimum information for the publication of quantitative real-time PCR experiments (MIQE). Ae. aegypti eggs from populations PPCampos (captive; maintained approximately 15 years in an insectary at the Laboratório de Entomologia Médica, CPqRR, FIOCRUZ, Belo Horizonte-MG, from Campos dos Goytacazes-RJ) and Dom Pedro (wild; collected in 2014 in district Dom Pedro of Manaus-AM) were used in this study according to a previous description [43]. Briefly, the larvae were hatched in an insectary at a temperature of 28°C and relative humidity of 80%, and infections were performed using 3 to 5-day-old female mosquitoes (Dom Pedro from the F2 generation) using a glass feeding device containing 2/3 of blood mouse (Mus musculus) and 1/3 of C6/36 cells suspension infected with either L1 or L6 lineages. The mean viral titer used for infection with L1 or L6 isolates was 5×105 TCID50/mL. Infected PPCampos (n = 80) and Dom Pedro (n = 60) females were maintained in cages with 10% glucose solution until day 14 after the experimental infection (complete extrinsic incubation period). They were then dissected, and total RNA was extracted from their bodies and heads (with attached salivary glands) using TRIzol (Invitrogen) as described previously [40], followed by one-step qRT-PCR [43]. The infection rate (IR) was then calculated as the individual proportion of all experimentally infected mosquitoes, in which DENV was detected in the body. Similarly, the vector competence (VC) was calculated, in which DENV was detected in the head (indicating the virus escaped the midgut barrier, completing its life cycle). However, the disseminated infection rate (DIR) is the proportion of DENV-infected mosquito heads of all infected mosquitoes with virus dispersed in the body (DIR = VC/IR). Aag-2 cells were coinfected with L1 and L6 isolates mixed at an equal ratio (1: 1) at an MOI of 0. 1, and the supernatants were harvested at 24,48 and 72 hpi, as previously described for the growth curves. Total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s instructions, and the relative amounts of L1 and L6 viruses in each dual infection were calculated based on the developed genotyping assay. HepG2 cells were infected at an MOI of 1. 0, and the supernatants and cells were harvested at 24,48 and 72 hpi. The sfRNA: gRNA ratio of the cells was obtained by real-time PCR using GoTaq qPCR Master Mix (Promega) with primers D1GSF, D1SF and D1GSR (described in S4 Table). Cells total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s protocol. First-strand cDNA was synthesized using M-MLV Reverse Transcriptase (Invitrogen) following the manufacturer’s instructions with primers D1GSR. The qRT-PCR reaction was performed using two different mixes, and it consisted of 5 μL of cDNA, 12. 5 μL of 2X GoTaq qPCR Master Mix, 2 μL of forward and reverse primers (10 μM), 0. 25 μL of CXR reference dye and nuclease-free water to a final volume of 25 μL per reaction. To quantify gRNA, primers D1GSF and D1GSR were used; otherwise, D1SF and D1GSR were used for sfRNA quantification. The reactions were subjected to the following cycle conditions, with data collected at 55°C and 60°C: 95°C for 5 min, followed by 40 cycles of 95°C for 15 sec and 55°C for 1 min. A final dissociation step was conducted at 60–95°C. The results were obtained using a standard curve and by analyzing the melting curve (~ 84. 5°C) to approximately CT 35. Quantification of gRNA and sfRNA levels was performed as described previously [44]. In addition, IFN-α1/13 production was measured in the supernatants of infected HepG2 as recommended for sample cultures by the Human IFNA1/Interferon Alpha-1/13 ELISA Kit (RAB0541; Sigma-Aldrich). HBMEC-ISRE-Luc cells were mock-treated or infected with DENV-2 (strain 16681), L1 or L6 DENV-1, in the presence or absence of IFN-β (1000 U– 2 ng/mL; PeproTech). After 48 hpi, the cells were lysed using cell culture lysis reagent (CCLR; Promega), and the supernatants were collected after centrifugation. The luciferase activity was measured by mixing 20 μL of cell lysate with 50 μL of Luciferase Assay Reagent (Promega), and the light produced was measured using a GloMax 96 Microplate Luminometer (Promega). The results are shown in relative light units (RLUs). Seventy-two samples of DENV-1 human sera (L1 or L6) were subjected to a selected panel to measure cytokines, chemokines, adhesion molecules and growth factors (EGF, VEGF, TNF-β, TNF-α, MIP-1β, MIP-1α, MCP-1, IP-10, IL-17, IL-15, IL-13, IL-12 (p70), IL-12 (p40), IL-10, IL-8, IL-7, IL-6, IL-5, IL-3, IL-2, IL-1RA, IL-1β, IL-1α, IFN-γ, IFN-α2, GM-CSF, G-CSF and CCL11) using the MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel–Premixed 29 Plex–Immunology Multiplex Assay (HCYTMAG-60K-PX29; Millipore) by the Luminex system in the MAGPIX instrument, according to the manufacturer’s instructions. In silico analyses of the putative antigenic potential of L1 and L6 lineages were performed as previously described [21]. Briefly, 20 amino acid sequences encoding the DENV-1 polyprotein were first aligned using the Multalin interface (http: //multalin. toulouse. inra. fr/multalin/multalin. html) with default parameters. The consensus sequence coding for Capsid, Envelope and NS1 proteins was submitted to the BepiPred 1. 0 server (http: //www. cbs. dtu. dk/services/BepiPred/) for Linear B epitope prediction. The consensus sequence coding for DENV-1 polyprotein was run into the NetCTL server (http: //www. cbs. dtu. dk/services/NetCTL/) to predict T-cell epitopes. Both algorithms were run using default settings. The Allele Frequency Net Database (http: //www. allelefrequencies. net/) was used to select the most predominant HLA classes in the Southeast of Brazil and to set up the NetCTL server. The mean value of the epitope propensity scores for each sequence was then classified and plotted according to its potential immunogenicity. Eight-week-old male C57BL/6 mice were divided into 3 groups (5 animals/group) and intraperitoneally immunized with 5x105 IP per mouse of DENV-1: L1 (group 1) or L6 (group 2). Mock C6/36 injections in L-15 medium (group 3) were used as controls. At 7 days p. i. , the spleens from each group of mice were extracted, immersed in cold RPMI 1640 medium (Cultilab) and macerated. After centrifugation at 377 x g for 10 min, the erythrocytes were lysed in 9 mL of cold water. Lysis was stopped by adding 1 mL of 1. 5 M PBS. The spleen cells were collected after centrifugation and resuspended in 1 mL of RPMI supplemented with 10% FBS (Gibco). To assess viability, an aliquot of cells was diluted 1: 20 in 0. 4% Trypan Blue solution (Invitrogen) and counted using a Neubauer chamber. A panel of monoclonal antibodies (Becton Dickinson, USA) specific for CD4+, CD8+ and CD21+ lymphocyte subsets and activation markers (CD25, CD69, CD28 and CTLA-4) was then used for cell staining (see S5 Table). Briefly, 1x106 spleen cells were distributed in 96-well polystyrene conical bottom microwell plates and centrifuged at 377 x g for 10 min. After 30 min of incubation with the antibodies at 4°C, the cells were washed twice with 0. 15 M PBS and fixed in 2% paraformaldehyde in PBS. Flow cytometry acquisition of 30,000 events/tube and analysis were performed using FACSCalibur (BD Biosciences) equipment. Distinct gating strategies were used to analyze the lymphocyte subsets (CD4+, CD8+ T-cell subsets and CD21+ B cells) with FlowJo software. The T lymphocyte CD8+ and CD4+ subsets were selected from the CD3+ cell population, and B lymphocytes were selected using the CD21+ marker. The frequency of cells was determined using quadrant statistics. Limits for the quadrant markers were always set based on negative populations and isotype controls. To analyze the CD8+ lymphocytes, the quadrants were always set for CD8 high populations to avoid including CD8 low NK cells. The results were expressed as percentages of cells for the different gated lymphocyte subpopulations analyzed. The expression of activation markers was evaluated inside each lymphocyte population by measuring the mean intensity fluorescence (MIF), which represents the number of molecules per cell. The human PBMCs (2x105 cells) were cultured with L1 or L6 DENV-1 at an MOI of 1. 0 for 2 h. Control cultures were performed using supernatant from noninfected C6/36 cells (mock-infected). At 48 hpi, the cells were harvested and incubated with CD38-APC, HLA-DR-FITC and CD3-Pacific Blue (eBiosciences). The frequency of CD38+/HLA-DR+ among the total PBMCs, CD4+ or CD8+ cells for the analysis of PBMC activation was determined by flow cytometry using FACSCalibur (BD Biosciences) equipment and FlowJo software. Additionally, after 12 days of culture, the supernatants were harvested and the IgM or IgG levels were analyzed by capture ELISA. Briefly, the ELISA plates were incubated with anti-human IgM or IgG antibodies at 3 mg/mL (Sigma-Aldrich) overnight at 4°C. The plates were blocked with PBS containing 10% FCS for 2 h at 37°C and washed with PBS-0. 05% Tween 20 (PBS-T), and serial dilutions of the supernatant samples were added and incubated overnight at 4°C. Serial dilutions of purified human IgM or IgG were also added to generate a standard curve. The plates were incubated with alkaline phosphatase (AP) -conjugated anti-human IgM or IgG (1 mg/mL; Invitrogen) for 2 h at 37°C, washed and developed using pNPP substrate (Sigma-Aldrich). The reaction was read at 405 nm using an ELISA reader (Bio-Rad Laboratories). Also, the secretion levels of IFN-γ, IL-6 and IL-8 were evaluated by ELISA, according to manufacturer’s protocol (PeproTech). The comparative analysis of the two lineages was performed using Student’s T test and Chi-squared test. Fisher’s exact test was used to compare proportions, and the Mann-Whitney nonparametric test was also applied. P values ≤0. 05 were considered significant. All analyses were performed using GraphPad Prism version 6. 01. The envelope sequences of the DENV-1 isolates analyzed in this study are in GenBank under the following accession numbers: KT438562, KT438564, KT438565, KT438566, KT438567, KT438568, KT438569, KT438570, KT438571, KT438572, KT438573, KT438574, KT438575, KT438576, KT438577, KT438578, KT438579, KT438581, KT438582 and KT438583. Additional genomic sequences used in the aa substitutions analysis can be found as GenBank: KP188543 and KP188540. | Since 2008 L6 is the endemic lineage that has circulated in SJRP. In 2010, the L1 lineage was first identified in the city. For a period, both lineages co-circulated, and then in 2013, L1 began to diminish until it was no longer detected in the population. Differences in replicative fitness are usually the main factor for clade replacement (CR) in DENV epidemics. However, despite the better viral fitness of the emerging lineage, the absence of CR could not be explained by these differences alone. Here, we combine epidemiological, phylogenetic, molecular and immunological analyses to provide a more precise understanding of the role of fitness in lineage dynamics with the persistence of L6 even after the introduction of L1 without CR. Differences in immune responses elicited by DENV-1 L1 and L6 lineages (genotype V), but not viral fitness in mosquito or human cells, explain the dynamics of circulating DENV in a city of Brazil. | Abstract
Introduction
Results
Discussion
Materials and methods | blood cells
invertebrates
innate immune system
medicine and health sciences
immune physiology
immune cells
cytokines
immunology
microbiology
animals
developmental biology
molecular development
insect vectors
research and analysis methods
sequence analysis
immune system proteins
infectious diseases
white blood cells
animal cells
proteins
bioinformatics
antigens
t cells
viral replication
disease vectors
immune response
insects
immune system
arthropoda
biochemistry
mosquitoes
eukaryota
cell biology
virology
physiology
database and informatics methods
interferons
biology and life sciences
cellular types
species interactions
organisms | 2018 | Viral immunogenicity determines epidemiological fitness in a cohort of DENV-1 infection in Brazil | 10,848 | 226 |
During development, animals usually undergo a rapid growth phase followed by a homeostatic stage when growth has ceased. The increase in cell size and number during the growth phase requires a large amount of lipids; while in the static state, excess lipids are usually stored in adipose tissues in preparation for nutrient-limited conditions. How cells coordinate growth and fat storage is not fully understood. Through a genetic screen we identified Drosophila melanogaster CDP-diacylglycerol synthetase (CDS/CdsA), which diverts phosphatidic acid from triacylglycerol synthesis to phosphatidylinositol (PI) synthesis and coordinates cell growth and fat storage. Loss of CdsA function causes significant accumulation of neutral lipids in many tissues along with reduced cell/organ size. These phenotypes can be traced back to reduced PI levels and, subsequently, low insulin pathway activity. Overexpressing CdsA rescues the fat storage and cell growth phenotypes of insulin pathway mutants, suggesting that CdsA coordinates cell/tissue growth and lipid storage through the insulin pathway. We also revealed that a DAG-to-PE route mediated by the choline/ethanolamine phosphotransferase Bbc may contribute to the growth of fat cells in CdsA RNAi.
During development, animals grow rapidly by both cell proliferation and cell growth. Cell growth is a heavily energy-dependent process and requires large amounts of phospholipids for expansion of cellular membranes and other cellular needs. Fat storage on the other hand is an energy-saving process which stores neutral lipids in the form of triacylglycerol (TAG) for utilization under nutrient-limited conditions such as starvation. In the de novo biosynthetic pathway, fatty acyl-CoA is utilized either for TAG synthesis or for phospholipid production (Figure 1A), so it is reasonable to propose that growth and fat storage might be balanced during normal development [1]. Indeed, numerous observations support such a balance. For example, Caenorhabditis elegans Tor mutants grow slowly and exhibit excess fat storage [2]. Drosophila melanogaster insulin pathway chico mutants are less than half the size of wild type, but show an almost 2-fold increase in lipid levels [3]. The mechanisms that coordinate cell growth and neutral lipid storage are largely unknown and many intriguing questions remain to be addressed. During development, animals usually have an initial phase of rapid growth, which involves expansion of cell number and size, followed by a homeostatic state, when growth ceases and excess fat is stored in adipose tissue. How is the balance between cell growth and neutral lipid storage regulated in these two different developmental stages? How is the balance regulated in different tissues or cells? For example, how is the balance regulated in the adipocyte, since it both grows large and stores fat? In some disease states, ectopic lipid accumulation in non-adipose tissues such as muscle, pancreas, and liver is often observed [4]. How is the balance regulated in non-adipose tissues? Answering these questions would definitely lead to a significant advance of our knowledge in the fields of both fat storage and cell growth. The insulin pathway is a conserved signaling pathway that is essential for cell growth in response to nutrient conditions [5]–[9]. The core components of the insulin pathway include the insulin receptor (InR), insulin receptor substrates (IRS), phosphatidylinositol 3-kinase (PI3K), the protein kinase Akt, and the transcription factor FOXO. PI3K, which generates phosphatidylinositol 3,4, 5-trisphosphate (PIP3) from phosphatidylinositol 4,5-bisphosphate (PIP2), and PTEN, the phosphoprotein phosphatase that converts PIP3 to PIP2, are positive and negative regulators of the insulin pathway, respectively [10], [11]. PIP4K, which generates PIP2 from phosphatidylinositol 5-phosphate (PIP) also positively regulates insulin pathway activity [12]. Therefore, various enzymes that affect the level of PIP3 provide an important layer of regulation of insulin pathway activity. Recently, several novel components of the insulin pathway were identified, including miRNAs (miR-8 and miR-14) and secreted proteins (Upd and SDR) [13]–[16]. It is well known that the insulin pathway regulates lipid homeostasis and couples nutritional conditions with systemic growth and metabolism. Besides inhibiting lipolysis, it also promotes fatty acid synthesis through activating the expression of different target genes, including acetyl-CoA carboxylase (ACC), fatty acid synthase (FAS), and sterol regulatory element binding protein (SREBP) [17]–[19]. Recently acyl-CoA synthetase (ACS) /Pudgy has been identified as a direct target of FOXO [20]. Since many of these regulators have been studied during the homeostatic stage, it is not known whether they contribute to the balance of cell growth and neutral lipid storage during the developmental stage. Furthermore, additional regulators or targets of the insulin pathway in balancing cell growth and fat storage remain to be identified. With its exquisite genetic tools and evolutionarily conserved metabolic pathways, Drosophila melanogaster has been accepted as a model for studying lipid metabolism [21]–[24]. Through continuous feeding, Drosophila larvae grow rapidly with a nearly 200-fold increase in body mass during the 4-day larval stage. During this time most tissues, such as brain, salivary gland, and imaginal discs, store very little neutral lipid, while the fat body accumulates large amounts of fat [23]. The fat body is a specialized energy storage organ, equivalent to the vertebrate adipose tissue, and most of the dietary fats and de novo-synthesized fats from the intestine are transported to the fat body via lipoproteins in the hemolymph [25]. As in mammals, many questions remain to be answered in Drosophila. How do Drosophila larvae coordinate cell growth and neutral lipid storage in non-adipose tissues? What mechanisms are used in the fat body to balance the storage of fat along with cell growth? In this study, we identified that Drosophila CDP-diacylglycerol synthetase, CdsA, coordinates cell growth and neutral lipid storage through phosphatidylinositol (PI) metabolism and the insulin pathway. We also found that when CdsA is defective, a DAG-to-PE route mediated by the choline/ethanolamine phosphotransferase Bbc may contribute to the growth of fat cells.
We previously reported that mutants of dSeipin, the Drosophila homolog of human lipodystrophy gene Seipin which is important for fat storage and lipid droplet size [26], [27], exhibit ectopic lipid droplets in salivary glands, which can be stained by the neutral lipid dyes Nile red or Bodipy [28]. The Drosophila salivary gland is large and easy to manipulate and has been previously used as an in vivo system to study cell death and autophagy, cell growth and proliferation, and signal transduction, for example in Hedgehog signaling [29]. Hence, we investigated neutral lipid storage regulation in salivary glands by screening for genes which, when knocked down or overexpressed, caused an ectopic lipid droplet phenotype in salivary gland. We used ppl-Gal4, which drives strong gene expression in both larval salivary gland and fat body. The strongest ectopic lipid storage phenotype was caused by CdsA RNAi (Figure 1B). The phenotype was also verified with an independent CdsA RNAi line (Figure S1A). In addition, CdsA1, a weak allele of CdsA, was previously found to have a dSeipin-like ectopic lipid storage phenotype in the salivary gland [28], demonstrating that CdsA affects fat storage. CdsA is the sole Drosophila CDP diglyceride synthetase (Cds), and diverts phosphatidic acid (PA) from TAG synthesis to the synthesis of cytidine diphosphate diacylglycerol (CDP-DAG), the precursor of two phospholipids, PI and phosphatidylglycerol (PG) (Figure 1A). According to results from the Drosophila gene expression database FlyAtlas (http: //www. flyatlas. org), CdsA is widely expressed in different larval tissues. In adults, CdsA is highly enriched in the eye and is important for the phototransduction pathway [30]. We confirmed the larval expression profile with Q-RT-PCR (Figure S1B) and a Lac-Z enhancer trap line. The Lac-Z signal was detected in many places, including salivary gland, fat body, proventriculus, hind gut, brain, muscle, Malpighian tubules, and tracheal tubes (Figure S1C). The broad expression of CdsA and the strong salivary gland fat storage phenotype of CdsA RNAi provide a unique opportunity to address whether other non-adipose tissues have the same capacity as salivary glands to store excess fat. When we drove UAS-CdsA RNAi transgene expression using the ubiquitous tub-Gal4 driver, a strong, fully penetrant phenotype was observed. The level of CdsA transcripts was reduced to about 40% with RNAi. The tub>CdsA RNAi larvae accumulated excess fat in many non-adipose tissues, including salivary gland, proventriculus, hind gut, Malpighian tubules, and trachea (Figure 1C). Among the tissues examined, the most prominent lipid storage phenotype was found in salivary gland and prothoracic gland, both of which have polyploid cells. The excess lipid accumulation phenotype of tub>CdsA RNAi was reproduced using various tissue-specific Gal4 lines, suggesting that CdsA acts tissue-autonomously (data not shown). To direct measure the accumulation of TAG in non-adipose tissues, we measured the level of TAG from the fat body-removed whole larval samples. Compared to control, tub>CdsA RNAi animals have much higher level of TAG, consistent with the Nile red staining results (Figure 1B). Together, these results indicate that although the fat body is the main energy reservoir for larvae, neutral lipids can still be stored in significant amounts in many non-adipose tissues when CdsA function is lost. However, it remains to be determined whether excess lipid accumulation in these non-adipose tissues can fulfill the energy reservoir function of the fat body, or causes deterioration of these tissues. Beside the fat storage phenotype, we noticed a significantly reduced organ size phenotype in tub>CdsA RNAi larvae. The imaginal discs, salivary gland, and brain are all smaller in tub>CdsA RNAi larvae than controls (Figure 1D). This phenotype could be due to a decrease in cell size, or cell number, or both. We quantified the cell size and cell number in the salivary gland and found that while the cell number is not changed, the cell size is greatly reduced in tub>CdsA RNAi larvae (data not shown). Therefore, reduced cell size is the likely cause of the small organ size in CdsA RNAi larvae. To rule out a possible off-target effect of RNAi, we turned to CdsA mutants. CdsA1 is a weak and viable allele of CdsA and does not cause an obvious small cell size phenotype (data not shown). In the CdsAGS8005 allele, a transposon element inserted into the second exon of CdsA disrupts transcription (Figure 2A). CdsAGS8005 animals died at late embryonic to early larval stages. By RT-PCR, CdsAGS8005 is likely a strong loss-of-function or null allele (Figure 2B). We generated CdsAGS8005 mutant clones in the salivary gland. Consistent with the RNAi result, CdsAGS8005 mutant salivary gland cells are smaller in size and accumulate large amounts of fat compared to neighboring control cells (Figure 2C), further demonstrating that CdsA functions cell-autonomously. On average, the size of CdsAGS8005 mutant salivary gland cells is only about 20% of wild-type cells. Together, these results suggest that CdsA plays a key role in balancing fat storage and cell growth. Cell growth depends on nutrient conditions and is well-known to be regulated by the insulin signaling pathway. The activity of the insulin pathway is regulated by the PI-derived lipid PIP3. Since CdsA is a key enzyme involved in the production of PI from PA through CDP-DAG (Figure 1A), and PI is the precursor of PIP, PIP2, and PIP3, we speculated that in CdsA RNAi or mutant animals, the total level of PI is reduced, resulting in a reduction in PIP3, which would lead to low insulin pathway activity and defective cell growth. To test this hypothesis, we first examined the level of PIP3 with a PIP3-specific GFP reporter, tGPH (tubulin-GFP-Pleckstrin homology), in the salivary gland [10]. GFP is recruited to the plasma membrane by binding to PIP3; therefore, the ratio of plasma membrane to cytosol GFP intensity reflects the relative level of PIP3. In controls, the GFP signal is strong in both plasma membranes and nuclei. In CdsA RNAi larvae, the overall GFP intensity is much lower due to unknown reasons and the plasma membrane to cytosol GFP intensity ratio is only 30% of wild type, indicating reduced levels of PIP3 (Figure 3A). To further probe the activity of the insulin pathway in CdsA RNAi, we also determined the level of phosphorylation of residue S505 of the protein kinase Akt (P-AktS505, equivalent to S473 in mammalian Akt), which has often been used as an indicator of insulin pathway activity. The P-AktS505 level is significantly decreased in CdsA RNAi whole larvae and salivary glands (Figure 3B). To obtain direct evidence of reduced PI levels, we quantified PI lipids in whole larval and salivary gland samples using mass spectrometry. We found that in both whole larvae and salivary glands, the total level of PI is greatly reduced in CdsA RNAi samples compared to controls and the level of PI in salivary glands is reduced more than ten-fold (Figure 3C). CdsA RNAi also causes a reduction in the total level of PG, accompanied by an increase in PA (Figure 3C). In addition, ubiquitous overexpression of CdsA slightly increases the levels of PI and PG in whole larval samples (Figure 3C). These data are consistent with the role of CdsA in catalyzing the conversion of PA to CDP-DAG, the precursor of PI and PG. Besides, compared to measurable amount of PIP2 in control, the level of PIP2 in CdsA RNAi salivary gland samples is below the threshold level detected by mass spectrometry (Figure 3D). Moreover, likely due to the small size of the salivary glands, we were unable to measure the abundance of PIP3 in both control and CdsA RNAi samples with mass spectrometry. Instead, we measured the level of PIP3 in whole larval samples using an ELISA method. In CdsA knockdown larvae, the PIP3 level is reduced to less than half of the control (Figure 3D). Taken together, our results indicate that CdsA influences insulin pathway activity and PI metabolism in the salivary gland. These results predict that blocking de novo PI synthesis in the salivary gland should mimic the CdsA mutant clone phenotype in affecting salivary gland cell growth. Phosphatidylinositol synthase (Pis) is the only enzyme which synthesizes PI from CDP-DAG and inositol (Figure 1C). Similar to CdsA RNAi salivary glands, Pis RNAi salivary glands are small and accumulate lipid droplets (Figure 3E). Similar to CdsA RNAi, the level of PI in salivary glands is greatly reduced in Pis RNAi samples compared to controls (Figure 3F). On the other hand, the level of PG is significantly increased in Pis RNAi (Figure 3F). We further examined Pis1, a likely null mutant of Pis [31]. The Pis1 mutation is lethal, like CdsAGS8005, so we generated Pis1 mutant clones. Interestingly, Pis1 mutant salivary gland cells do not accumulate large numbers of lipid droplets (Figure 3G), suggesting that Pis mutations likely increase the level of CDP-DAG, which is converted to PG, without significantly elevating the level of PA. Similar to CdsAGS8005 mutant salivary gland cells, Pis1 mutant salivary gland cells are significantly smaller than wild-type cells (Figure 3G). On average, the size of Pis1 mutant salivary gland cells is only about 25% of wild-type cells. The reduction of cell size in CdsA and Pis mutants is not as strong as in insulin pathway mutants, where salivary gland cell size can be reduced by over 90% [10], [32], suggesting that PI from other resources, such as uptaken externally or inherited from mother cells, may partly contribute to salivary gland cell growth. Taking these results together, we conclude that CdsA regulates salivary gland cell growth by affecting PI metabolism and insulin pathway activity. To further reveal the connections between CdsA, PI metabolism, and the insulin pathway, we turned to genetic interaction assays. In the salivary gland genetic screen, we identified two insulin pathway components, PI3K and Akt, in addition to CdsA. Knockdown of PI3K by dominant-negative PI3K (PI3K DN) or of Akt by RNAi leads to small cell size and ectopic lipid storage phenotypes in the salivary gland, although the lipid storage phenotype is weaker than that of CdsA RNAi (Figure 4A and S2A). Since both CdsA RNAi and reduced insulin pathway activity have the same phenotypes in the salivary gland, we examined their genetic interaction by gain of insulin pathway activity in CdsA RNAi larvae and vice versa. We found that overexpressing wild-type Akt or constitutively active PI3K (PI3K CA) fully suppressed the small cell size phenotype of CdsA RNAi (Figure 4B and 4C). Although we can' t rule out the possibility that CdsA and insulin pathway act redundantly in a common pathway, these results are consistent with the notion that CdsA acts upstream of PI3K in positively regulating insulin pathway activity. Interestingly, the ectopic lipid accumulation phenotype of CdsA RNAi is partially rescued by overexpression of PI3K CA, but not Akt (Figure 4B and S2B). We then tested whether overexpression of CdsA can rescue the cell size and ectopic lipid storage phenotypes caused by impaired insulin pathway activity. On its own, CdsA overexpression slightly increases the size of the salivary gland (Figure 4D and 4E). Interestingly, CdsA overexpression partially, but significantly, rescued the small cell size phenotype of PI3K DN (Figure 4D and 4E). Moreover, the salivary gland ectopic lipid storage phenotype of PI3K DN can be fully suppressed by CdsA overexpression (Figure 4D and S2B). These results raise the possibility that insulin pathway activity may influence the CdsA level. Since insulin signaling regulates lipid homeostasis mainly at the level of gene transcription and the forkhead box protein dFOXO can transcriptionally activate the upstream target InR via a feedback regulatory loop [33], we asked whether the transcription of CdsA is affected by the insulin pathway. We examined the level of CdsA under conditions which perturb insulin pathway activity. Compared to controls, CdsA transcription levels in PI3K DN or Akt RNAi salivary glands were reduced significantly to 40–50% as assayed by quantitative RT-PCR (Figure 4F). Conversely, CdsA transcription levels slightly increased when Akt was overexpressed (Figure 4F). Likewise, loss of FOXO slightly increased the level of CdsA (Figure 4F). Taken together, these results pinpoint a positive feedback loop between insulin signaling and CdsA in modulating the balance of cell growth and lipid storage. In this positive feedback loop (Figure 4G), CdsA positively regulates the activity of the insulin pathway through PI, and the insulin pathway affects the transcription of CdsA. Besides salivary gland, we also examined the role of CdsA in fat body by both loss-of-function and gain-of-function analyses. To our surprise, fat body-specific CdsA RNAi or CdsA overexpression using ppl-Gal4 or cg-Gal4 did not result in a significant lipid storage phenotype in the fat body (Figure 5A and data not shown). Consistent with this, the total levels of glycerides are not significantly different in control, CdsA RNAi, and CdsA-overexpressing animals (Figure 5B). The fat cell size is also unchanged in CdsA RNAi and CdsA-overexpressing animals (Figure 5A). Although the RNAi efficiency was verified by Q-RT-PCR (data not shown), it is still possible that knockdown by CdsA RNAi is not efficient enough to cause a cell growth or fat storage phenotype in the fat body. To rule out this possibility and to obtain more definitive answers, we analyzed CdsA mutants. In contrast to the small size and massive neutral lipid storage of CdsAGS8005 mutant salivary gland cells (Figure 2C), CdsAGS8005 mutant fat body cells are normal in size and fat storage (Figure 5C), which is consistent with the RNAi result. Therefore, we concluded that fat body cell growth and neutral lipid storage are not affected by CdsA under normal conditions. The lack of a neutral lipid storage phenotype in CdsA RNAi fat bodies suggests that PA and probably de novo lipogenesis from fatty acyl-CoA contribute little to final TAG content in the fat body under normal conditions. Indeed, it was reported that DAG is the major form in which lipids are transferred, via lipoproteins, from the intestine to the fat body for storage [25]. Our results are also consistent with a previous finding that unknown mechanism (s) maintain the level of fat body lipid storage within a narrow range [25]. Next, we asked why CdsA does not affect fat body cell growth. As we showed before, CdsA affects PI levels in the salivary gland and subsequently the activity of the insulin pathway (Figure 3). Is it possible that PI or insulin signaling is not important for fat body cell growth? Several previous reports involving manipulations of insulin pathway activity in fat cells clearly rule out this possibility [10], [16], [34]. Alternatively, it is possible that the PI level is not reduced in CdsA RNAi or mutant fat cells compared to salivary gland. We therefore examined the levels of PI and PIP3 in CdsA RNAi fat body cells. Similar to salivary gland, the overall tGPH reporter intensity is much lower in CdsA RNAi fat cell compare to control. The ratio of plasma membrane to cytosol GFP signal intensity from the tGPH reporter is reduced to around 50% of wild type in CdsA RNAi fat body cells (Figure 5D). By lipid content analysis, the level of PA is slightly increased in CdsA RNAi fat bodies (Figure 5E). However, compared to the >10-fold reduction of PI levels in the CdsA RNAi salivary gland, the level of PI is only reduced to 35% of wild type in the CdsA RNAi fat body (Figure 5E). Interestingly, the P-AktS505 level is not changed in the CdsA RNAi fat body (Figure 5F), consistent with the cell size phenotype. These results raise the possibility that although the level of PI, and probably PIP3, is reduced by CdsA RNAi, the reduction is not sufficient to affect Akt phosphorylation or other compensatory mechanism (s) that exist to support fat cell growth. To further explore why the fat cell growth is not affected in CdsA mutant, we turned to Pis null mutant and Pis RNAi again. Pis1 mutant fat body cells are about 30% smaller on average than control cells (Figure 6A) and in Pis RNAi fat body, the PI level reduced to less than half of the control (Figure 6B). Expression of dominant negative PI3K or Akt RNAi reduces the size of fat body cells by over 50% (Figure 6C). Together, these results suggest that de novo-synthesized PI contributes partly to fat cell growth and PI from other origins (i. e. transported from other cells/tissues or inherited from mother cells) may contribute significantly to the growth of fat cells. If so, why is a cell growth phenotype not exhibited by CdsA mutant clones? To examine the CdsA mutant phenotype in a sensitive genetic background, we crossed CdsA RNAi to PI3K or Akt mutants. We found that reducing PI3K or Akt to one copy in the CdsA RNAi background results in a reduction in fat cell size of around 50% (Figure 6D). These dosage-sensitive interactions suggest that CdsA RNAi mildly impairs fat cell growth. In addition, because CdsA acts one step earlier than Pis in PI biosynthesis (Figure 1A), it is possible that accumulation of or lack of certain metabolites compensates for the mild growth defect in CdsA mutants. Through lipid measurements, we found that the levels of DAG and PE are significantly increased in CdsA RNAi fat body samples. In fat body, the DAG level is doubled and the PE level is increased by about 50%, while in salivary gland, the PE level is unchanged (Figure 6E and 6F). In contrast, the levels of DAG and PE are only slightly increased in Pis RNAi fat body (Figure 6E and 6F). PEBP, a PE binding protein, has been reported to affect Akt phosphorylation, suggesting that PE may be linked to insulin signaling [35]. If the growth compensation is through the DAG-to-PE route, blocking the synthesis of PE from DAG in CdsA RNAi animals might lead to a more obvious fat cell growth phenotype. Choline/ethanolamine phosphotransferase (Cept) Bbc is a key enzyme in Drosophila that converts DAG to PE (Figure 1A) [36]. While bbc RNAi did not affect the P-AktS505 level (Figure 6G) and did not obviously affect the fat cell size (Figure 6C), CdsA and bbc double RNAi exhibits a synergistic effect: the fat cell size is drastically decreased to a level comparable to Akt or PI3K RNAi (Figure 6C). Moreover, the fat body P-AktS505 level is greatly reduced by CdsA and bbc double RNAi (Figure 6G). In contrast, the salivary gland cell size of CdsA and bbc double RNAi is comparable to CdsA single RNAi (Figure S3). Together, these results suggest that the elevated conversion of DAG to PE mediated by the choline/ethanolamine phosphotransferase Bbc may contribute to fat cell growth in CdsA RNAi larvae.
The regulation of cell growth in terms of both cell size and cell number has been studied intensively, and is known to involve many key core signaling pathways such as the insulin pathway, the Hippo pathway, and the mTOR pathway, as well as cross-talk between these pathways [5], [7], [37]. The downstream targets that execute and coordinate various specific aspects of cell growth are just starting to be elucidated. In this study we have revealed the intrinsic connections between CDP-diacylglycerol synthetase, PI metabolism, and insulin pathway activity in coordinating cell growth and neutral lipid storage. Lipid storage is a basic function of a cell. Within an organ, glycerol and fatty acyl-CoA is the precursor of all phospholipids, which are mainly utilized within the cell, and glycerolipids, which are largely destined for neutral lipid storage. Therefore, under normal developmental conditions when resources are not unlimited, a balance must be achieved between cell growth and storage (fast growth and less storage, or slow growth and increased storage). Moreover, at the systematic level, the tissue specificity of insulin action and the lipid flow between different tissues make this balance more complicated [38]. For non-adipose tissues, fast cell growth and low levels of storage are beneficial for maximizing growth potential. However, the balance encounters a problem in adipose tissue because both growth and storage are required. Several predictions can be made here. First, the key enzymes that act at the branch points of phospholipid and TAG synthesis are good candidates for controlling the balance between growth and storage. Second, the factors that control the difference in balance regulation between adipose tissue and non-adipose tissue likely sit after the regulatory points. Alternatively, adipose tissue may have two developmental phases, a growing phase and a fat storage phase. The timing of the switch between these two phases would determine the final size and storage capacity of adipose tissue. Third, excess nutrients may saturate the balance mechanism, leading to overload in both directions. Indeed, our results depict a scenario in which CDP-diacylglycerol synthetase coordinates cell growth and fat storage by partitioning the flow of fatty acyl-CoA between PI synthesis and TAG synthesis. Moreover, our results reveal different contributions of de novo-synthesized PI and PI derived from external sources to the growth of fat cells and salivary gland cells. De novo-synthesized PI contributes most to the growth of salivary gland cells, while fat body cell growth is mainly stimulated by externally-derived PI (Figure 6H). Therefore, the role of CdsA in fat cell growth is only revealed under conditions in which growth is mildly compromised. Besides that, an external supply of DAG, most likely from the intestine, may bypass the CdsA-gated coordination, allowing adipose tissue (fat body) cells to both grow large and store fat (Figure 6H). Consistent with this, it was previously reported that intestine-derived lipids, including DAG, are mainly transported to the fat body via lipoproteins, while relatively low levels of lipids are transported to non-adipose tissues either directly from the intestine or indirectly via the fat body [25], [39]. By responding to environmental and nutritional cues, the insulin pathway regulates cell growth [5], [9]. Through genetic analyses, GFP reporter assays, and lipidomics, we have presented evidence for a positive feedback loop between the insulin pathway and CdsA levels (Figure 4G). In this loop, lowering insulin pathway activity reduces CdsA transcription, leading to less PI production, which further decreases insulin pathway activity. There are numerous previous findings of FOXO as a transcriptional activator [20], [33], [40], but dFOXO appears to be a repressor of CdsA transcription. Microarray and ChIP studies have identified many other targets of FOXO repression [41]. At this point, we do not know whether dFOXO represses CdsA directly or indirectly. The widespread cell growth and fat storage phenotype caused by CdsA RNAi suggests that the insulin pathway-CdsA feedback loop operates widely in various tissues. In addition, it shows that the fat storage capacity of most non-adipose tissues is enormous. The insulin pathway-CdsA feedback regulation allows cells to take full advantage of the environmental conditions to grow fast and stop growth quickly when necessary. For example, in a nutrient-rich environment, animals that can grow quickly have a tremendous competitive advantage over slow-growing ones. In nutrient-poor conditions, the ability to quickly convert growth to fat storage is beneficial for animals to survive through harsh times. This is not the first time that feedback regulation of the insulin pathway has been reported [33], suggesting that it might be a common mechanism to augment pathway sensitivity and ensure a rapid response to changes in nutrient conditions. Regulation of the insulin pathway by lipids is already known to occur at the points of PIP to PIP2 and PIP2 to PIP3 conversion. Compared to regulation by PI3K and PTEN, CdsA provides a new layer of control at the level of available lipids. Together with other pathways regulated by available nutrients/small molecules, such as the mTOR pathway, which responds to amino acid levels, and the AMPK pathway, which responds to ATP levels [42], [43], this new regulatory mechanism helps cells to cope with diverse environmental changes. The insulin-CdsA feedback loop is important during cell development. When cells reach the homeostasis phase, the growth-favored loop must be switched towards fat storage. In the future, it will be interesting to determine what triggers the switch. In addition, the existence of the positive feedback loop suggests that excess nutrients will overload the fatty acyl-CoA partitioning mechanism and disrupt the balance. Therefore, impairing the insulin-CdsA feedback loop by genetic mutations or environmental conditions may cause an imbalance in fat storage, resulting in metabolic diseases such as obesity and related disorders. Re-establishing the feedback loop could be a way of treating these disorders.
Drosophila stocks were maintained in standard cornmeal food. w1118 was used as wild type. All RNAi stocks were obtained from the VDRC or NIG RNAi stock centers. ppl-Gal4 was generously provided by Dr. Pierre Leopold. The expression pattern of ppl-Gal4 was confirmed with UAS-GFP. Other stocks used were: tub-Gal4, y1 w67c23; P{EPgy2}CdsAEY08412, y1 w67c23; P{GSV3}GS8005/TM3 Sb1 Ser1, y1 w*; P{w[+mC] = Dp110[D954A]}2, y1 w1118; P{w[+mC] = UAS-Akt1. Exel}2, P{w[+mC] = Dp110-CAAX}1, dFOXO21, dFOXO25, w1118; P{w[+mC] = tGPH}2; Sb1/TM3 Ser1, Pis1/FM7a; P{w[+mC] = hs-Pis. MYC}3/TM2, w*; Df (3R) Pi3K92E[A] H[A] Pi3K92E[A], P{ry[+7. 2] = HBS}3/TM6B Tb1, ry506 P{ry[+t7. 2] = PZ}Akt1[04226]/TM3 ryRK Sb1 Ser1, P{w[+mC] = Ubi-mRFP. nls}1, w* P{ry[+t7. 2] = hsFLP}122 P{ry[+t7. 2] = neoFRT}19A, hsflp122; sp/Cyo; FRT2A histone: GFP/TM6B, ry506 P{HZ}CdsA1. ppl-Gal4 virgin females were crossed to males harboring UAS-RNAi inserts to knock down target gene expression in the larval salivary gland. Wandering 3rd instar larvae harboring both the Gal4 and the UAS-RNAi transgene were dissected and mounted on glass slides with PBS. For visualization, specimens were examined with a Zeiss DIC microscope. For GAL4/UAS experiments, flies were grown at 29°C to allow for maximal GAL4 activity. For tissue or clonal analysis, unless stated otherwise, all larvae were dissected at the wandering 3rd instar stage. Salivary gland clones were generated by heat shock at 37°C for 1 hr immediately after collection of eggs for 4 hr. Fat body clones were induced 8 hr after egg laying for 1 hr at 37°C. After crossing ppl-Gal4 with RNAi lines, wandering 3rd instar larval progeny were dissected and stained with Nile red or Bodipy, which are fluorescent dyes that specifically mark neutral lipids. For lipid droplet staining, larvae were dissected in PBS and fixed in 4% paraformaldehyde for 30 min at room temperature. Tissues were then rinsed twice with PBS, incubated for 30 min in a 1∶2500 dilution with PBS of 0. 5 mg/ml Nile red (Sigma), and then rinsed twice with distilled water. Stained samples were mounted in 80% glycerol for photo-taking. For β-galactosidase staining, dissected tissues were fixed for 2 min in 0. 5% glutaraldehyde on ice. The tissues were washed three times for 10 min each in PBS containing 0. 3% Triton X-100 (PBT) and then placed in X-gal staining solution (0. 2% X-gal, 1 mM MgCl2,5 mM K4[Fe (CN) 6], 5 mM K3[Fe (CN) 6]) overnight at 37°C. All images were taken using a Nikon confocal scope or Zeiss fluorescent scope. Quantification of mutant clone areas was performed by calculating the size of the area occupied by mutant clones versus wild-type cells. Quantification of the size of whole salivary gland, brain, and wing disc or the size of fat body cells was performed by measuring the area occupied by these organs/cells. To quantify lipid storage in salivary glands, the Nile red positive areas of salivary glands per genotype were measured and normalized to the whole cell area as previously described [25]. The average lipid storage in salivary glands of CdsA RNAi was set as 1. For tGPH signal quantification, membrane-to-cytoplasmic GFP ratios were calculated from measurements of mean pixel intensities within equal areas of membrane versus cytoplasm. All measurements were done using Nikon Br analysis software. Total RNA was isolated using an RNeasy kit (Qiagen) according to the manufacturer' s protocol. To quantify CdsA transcript levels in tub>CdsA RNAi and tub/+ controls, three groups of 5 larvae per genotype were collected at the wandering 3rd instar larval stage. To quantify CdsA transcript levels in different tissues such as salivary gland, fat body, brain, and gut, three groups of 30 tissues per genotype were dissected and collected from wandering 3rd instar larvae. To quantify CdsA transcript levels in larval salivary gland with loss of function or overexpression of insulin pathway components, salivary glands were dissected and collected in three groups of 50 pairs per genotype from wandering 3rd instar larvae. For each sample, 3 µg total RNA was used to synthesize single-stranded cDNA using a SuperScript II reverse transcriptase kit (Invitrogen). Quantitative RT-PCR (qRT-PCR) experiments (Agilent Stratagene Mx3000P system, SYBR Green PCR mix) were performed using specific primer pairs (5′ GCAATGATGTCATGGCGTAC 3′ and 5′ AAATCCAGAGGCAAAGAAGC 3′). The level of each transcript was normalized to rp49 in the same sample. All qRT-PCR experiments were repeated at least three times. Whole larva, salivary gland, and fat body extracts were prepared by dissecting wandering 3rd instar larvae in groups of 6,120, and 30 larvae/tissues respectively. Extracts were lysed by homogenizing equal masses of control or RNAi larval tissues on ice in 200 µl of ice-cold 1% SDS lysis buffer (1% SDS, 40 mM Tris-Cl PH7. 45,1 mM PMSF, 1 tablet Protease inhibitor, PH7-8). For western blotting, 30 µg of protein sample were loaded, blotted, and detected with the following antibodies: rabbit anti-Akt (Cell Signaling, diluted at 1∶1000), rabbit anti-phospho-Akt (Ser473) (Cell Signaling, diluted at 1∶1000), and rabbit anti-α-tubulin (Abcam, diluted at 1∶4000). Quantification of band intensities was done using Image J (National Institutes of Health) software. The extraction and measurement of PIP3 were performed using the PIP3 Mass ELISA Kit (Echelon), according to the manufacturer' s instructions. For extraction of PIP3, four groups of 20 larvae per genotype were collected at the wandering 3rd instar larval stage. The level of PIP3 was normalized to larval dry mass in each group. The PIP3 ELISA assay was repeated four times. Lipids were extracted from salivary glands of 50 larvae, fat bodies of 50 larvae, 12 whole larvae or 6 fat body-removed larvae carcass (three sets of samples per genotype). An Agilent high performance liquid chromatography (HPLC) 1260 system coupled with an Applied Biosystem Triple Quadrupole/Ion Trap mass spectrometer (4500Qtrap) was used to quantify individual lipids. Polar lipids were analyzed using multiple reaction monitoring (MRM) scans [44]. Separation of individual polar lipids was carried out using a Phenomenex Luna 3u silica column (i. d. 150×2. 0 mm). Individual lipid species were quantified by referencing to spiked internal standards. PC-14: 0/14: 0, PE-14: 0/14: 0, PS-34: 1/d31, PA-17: 0/17: 0, PG-14: 0/14: 0 and PI-34: 1/d31 were obtained from Avanti Polar Lipids. Neutral lipids were analyzed using a sensitive HPLC/ESI/MS method modified from a previous method [45]. DAGs were quantified using 4ME 16∶0 Diether DG (Avanti) as an internal standard. Quantitative analysis of PIP2 was carried out as described [46] using a Dionex Ion Chromatography 5000 system. Lipid extracts were deacylated by incubation with 0. 5 ml of methylamine reagent [MeOH/40% methylamine in water/1-butanol/water (47∶36∶9∶8) ] at 50°C for 45 min. The aqueous phase was dried, resuspended in 0. 5 ml of 1-butanol/petroleum ether/ethyl formate (20∶40∶1), and extracted twice with an equal volume of water. Aqueous extracts were dried, resuspended in water, and subjected to anion-exchange HPLC on an Ionpac AS11-HC column. Lipid levels were calculated using deacylated anionic phospholipids as standards. Total body glycerides were measured by homogenizing six groups of 5 male larvae per genotype in 100 µl PBS containing 0. 05% Tween 20 (PBST). The homogenate was immediately heat-treated at 70°C for 10 min. Then, 20 µl samples were incubated with 200 µl Triglyceride Reagent (Triglycerides Kit, ZhongShengBeiKong) for 10 min at 37°C and the absorbance at 540 nm was measured using a spectrophotometer. The total protein content of the same samples was measured by Bradford assay (Sigma). Glyceride levels were normalized to total protein amounts in each sample. | During development, animals undergo a rapid increase in cell size and number, which requires large amounts of lipids, in the form of phospholipids, for the expansion of cell membranes. Once the growth phase ends, excess lipids are usually stored as body fat, in the form of triacylglycerol (TAG), for use when nutrients are limited. How cells coordinate growth and fat storage is not fully understood. By screening for genes that affect lipid storage in the fruitfly Drosophila we discovered that the enzyme CDP-diacylglycerol synthetase (CdsA) coordinates cell growth and fat storage. Phospholipids and TAG have a common precursor, phosphatidic acid, which is diverted by CdsA from TAG synthesis to synthesis of the phospholipid phosphatidylinositol (PI). We also uncovered a link between CdsA and the insulin signaling pathway, which plays a major role in regulating cell and tissue growth. CdsA regulates the level of PI, which modulates insulin pathway activity; insulin pathway activity, in turn, influences the level of CdsA. The lipid metabolism pathways and the insulin signaling pathway are conserved in other animals including humans. Our findings may therefore provide further insights into clinically important imbalances in fat storage such as obesity. | Abstract
Introduction
Results
Discussion
Materials and Methods | animal models
developmental biology
drosophila melanogaster
model organisms
signaling
cell growth
molecular development
biology
molecular cell biology | 2014 | CDP-Diacylglycerol Synthetase Coordinates Cell Growth and Fat Storage through Phosphatidylinositol Metabolism and the Insulin Pathway | 10,783 | 312 |
Circadian clocks coordinate time-of-day-specific metabolic and physiological processes to maximize organismal performance and fitness. In addition to light and temperature, which are regarded as strong zeitgebers for circadian clock entrainment, metabolic input has now emerged as an important signal for clock entrainment and modulation. Circadian clock proteins have been identified to be substrates of O-GlcNAcylation, a nutrient sensitive post-translational modification (PTM), and the interplay between clock protein O-GlcNAcylation and other PTMs is now recognized as an important mechanism by which metabolic input regulates circadian physiology. To better understand the role of O-GlcNAcylation in modulating clock protein function within the molecular oscillator, we used mass spectrometry proteomics to identify O-GlcNAcylation sites of PERIOD (PER), a repressor of the circadian transcriptome and a critical biochemical timer of the Drosophila clock. In vivo functional characterization of PER O-GlcNAcylation sites indicates that O-GlcNAcylation at PER (S942) reduces interactions between PER and CLOCK (CLK), the key transcriptional activator of clock-controlled genes. Since we observe a correlation between clock-controlled daytime feeding activity and higher level of PER O-GlcNAcylation, we propose that PER (S942) O-GlcNAcylation during the day functions to prevent premature initiation of circadian repression phase. This is consistent with the period-shortening behavioral phenotype of per (S942A) flies. Taken together, our results support that clock-controlled feeding activity provides metabolic signals to reinforce light entrainment to regulate circadian physiology at the post-translational level. The interplay between O-GlcNAcylation and other PTMs to regulate circadian physiology is expected to be complex and extensive, and reach far beyond the molecular oscillator.
Circadian clocks are endogenous protein machines that integrate external time cues and internal metabolic states to impose temporal organization on physiology, metabolism, and behavior (reviewed in [1–2]). They allow organisms from all kingdoms of life, which experience perpetual 24-hour day-night cycles, to anticipate daily environmental changes and execute biological tasks, from molecular to behavioral levels, at the optimal time of day. Over the years, great progress in elucidating the molecular mechanisms driving circadian rhythms has been made by studying the core circadian oscillator. The molecular oscillator consists of two interlocked transcriptional translational feedback loops (TTFLs) that produce daily oscillations in clock mRNAs and proteins to drive rhythms in diverse cellular processes. During the day and into the early parts of the night, two basic helix-loop-helix (bHLH) -PAS transcription factors, CLOCK (CLK) and CYCLE (CYC; homolog of BMAL1 in mammals), activate transcription of their own repressors, period (per) and timeless (tim) and other clock-controlled output genes [3]. After a time-delay and TIM-assisted entry into the nucleus [4], PER interacts with and inhibits CLK-CYC activity [5]. This repression is relieved upon proteasomal-dependent degradation of PER during late night into early morning, thus initiating another round of CLK-CYC-mediated transcription [6–7]. Among CLK-activated genes are two bZIP transcription factors, vrille (vri) and par domain protein 1ε (Pdp1ε) [8–9]. Due to differential kinetics of VRI and PDP1ε protein accumulation, VRI accumulates first and inhibits Clk expression. As VRI level decreases, PDP1ε accumulates and activates Clk transcription, and the cycle of per/tim expression starts again. More recent evidence however suggested that the main role of VRILLE could be to control clock output by driving rhythms in expression of the neuropeptide PDF (Pigment Dispersing Factor) and neuronal arborization [10]. CLOCKWORK ORANGE (CWO) is another direct CLK target that feedbacks and represses CLK activity by competing with CLK-CYC complexes for E-box binding at circadian promoters [11]. The TTFLs are synchronized to the 24-hour day-night cycle through light-dependent degradation of TIM [12–14], which interacts with the photoreceptor CRYPTOCHROME 1 (CRY1) [15]. To extend the duration of the TTFL to last a 24-hr circadian cycle, post-translational regulation of core clock proteins overlays on the TTFL and has been recognized to be critical in maintaining the functionality of the circadian oscillator (reviewed in [1,16]). In Drosophila, the phase-specific phosphorylation state of PER is closely linked to its time-of-day specific function and the speed of the oscillator [17–21]. De novo synthesized hypophosphorylated PER goes through a multi-site phosphorylation program that progressively increases its phospho-occupancy until it gets hyperphosphorylated. In particular, phosphorylation at a N-terminal phosphodegron targets PER for degradation in a proteasome-dependent manner [17]. In a study by Robles et al. [22], 25% of the 20,000 phosphosites identified in mouse liver proteins were found to oscillate over the circadian cycle. This suggests that widespread and dynamic oscillations in phosphorylation occur beyond core circadian transcription factors to transition cellular proteins between functional states to regulate circadian physiology. More recently, O-GlcNAcylation has emerged as another PTM that can regulate the temporal function and activity of circadian transcription factors [23–25]. In contrast to protein phosphorylation, which is mediated by a wide selection of kinases and phosphatases, protein O-GlcNAcylation is regulated by a single pair of enzymes with opposing functions [26]. O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA) facilitate the O-GlcNAcylation and de-O-GlcNAcylation of cellular proteins respectively. In Drosophila, PER and CLK have been identified as substrates of O-GlcNAcylation, and there is evidence from overexpression of ogt to support that O-GlcNAcylation of these clock transcription factors regulates their nuclear translocation, stability, and transcriptional activity [23–24]. In mammalian clocks, BMAL1 and CLOCK have been shown to be rhythmically O-GlcNAcylated over the circadian cycle, and O-GlcNAcylation functions to counteract ubiquitination to stabilize these proteins [25]. In a separate study, PER2 was shown to be modified by O-GlcNAcylation at the S662-S674 region, which is important for regulating clock speed via CK1 phosphorylation [24]. A S662G mutation in humans is known to cause the familial advanced sleep phase syndrome (FASPS) [27]. Interestingly, S662 can also be O-GlcNAcylated, suggesting in vivo interplay between phosphorylation and O-GlcNAcylation in this domain. Increasingly, the interplay between phosphorylation and O-GlcNAcylation is shown to be prevalent in the regulation of diverse cellular processes (reviewed in [1]). As O-GlcNAcylation is a nutrient-sensitive PTM that relies on the availability of UDP-GlcNAc, an end product of the hexamine biosynthetic pathway (HBP), it is expected that levels of cellular protein O-GlcNAcylation may be highly dependent on daily feeding-fasting cycles. Many unknowns regarding PTM regulation of clock proteins remain, including the mechanisms by which phase-specific phosphorylation and O-GlcNAcylation collaborate to regulate their time-of-day functions. A significant barrier to understanding these mechanisms is the identification of O-GlcNAcylated residues. In fact, although both Drosophila PER and CLK are known to be O-GlcNAcylated, specific O-GlcNAcylated residues have not been identified [23–24]. The effects of O-GlcNAcylation on these clock proteins have only been investigated by global overexpression and knockdown of ogt and oga, which impact all O-GlcNAcylated residues simultaneously. Lessons learned from previous studies on clock kinases [17–21] highlight the likelihood that valuable mechanistic insights may be overlooked by global manipulation in PTM enzyme expression. Other important questions concern the temporal requirement of clock protein O-GlcNAcylation and the relationship between protein O-GlcNAcylation status and feeding-fasting cycles. In this study, we sort to understand how metabolic input influences the O-GlcNAcylation status of PER and regulates its function. We used Mass Spectrometry (MS) proteomics to identify PER O-GlcNAcylation and phosphorylation sites from adult flies, and characterized the function of site-specific O-GlcNAcylation events in vivo. We focused on PER as its phase-specific function has been shown to be highly dependent on its progressive phosphorylation program over the circadian cycle. Investigating PER O-GlcNAcylation can therefore set the stage for understanding the interactions between phosphorylation and O-GlcNAcylation in regulating its circadian function. We observe that loss of O-GlcNAcylation at multiple residues affect PER repressor function. In particular, loss of O-GlcNAcylation at PER (S942), which is located in the PER-CLK interaction or CLK binding domain (CBD), leads to stronger PER-CLK interaction and premature entry into the circadian repression phase. Conversely, overexpression of OGT in clock neurons weakens PER-CLK interaction, contributing to its period-lengthening phenotype in locomotor activity rhythms. Finally, we report that daily rhythms of PER O-GlcNAcylation in adult head tissues correlate with feeding-fasting cycles. This correlation is expected to be even stronger in peripheral tissues, which are more sensitive to metabolic signals. Specifically, PER O-GlcNAcylation exhibits circadian rhythmicity and is higher during the day when flies are actively feeding. Our results suggest that metabolic input collaborates with other entrainment signals to regulate time-of-day PER function in circadian transcription.
Although O-GlcNAcylation has been shown to influence PER function, specific residues that are modified by O-GlcNAcylation have not been identified [23–24]. This represents a critical barrier to understand the function of site-specific O-GlcNAcylation events and the interplay between O-GlcNAcylation and phosphorylation to regulate the phase-specific functions of PER. We therefore sought to identify PER O-GlcNAcylation sites and obtain temporal data on their occupancy. We purified FLAG-tagged PER from heads of wper0; p{3XFLAG-per (WT) } flies at specific time-points over the circadian cycle using FLAG affinity purification and performed quantitative mass spectrometry (MS) to obtain a circadian profile of PER phosphorylation and O-GlcNAcylation using fly tissues as starting materials. Although we attempted to perform this study using both fly head and body tissues, we were only able to identify PER O-GlcNAcylation sites in head tissues since we were unable to pull down sufficient PER proteins in bodies for comprehensive PTM identification. Nevertheless, we postulate that PER residues that are O-GlcNAcylated in head tissues will also be O-GlcNAcylated in peripheral tissues due to the ubiquitous expression of ogt [28]. To enable quantitation of PTM sites, flies were fed with 15N-labeled (heavy) yeast or 14N (light) yeast for two generations to ensure complete labeling in flies (herein termed 15N and 14N flies) [29]. To ensure that 15N-fed and 14N-fed flies show similar behavioral rhythms, we examined their locomotor activity rhythms using Drosophila activity monitoring [30]. Both types of flies displayed strong behavioral rhythms with periods close to 24-hr in constant darkness (S1A and S1B Fig). Furthermore, we measured PER daily abundance in 15N and 14N flies to confirm that temporal expression of PER was not altered as a result of the diets and observed no difference (S1C Fig). To profile PER PTMs in vivo, 15N and 14N flies were collected at six time-points over the circadian cycle (ZT1,3, 12,16,20,24), and protein extracts from heads of 15N and 14N flies were separately subjected to FLAG affinity purification (AP) prior to sample preparation for LC-MS/MS analysis (S2A and S2B Fig). Relative PTM quantification was achieved by pooling purified PER from all 14N samples, and aliquoting equal amounts of pooled 14N fractions to each of the six 15N purified PER samples at a 1: 1 ratio prior to MS analysis (S2C Fig). Despite the large amount of fly head tissues we used for our protein extraction and FLAG-PER purification, our 15N/14N quantitative MS did not yield satisfactory temporal resolution of PER PTM cycling. We therefore consolidated our MS data from multiple time-points with the goal of identifying PER O-GlcNAcylation sites qualitatively (Fig 1 and Table 1). We observed that multiple residues located in the CBD and within the CLK: CYC inhibition domain (CCID) [31,32] are O-GlcNAcylated. These include PER (S942) as well as a potential sites at S951, T952, or T954. Our MS analysis yielded O-GlcNAc-modified peptides in which only PER (S942) is modified, providing unambiguous identification of PER (S942) as an O-GlcNAcylated residue (Fig 1 and Table 1). In addition, O-GlcNAc-modified peptides spanning S942 to T954 were also identified, but we were not able to narrow down the single modified residue within these peptides. It is possible that only S942 is O-GlcNAcylated within this region. Alternatively, either S951, T952, or T954 may represent a second O-GlcNAcylated residue within this region. Taken together, we hypothesize that O-GlcNAcylation in the CBD may be important in modulating PER-CLK interaction and PER repressor activity. In addition to the O-GlcNAcylated residues in the CBD, a number of other potential sites were identified in other parts of the PER protein, but most of them are not located in characterized functional domains. In addition to identifying PER O-GlcNAcylation sites, we took this opportunity to identify PER phosphorylation sites in fly tissues for the first time and confirm phosphorylation sites that have previously been identified in Drosophila S2 cells [17–19,21]. Our analysis provides evidence that the majority of phosphorylation sites identified in Drosophila S2 cells are bona fide phosphorylation sites in fly tissues (Table 1). During the process of optimizing PER affinity purification and MS analysis using fly tissues as starting materials, we generated additional qualitative label-free MS datasets by analyzing PER phosphorylation sites at multiple time-points over the circadian cycle (ZT3,16,20,24). S1 Table summarizes the label-free data and compares it to the phosphorylation sites identified in the 15N/14N-labeled MS analysis. PER O-GlcNAcylation site identification was not performed as part of the label-free MS analysis. We observed a high level of congruence between the PER phosphorylation sites identified in our two fly head tissue data sets, as well as those generated using Drosophila S2 cells as starting materials [17–19,21]. In summary, our MS analysis sets the stage for future studies to understand the functional interplay between PER O-GlcNAcylation and phosphorylation. Subsequent to the identification of PER O-GlcNAcylation sites, we proceeded to analyze the function of site-specific PER O-GlcNAcylation by mutating one or a cluster of S/T residues to non-O-GlcNAcylatable alanine on the per gene and evaluated PER repressor function on CLK activity using the per-luciferase (per-luc) reporter assay in Drosophila S2 cells [3,37]. To prioritize, we analyzed PER O-GlcNAcylation residues with a Mascot score >50 (Table 1). We compared per-luc activity in S2 cells expressing per wild type or mutant variants (Fig 2A). In order to take into account the varying expression level of the PER variants (Fig 2B top panel, S3A and S3B Fig), we normalized CLK-activated per-luc activity observed for PER (WT) and PER mutant variants to their respective protein expression levels to more accurately assess the impact of blocking PER O-GlcNAcylation at specific residues on repressor activity (Fig 2C). We observed that all but one PER O-GlcNAcylation site mutant (PER (S174A) ) exhibited significant increase in repressor activity. Interestingly, PER (S942A) and PER (T951A-S954A) are the two O-GlcNAc site mutants that exhibited the strongest repressor activity, further strengthening the rationale for testing the hypothesis that O-GlcNAcylation in the CBD may be important in modulating PER-CLK interaction and PER repressor activity. For this reason, this current study will focus on investigating the function of PER O-GlcNAcylation in the CBD, while analysis of other PER O-GlcNAcylation residues will be pursued in the future. To confirm this finding in whole animals, we generated transgenic flies expressing p{per (WT) -HA10HIS} (herein referred to as per (WT) and mutant variants) in the per0 genetic background [38] so that only transgenic per (WT or mutant) was expressed. First, we performed quantitative RT-PCR to measure clock gene expression (per and tim) in heads of per (S942A) flies to examine the function of PER (S942) O-GlcNAcylation in central clock oscillators. All flies used for molecular analysis are homozygous for the per transgene. As predicted from the elevated PER repressor activity observed in S2 cells, we found that cycling of per and tim mRNAs was significantly dampened in per (S942A) flies as compared to per (WT) flies in both LD cycles and constant darkness (DD) (Fig 2D and 2E). Furthermore, per and tim mRNAs exhibited earlier initiation of repression phase in DD (Fig 2E). Since the per (S942A) mutant exhibited lowered levels of clock gene mRNAs, PER protein level is expected to decrease. As expected, peak PER abundance was significantly reduced in the heads of per (S942A) flies than in per (WT) flies in both LD and DD conditions (Fig 2F and 2G). To ensure that lower level of PER protein is a result of increased PER (S942A) repression on CLK-dependent transcription of per rather than reduced PER stability, we monitored the rate of PER degradation by cycloheximide (CHX) chase assay in S2 cells. The PER (S942A) mutant degraded at a similar rate as PER (WT) in the presence of OGT, demonstrating that the S942A mutation has no significant effect on PER stability (S3C and S3D Fig). Since peripheral tissues are known to be more sensitive to metabolic fluxes [39,40] and O-GlcNAcylation is a nutrient-sensitive PTM, we also analyzed the effect of blocking PER (S942) O-GlcNAcylation in oscillators of peripheral tissues, specifically the fat body. The fat body, which is analogous to mammalian liver and adipose tissue, plays an essential role in regulating energy metabolism in insects [39,40]. We assayed per and tim mRNA levels in the fat body of per (WT) and per (S942A) flies on DD1 after LD entrainment to evaluate PER-dependent repression of CLK activity. Consistent with what we observed in fly heads, per (S942A) flies displayed significantly dampening of per and tim mRNA cycling as compared to per (WT) flies in the fat body (Fig 2H and 2I). This suggests that PER (S942) O-GlcNAcylation normally weakens the activity of PER to repress CLK-dependent transcription in both head and fat body oscillators. As in PER (S942A) mutant, PER (S951A/T952A/T954A) mutant also exhibited a significant difference in repressor activity as compared to PER (WT) in per-luc reporter assay (Fig 2C). To evaluate the effects of these residues, which are also located in the CBD, we generated wper0; per (S951A/T942A/T954A) transgenic flies to confirm the effects of the PER (S951A/T952A/T954A) mutations in whole animals. We measured temporal cycling of per and tim mRNAs in heads of per (S951A/T952A/T954A) flies and observed dampened expression of per and tim mRNA as compared to per (WT) flies in both LD and DD conditions (S4A and S4B Fig), although not to the extent observed in per (S942A) flies. However, this did not translate into significant differences in PER protein abundance and cycling (S4C and S4D Fig). The differential effects observed when blocking O-GlcNAcylation at PER (S942) and PER (S951/T942/T954) suggest that PER (S942) O-GlcNAcylation is a key event in regulating PER repression within the CBD. It is important to stress that PER (S942) has never been identified as a phosphorylation site in all previous comprehensive mapping studies [17–19,21], so it is highly unlikely that the effects for S942A mutation is due to disruption of PER (S942) phosphorylation. Given that several of the non-O-GlcNAcylatable per mutants exhibited elevation in PER repressor activity in S2 cells and in flies, we proceeded to investigate if this molecular phenotype can translate to alterations in output behavioral rhythms. We first evaluated daily locomotor activity rhythms of per (WT) and mutant flies. Activity rhythm is a robust readout that reflects the function and speed of the central oscillators located in fly heads [30]. Flies were entrained for 3 days in LD cycles followed by 7 days in DD. As expected, heterozygous per (WT) flies manifested robust rhythms with ~24-hr periods, indicating full rescue of arrhythmic per0 mutation (Fig 3A). Homozygous per (WT) flies displayed somewhat shorter behavioral rhythms due to the extra copy of per. Interestingly, as compared to per (WT), per (S942A) and the triple per (S951A/T952A/T954A) mutants homozygous for their respective transgenes exhibited shorter periods by 1. 4 and 0. 7 hrs. The earlier initiation of repression previously observed in per (S942A) mutant likely contributes to its short-period phenotype. In comparison, per (S951A/T952A/T954A) mutants displayed a smaller though significant change in repression activity, which likely accounts for the smaller change in period length. In addition to monitoring locomotor activity rhythms of per (S942A) flies, we also examined the effects of blocking PER (S942) O-GlcNAcylation in peripheral oscillators by measuring feeding rhythms. Feeding assays were only performed to compare per (WT) and per (S942A) flies as per (S951A/T952A/T954A) mutants displayed minor changes in activity rhythms and clock gene repression. Feeding activity rhythms are governed by oscillators of metabolic tissues (i. e. in the fat body) [40], which are more sensitive to nutrient flux than central oscillators in the brain. Since we demonstrated earlier that clock gene cycling was dampened in the fat body of per (S942A) flies as a result of PER being a stronger repressor, we expected that feeding rhythms will also be disrupted. We entrained per (WT) and per (S942A) flies in 12h: 12h LD cycles and assayed their feeding activity rhythms for two consecutive days either in LD or DD conditions using the CAFE assay [41]. The amount of food consumed over a 2-hr period was determined at 4-hr intervals over the circadian cycle. As control, per0 flies were measured in parallel, and as expected they exhibited arrhythmic feeding activity in LD and DD (Fig 3B and 3C) as determined by JTK-cycle (P = 1) [42]. per (WT) flies displayed robust rhythms of feeding behavior (JTK-cycle (P < 0. 01) ); they feed during daytime and fast at night in both LD and DD conditions (Fig 3B and 3C). Robust feeding rhythms were also observed in per (S942A) flies in LD (JTK-cycle (P < 0. 01) ), but the peak was phase advanced. This is consistent with the shorter period length in locomotor activity observed in the per (S942A) mutant. The consequence of blocking O-GlcNAcylation at PER (S942) on feeding behavior is even more severe in DD in the absence of light cues as rhythmic feeding was abolished in DD (JTK-cycle (P = 1) ). Together, our data demonstrates that O-GlcNAcylation at PER (S942) modulates the function of central and peripheral oscillators by regulating PER activity. Premature PER nuclear entry during the night could explain why per (S942A) mutant exhibits stronger and/or premature initiation of repression on CLK activity. A previous study showed that PER O-GlcNAcylation is involved in regulating nuclear entry [23]. To examine this possibility, we monitored timing of PER nuclear entry in adult clock neurons in LD cycles. In the fly brain, rhythmic expression and nuclear localization of PER in lateral clock neurons (LNvs) are necessary for proper clock function and maintenance of rhythmic locomotor activity (reviewed in [2]). The lateral neurons (small and large; l-LNvs and s-LNvs) express pigment dispersing factor (PDF), which has been commonly used to label the cytoplasm of the LNvs [43]. Despite having shorter behavioral rhythms, we found that per (S942A) mutants did not exhibit a significant difference in accumulation of nuclear PER as compared to per (WT) flies at time-points when nuclear entry is most prominent, ZT18 to ZT22 [4] (S5A and S5B Fig). Although a previous report demonstrates that global knockdown or overexpression of OGT in clock neurons affects the timing of PER nuclear entry [23], it is not surprising that site-specific non-O-GlcNAcylatable per mutants, in this case per (S942A), may not show the same phenotype. As such, our results suggest that the stronger and premature initiation of CLK repression observed in heads and fat bodies of short-period per (S942A) flies cannot be explained by altered timing of PER nuclear localization in circadian oscillators. Given that PER does not enter the nucleus prematurely in the LNvs of per (S942A) mutant flies, we speculate that the PER (S942) may have a higher affinity to CLK that results in stronger and/or premature initiation of repression. We therefore examined PER-CLK interactions by performing co-immunoprecipitations (coIP) using head extracts from per (WT) and per (S942A) flies. These experiments were performed in head tissues as IP reactions of clock proteins are more efficient in head tissues and produce more robust results. Indeed, the per (S942A) flies showed significantly higher PER-CLK interaction at ZT16 than per (WT) flies (Figs 4A, 4B and S6A). To validate our results, we performed coIPs using S2 cells coexpressing per (WT) -cmyc or per (S942A) -cmyc with clk-V5, in the presence or absence of pMT-ogt-FLAG. Consistent with the fly data, we observed that PER (S942A) proteins exhibited significantly stronger binding to CLK as compared to PER (WT) (Fig 4C and 4D). Interestingly, when PER (S942A) mutant was coexpressed with CLK in the presence of OGT, the level of PER-CLK binding was lower than the levels when PER (S942A) mutant was coexpressed with CLK in the absence of OGT. This suggests that O-GlcNAcylation of CLK or perhaps at other PER residues may negatively impact PER-CLK interaction. Nevertheless, our results suggest that stronger and/or premature initiation of repression phase in per (S942A) flies is due to higher affinity between CLK and PER (S942A). In comparison to per (S942A) mutant flies, per (S951A/T952A/T954A) flies only displayed minor changes in activity rhythms and clock gene expression in flies. Based on these observations, the PER (S951A/T952A/T954A) mutant protein is predicted to exhibit minor changes on PER-CLK interactions, despite that these three residues are also localized within the CBD. As expected, no significant changes in PER-CLK interactions were observed between per (WT) and per (S951A/T952A/T954A) flies at the indicated time-points (S7A and S7B Fig). Thus, our results suggest that O-GlcNAcylation events at PER (S951/T952/T954) only have minor modulatory effects on PER-CLK interactions. Alternatively, it is possible that only PER (S942) is O-GlcNAcylated within the CBD (Fig 1). Since part of PER repressor function is to remove CLK from clock gene promoters, we would expect that CLK may be removed prematurely when PER (S942) O-GlcNAcylation is blocked. We therefore measured CLK occupancy at the E-box elements of per and tim promoters by performing chromatin immunoprecipitation (CLK-ChIP) using extracts from adult fly heads. Consistent with our hypothesis, we found that flies expressing per (S942A) showed significantly reduced CLK occupancy on per and tim promoters at ZT18 and ZT20 as compared to per (WT) flies (Fig 4E). To rule out the possibility that decreased CLK occupancy observed at ZT18 and ZT20 in per (S942A) flies was due to decreased CLK levels, we examined CLK levels after the ChIP reactions and found that CLK was not limiting in flies expressing per (WT) or per (S942A) (S6B Fig). To further support our hypothesis that dynamic O-GlcNAcylation at PER (S942) is critical for regulating the strength and/or timing of PER-dependent clock gene repression, we overexpressed 3XFLAG-ogt in tim-expressing clock neurons using a tim (UAS) -gal4 driver (referred to as TUG) [44] and examined PER-CLK interactions in head extracts of these flies. First, we verified that ogt overexpressors (TUG>FLAG-ogt) exhibited a significant ~2-hr period-lengthening of behavioral rhythms, as previously observed [23,24] (Fig 5A). All 4 independent ogt overexpressor fly lines we generated showed similar period-lengthening phenotypes as compared to parental controls. To confirm that the overexpressed OGT enzyme is functional, we optimized O-GlcNAc chemoenzymatic labeling to measure levels of O-GlcNAcylated proteins in fly head extracts. Currently, detecting protein O-GlcNAcylation remains to be a challenge when using traditional protein analytical techniques because the addition of a sugar group does not influence the migration of a polypeptide in gel electrophoresis (reviewed in [45]). Additionally, commercially available O-GlcNAc-specific antibodies or lectin yield non-specific signals in our hands. Thus, we chose to use a more sensitive and specific chemoenzymatic labeling method to examine in vivo protein O-GlcNAcylation status in this study [46–48]. We first tested the specificity of this approach by detecting O-GlcNAcylated PER and OGT in Drosophila S2 cell culture. S2 cells were transiently transfected with V5 tagged per with or without pMT-ogt-FLAG. Immunoprecipitated PER and OGT were labeled with resolvable mass tags prior to SDS-PAGE and western blotting. Results showed successful labeling of O-GlcNAcylated PER with 20 kD PEG mass tag in the presence of OGT, leading to clear mobility shift of O-GlcNAcylated PER isoforms (S8A Fig). As OGT is known to be auto-O-GlcNAcylated, addition of 10kD PEG mass tag in the reaction resulted in O-GlcNAcylated OGT isoforms that appeared as slower migrating bands as detected by Western blots (S8B Fig). The use of PEG mass tag labeling provides two advantages. First, both modified and non-modified isoforms can be detected, providing information on stoichiometry. Second, epitope-tagged antibodies or target-specific antibodies can be used for detection of all isoforms. Alternatively, O-GlcNAcylated proteins can also be labeled with biotin to facilitate detection with α-streptavidin antibodies [46–48]. Our success in detecting and visualizing in vivo O-GlcNAcylation status of proteins in S2 cells by chemoenzymatic labeling allowed us to proceed and confirmed that PER O-GlcNAcylation was elevated in head extracts of ogt overexpressor flies, an indication that overexpression of ogt led to increase in enzyme activity (Fig 5B and 5C). Since blocking PER (S942) O-GlcNAcylation promotes PER-CLK interaction, we expect that PER-CLK interaction should be reduced in ogt overexpressors assuming PER (S942) is hyper-O-GlcNAcylated. As anticipated, we observed that control TUG flies exhibited significantly higher PER-CLK interaction at ZT18 as compared to ogt overexpressor flies in coIP assays using head extracts of flies (Fig 5D and 5E). For ogt overexpressor flies, the reduction in PER-CLK interaction observed in fly heads may partially account for the physiological effect of OGT overexpression, i. e. period-lengthening of clock-controlled locomotor activity rhythms (Fig 5A). We speculate that PER (S942) O-GlcNAcylation occurs either during the daytime or early night to prevent de novo PER from prematurely binding to CLK or at late night to facilitate CLK dissociation from PER after PER-dependent repression. Both of these scenarios would result in stronger clock gene repression in per (S942A) flies. Since we observed that the differences in PER-CLK interactions between per (WT) and per (S942A) flies as well as between WT TUG control and ogt overexpressor flies occurred at around the start of the circadian repression phase (i. e. ZT16 and ZT18 respectively; Fig 4B and 5E), we postulated that the former scenario may be more likely. Temporal data on PER (S942) O-GlcNAcylation status would certainly help to rule out one of the two scenarios. It is therefore unfortunate that we are not able to determine the timing of PER (S942) O-GlcNAcylation by MS. Kim et al. [23] and Kaasik et al. [24] have previously examined global PER O-GlcNAcylation status over the circadian cycle using anti-O-GlcNAc antibodies in combination with Western blotting, but their results were incongruent. Whereas Kim et al. observed that PER O-GlcNAcylation peaks at ZT20 [23], Kaasik et al. observed a peak in PER O-GlcNAcylation at ZT10 [24]. We therefore opted to profile global PER O-GlcNAcylation status by chemoenzymatic labeling to gain insights into the timing and function of PER (S942) O-GlcNAcylation. We examined the temporal profile of PER O-GlcNAcylation in head extracts of per (WT) flies in LD condition by labeling PER with either 20kD PEG or a biotin tag [46–48]. In both cases, maximal PER O-GlcNAcylation occurred at around ZT4 to ZT8 with subsequent decline from ZT12 to ZT24 (Fig 6A and 6B). Our results were more in line with the temporal PER O-GlcNAcylation profile observed in Kaasik et al. [24]. Daily PER O-GlcNAcylation cycle was found to be significantly rhythmic (JTK cycle; P < 0. 0005) (Fig 6C). Our results reveal that PER is more highly O-GlcNAcylated during the day. This suggests that PER (S942) is likely O-GlcNAcylated during the day and may persist into early night to prevent de novo PER from prematurely interacting with CLK to initiate the repression phase. O-GlcNAcylation of cellular protein is sensitive to nutrient input [25,49]. Since we observed that PER O-GlcNAcylation is higher during the day and gradually decreases over the circadian cycle, we expect that this temporal pattern may correlate to daily feeding activity. We measured feeding rhythms in per (WT) flies fed ad libitum using the CAFE assay [41]. The flies used for these assays were entrained simultaneously with flies used for PER O-GlcNAc labeling to better assess correlation between feeding activity and PER O-GlcNAcylation. We found that mixed sexes of per (WT) flies displayed rhythmic feeding that peaked during early day in LD condition (S9A Fig). Similar results were obtained when male and female per (WT) flies were housed and tested separately (S9B Fig). Furthermore, a separate experiment comparing feeding activity rhythms of per (WT) and per (S942A) flies also showed higher daytime feeding activity in per (WT) flies (Fig 3B). Taken together, our data suggest that nutrient flux via feeding activity provides time-of-day metabolic signals to the circadian oscillator via temporal O-GlcNAcylation of PER. Specifically, PER (S942) O-GlcNAcylation, which is expected to occur during the day and persist into early night, may prevent interaction of PER and CLK prematurely to regulate timing of circadian repression.
Recent studies reveal that O-GlcNAcylation of circadian transcription factors, PER and CLK in Drosophila and BMAL1 and CLOCK in mice, plays an important role in modulating their function in the circadian oscillator [23–25,50]. However, to more fully understand the mechanisms by which site-specific O-GlcNAcylation events regulate circadian physiology and to set the stage for investigating the interplay between phosphorylation and O-GlcNAcylation, it is necessary to identify O-GlcNAcylated residues in core clock proteins and other cellular proteins and characterize their site-specific functions. Furthermore, the relationship between feeding-induced nutrient influx and the temporal regulation of clock protein O-GlcNAcylation warrants investigation in whole animals. Since feeding activity is controlled by the circadian clock, we hypothesize that food intake will increase HBP influx, leading to increase in clock protein O-GlcNAcylation during the feeding period or soon after. This could serve as a mechanism by which metabolic input reinforces circadian entrainment by other zeitgebers and regulates oscillator function in a time-of-day specific manner. We therefore set out to (i) identify PER O-GlcNAcylation sites and characterize their site-specific functions in regulating the circadian oscillator; and (ii) determine if there is a correlation between time of feeding and O-GlcNAcylation levels of PER. Although overexpression or knockdown of OGT and OGA has provided insights into the global effects of protein O-GlcNAcylation on circadian clock regulation, we expect that site-specific characterization of O-GlcNAcylation events will alleviate confounding effects resulting from having multiple O-GlcNAcylated residues within a single protein or the involvement of multiple O-GlcNAcylated proteins in clock regulation, leading to new mechanistic insights. By utilizing MS-based proteomics, we observed that in addition to being heavily phosphorylated, PER is O-GlcNAcylated at at least 6 residues, some of them in the CBD. To understand the role of these O-GlcNAcylation events in regulating PER function and circadian physiology, we analyzed these O-GlcNAcylation sites by replacing serine or threonine to non-GlcNAcylatable alanine either singly or in clusters. Several of these non-O-GlcNAcylatable per mutants exhibit changes in PER repressor function, which consequently result in period-changing phenotypes in their corresponding transgenic fly lines. In particular, we observed that O-GlcNAcylation at PER (S942), which is located in the CBD, reduces PER-CLK interaction (Fig 7A and 7B). Using O-GlcNAc chemoenzymatic labeling, we show that PER O-GlcNAcylation primarily occurs during daytime and correlates with the time period when animals are feeding. We therefore postulate that PER (S942), as in the case for most PER O-GlcNAcylation sites, is O-GlcNAcylated during the day. This ensures that PER does not interfere with CLK activity in the circadian activation phase and its repression of CLK activity does not initiate prematurely when de novo PER starts to translocate into the nucleus. This suggests that OGA may need to actively remove O-GlcNAc from PER residues prior to circadian repression phase. Indeed, OGA level has been shown to oscillate over the circadian day, peaking prior to initiation of circadian repression phase [24]. Moreover, O-GlcNAcylation has been shown to regulate PER nuclear entry [23], suggesting that OGA-dependent removal of O-GlcNAc at unknown PER residue (s) is likely required to facilitate PER nuclear translocation independent of OGA activity on PER (S942). Finally in the evening, since flies are fasting, the level of O-GlcNAcylation at PER (S942) will remain low allowing PER to bind strongly to CLK to repress its activity. Our findings that PER O-GlcNAcylation at S942 reduces PER repression of CLK activity is not congruent to the observation in [24], where they observed that the repressor activity of PER is enhanced when coexpressed with OGT in S2 cell per-luc reporter assay. However, this apparent incongruence could be explained by the combined activities of other O-GlcNAcylation events on PER, CLK, or other cellular proteins that impact per-luc reporter gene expression. Nevertheless, it is important to point out that none of our non-O-GlcNAcylatable per mutants showed a decrease in repressor activity (Fig 2C). It is interesting to note that unlike some phosphorylation sites previously identified on PER proteins, residues we identified to be O-GlcNAcylated in Drosophila PER are not conserved in mouse PER proteins. This is perhaps not surprising since nutrient-dependent O-GlcNAcylation on cellular proteins likely depends on when organisms are actively feeding, i. e. whether they are diurnal or nocturnal. For instance, mice are nocturnal, which may restrict protein O-GlcNAcylation of cellular proteins to nighttime. This nighttime peak in O-GlcNAcylation levels corresponds to when mouse PER2 proteins are abundant and active as repressors, suggesting that O-GlcNAcylation may act to promote PER2-dependent repression on circadian transcription. On the other hand, given that flies are diurnal and feed during the day (Fig 3B), nutrient flux promoting O-GlcNAcylation of Drosophila PER inhibits PER function as a transcriptional repressor (Fig 7). Overall, the differences in timing of feeding activity and nutrient flux between diurnal and nocturnal animals will present interesting opportunities for comparative analysis with regard to site-specific and global effects of O-GlcNAcylation on cellular protein function. In addition to regulating PER-CLK interaction in the Drosophila circadian clock, O-GlcNAcylation is expected to affect oscillator function via other mechanisms [23–25]. Besides regulating clock proteins directly, it is likely that O-GlcNAcylation can modify the activity of clock kinases, just as GSK3β can regulate the activity of OGT [24]. CK2 and GSK3β are two clock kinases that have been shown to be substrates of OGT [51–52], and currently it is unclear how OGT might modulate their activities in clock regulation. Furthermore, O-GlcNAcylation can also impact the activities of chromatin modifiers and transcription machineries, including RNAPII [53]. The interplay between O-GlcNAcylation and other PTMs to regulate circadian physiology is expected to be complex and extensive, and reach far beyond the molecular oscillator and circadian transcription. In summary, our results support that clock-controlled feeding activity provides metabolic input to reinforce entrainment signals by light-dark cycles to regulate circadian physiology via clock protein O-GlcNAcylation. We expect that circadian rhythms in peripheral systems, where oscillators are more sensitive to metabolic input, to be particularly sensitive to modulation via O-GlcNAcylation of clock proteins. Finally, our results imply that disruptions in daily feeding rhythms, e. g. irregular meal times and late night eating common in modern societies, will likely affect rhythms in protein O-GlcNAcylation and interplay with other PTMs, thereby disrupting circadian rhythms in physiology. Future experiments to manipulate feeding schedules by time-restricted feeding (TRF) [54] can further solidify the causal relationship between feeding-induced nutrient influx and O-GlcNAcylation of cellular proteins. They can also provide mechanistic insights into the benefits of TRF.
To profile PER PTMs, we generated transgenic flies that expressed a 13. 2 kb genomic clone of per in w1118 per0 (wper0) background. A previously characterized vector that contains a 13. 2kb per genomic fragment tagged with HA and 10X histidine at the carboxyl terminal (pCaSpeR-per (13. 2WT) -HA-10HIS) [55] was used as the template for inserting 3XFLAG at the amino terminal before the starting Methionine to facilitate FLAG Affinity Purification. Transformants were generated by P-element transformation (BestGene Inc. , Chino Hills, CA), and the 3XFLAG-per (13. 2WT) -HA-10HIS transgene was tested for functionality by determining whether it can rescue wper0 flies in behavioral and molecular assays. To generate transgenic flies carrying wild-type (WT) or O-GlcNAcylation site mutants of per, we opted to use PhiC31 site-directed recombination [56]. The genomic per (13. 2WT) -HA-10HIS was excised from pCaSpeR-per (13. 2WT) -HA-10HIS using the restriction sites XhoI and BamHI and subcloned into pattB vector (kind gift from Amita Seghal) to yield pattB-per (13. 2WT) -HA-10HIS. The pattB vector was modified so that the sites KpnI and XbaI were removed, the BgIII site was replaced by a BamHI site, and the BamHI site was replaced by a BgIII site. For generating flies expressing non-O-GlcNAcylatable mutants on per, a 3. 4 kb XbaI-BamHI or 1kb BamHI-KpnI genomic fragment was excised from the pattB-per (13. 2WT) -HA-10HIS plasmid and subcloned into a pGem7 vector [20]. The resulting pGem7-per (XbaI-BamHI) or pGem7-per (BamHI-KpnI) plasmid served as a parent template for site-directed PCR mutagenesis (Agilent Technologies, Santa Clara, CA) depending on the location of the O-GlcNAc site (S2 Table for mutagenic primer sequences). After mutagenesis and confirmation by Sanger sequencing (UC Davis Sequencing), the mutant variants of 3. 4 kb or 1 kb per subfragments were used to replace the WT fragments in pattB-per (13. 2WT) -HA-10HIS. Plasmids were injected into w1118 fly embryos carrying attP sites on chromosome 3 (attP2) (BestGene, Inc.). Transformants were crossed with wper0 flies to remove endogenous copies of per prior to behavioral and molecular analyses. To generate flies overexpressing ogt, ogt was amplified and subcloned into pUAST-attB vector [57]. 3XFLAG was added to the N-terminus of ogt during the cloning process. Plasmids were injected into w1118 fly embryos carrying attP sites on chromosome 3 (attP2) (BestGene, Inc.). To express 3XFLAG-ogt in clock neurons, transgenic flies carrying the UAS-FLAG-ogt transgene were crossed to w; tim- (UAS) -GAL4 (referred to as TUG) driver line [44]. CAFE assay was performed as described [41] with modifications. Mixed-sex population of five male and five female wper0, wper0; per (13. 2WT), and wper0; per (S942A) flies or separately housed male or female (10 per group) flies were fed Bloomington Drosophila Stock Center standard fly food during entrainment in 12 h light/12 h dark day-night cycles and food consumption were measured starting on the third day of LD or the first day of DD. Prior to the day of measurement, grouped flies were transferred to a vial containing 2% agar as the medium with 5% sucrose solution maintained in calibrated glass micropipettes (VWR). After 24 hours of training, old micropipettes were replaced by fresh experimental micropipettes filled with 5% sucrose solution approximately 2 hours before each of the indicated time-points. After 2 hours, the amount of liquid consumed from the experimental micropipette was recorded, and the evaporation effect was evaluated by measuring the change in liquid volume in a micropipette placed in a vial without flies. Food consumption for each group/vial was determined by subtracting the amount of liquid consumed from the experimental micropipette with the amount of evaporated liquid. These values were normalized to the amounts of flies in the vial that survived until the end of the experiment. These experiments were performed in biological triplicates (one group/vial represents one independent experiment). Error bars = SEM. Rhythmicity of feeding was determined by JTK Cycle [42]. Proteins from S2 cells and fly heads were extracted using modified RIPA buffer as previously described [58]. Extracts were quantified and either directly analyzed by immunoblotting or incubated with 15μ α-V5 resin (Sigma) or 20μl of α-HA resin (Sigma) at 4°C for 4 hours. Beads were washed once in M-RIPA and twice with reaction buffer (20mM HEPES pH 7. 9,50mM NaCl, 1μM PUGNAc, 25mM NaF, 0. 5mM PMSF, and 5mM MnCl2) supplemented with 1x protease inhibitor (Sigma) [59]. Procedures for chemoenzymatic labeling with biotin or PEG (Polyethylene Glycol) 10kD or 20kD mass tag were performed as described [48] with modifications. Attachment of biotin or PEG mass tag to O-GlcNAc group requires a two-step derivatization process [59]: (1) a mutant galactosyltransferase, GalT1 (Y289L), utilizes UDP-azidogalactose (UDP-GalNAz) as substrate to add an azide onto the O-GlcNAc group; (2) Biotin alkyne or an alkyne-functionalized PEG mass tag indirectly attaches to the O-GlcNAc group via azide-alkyne cycloaddition chemistry. Briefly, after immunoprecipitation followed by washes, immune complexes were resuspended in 20 μl of reaction buffer containing 2 μl of Gal-T1 Y289L (Invitrogen) and 2μl of 0. 5mM UDP-GalNAz (Invitrogen) [59] following overnight incubation at 4°C with gentle rotation. Azide-labeled beads were washed twice with reaction buffer and subsequently resuspended in 50μl of labeling buffer (1% SDS and 50mM Tris-HCl pH 8. 0). The samples were reacted with biotin alkyne (Invitrogen) or an alkynyl-functionalized poly (ethylene glycol) (10-kDa or 20-kD) (Creative PEGWorks, Chapel Hill, NC) according to manufacturer’s protocol or a previously described protocol by [59] respectively. Samples were eluted in 50μl of 1XSDS sample buffer. For proteins labeled with a mass tag (PEG), PER or OGT was resolved using SDS-PAGE (5% or 8% minigel with a 40: 1 acrylamide/Bis-acrylamide solution, Bio-rad, Hercules, CA). For proteins labeled with biotin, PER was resolved using SDS-PAGE (5% Criterion gels, Bio-rad). Antibody dilutions to detect O-GlcNAcylated PER or OGT proteins are as follows: α-V5 (1: 5000), α-HA (1: 1000), α-FLAG (1: 7000), α-streptavidin (Cell Signaling Technologies, Danvers, MA) (1: 5000), and α-PER (GP5620; RRID: AB_2747405) (1: 2000). wper0; p{3XFLAG-per (13. 2WT) -HA10HIS} flies were fed with an 15N diet (0. 2g 15N yeast, 1% Bacto agar, 15% unsulfured molasses, phosphoric and propionic acid mix, and tegosept). Saccharomyces cerevisiae were metabolically 15N-labelled as described [29,60]. As control, flies were also fed with 14N diet. The adult progenies of 15N- or 14N-fed parental flies were reared in 15N- or 14N diet and entrained for 3 days in 12hr light: 12hr dark at 25°C and collected every four hours over a period of 24 hours on the fourth day. Upon collection, flies were immediately frozen on dry ice until protein extraction. For each time-point, 4ml of fly heads were homogenized into fine powder by a chilled mortar and pestle and resuspended in Lysis Buffer (20mM HEPES pH 7. 9,5% glycerol, 350mM NaCl, 0. 1% Triton X-100,1mM DTT, 1mM MgCl2,0. 5mM EDTA, 25mM sodium fluoride, 1x protease inhibitor (Sigma, St. Louis, MO), 1x PhosSTOP (Roche, South San Francisco, CA). Homogenate was dounced for 15 strokes using a 50ml tissue grind tight pestle (Kimble-Chase, Vineland, NJ) and were filtered using a 70μm cell strainer. Samples were spun at 300 rcf for 1 minute and then incubated at 4°C on a nutator for 30 minutes. Additional Lysis Buffer was added to dilute the sample from 350mM to 150mM NaCl before centrifuging at 15,000 rpm for 15 minutes at 4°C. Supernatant was collected and incubated with 200μl α-FLAG resin (Sigma) overnight over a nutator at 4°C. Next day, beads were washed twice for 15 mins with Lysis Buffer without EDTA, DTT, or PhosSTOP. Samples were eluted in 300ul R+A buffer (30% glycerol, 3% SDS, 6mM EDTA, 0. 15M Tris-HCl pH 6. 8) at 95°C. Eluate was reduced with 15μl 1M DTT for 10 minutes at 65°C and then alkylated with 35μl 1M IAA in room temperature for 20 minutes in the dark. Eluates were then flash frozen using liquid nitrogen immediately. We used the pooled standard approach to enable more accurate comparisons between different time points. 14N eluate from six time-points were pooled together and split evenly to mix with each 15N eluate at a 1: 1 ratio on ice. For each time-point, 600μl of cold acetone was added to the 14N/15N eluate mixture and placed in -20°C overnight. Precipitate were spun at 14,000 rpm for 10 minutes at 4°C and the resulting pellet were washed briefly with 1ml cold acetone. Precipitated eluate was resuspended in 80μl R+A sample buffer containing 3μl of 4X SDS sample buffer. The 14N/15N eluate was resolved in 12% SDS-PAGE and the excised PER band was used for protease digestion and analysis by mass spectrometry. For analysis of 14N/15N-labeled samples, proteins were digested in-gel with trypsin and elastase in separate reactions to result in overlapping peptides, such that individual modified sites can be determined. We have previously used this multi-protease approach [61] to maximize high sequence coverage when mapping PER phosphorylation sites [17]. For in-gel digestion the excised gel bands were destained with 30% ACN, shrunk with 100% ACN, and dried in a Vacuum Concentrator (Concentrator 5301, Eppendorf, Hamburg, Germany). Digests with trypsin and elastase were performed overnight at 37°C in 0. 05 M NH4HCO3 (pH 8). About 0. 1 μg of protease was used for one gel band. Peptides were extracted from the gel slices with 5% formic acid. NanoLC-MS/MS analyses were performed on an LTQ-Orbitrap Velos Pro or an Orbitrap Fusion (Thermo Fisher Scientific, Waltham, MA) equipped with an EASY-Spray Ion Source and coupled to an EASY-nLC 1000 (Thermo Fisher Scientific). Peptides were loaded on a trapping column (2 cm x 75 μm ID. PepMap C18,3 μm particles, 100 Å pore size) and separated on an EASY-Spray column (25 cm x 75 μm ID, PepMap C18,2 μm particles, 100 Å pore size) with a 90-minute linear gradient from 3% to 30% acetonitrile and 0. 1% formic acid. For the Oribtrap Velos MS scans were acquired in the Orbitrap analyzer with a resolution of 30,000 at m/z 400, MS/MS scans were acquired in the Orbitrap analyzer with a resolution of 7,500 at m/z 400 using HCD fragmentation with 30% normalized collision energy. A TOP5 data-dependent MS/MS method was used; dynamic exclusion was applied with a repeat count of 1 and an exclusion duration of 30 seconds; singly charged precursors were excluded from selection. Minimum signal threshold for precursor selection was set to 50,000. Predictive AGC was used with AGC target a value of 1e6 for MS scans and 5e4 for MS/MS scans. Lock mass option was applied for internal calibration in all runs using background ions from protonated decamethylcyclopentasiloxane (m/z 371. 10124). For the Orbitrap Fusion, both MS and MS/MS scans were acquired in the Orbitrap analyzer with a resolution of 60,000 for MS scans and 15,000 for MS/MS scans. HCD fragmentation with 35% normalized collision energy was applied. A Top Speed data-dependent MS/MS method applying HCD and ETD fragmentation from the same precursor with a fixed cycle time of 3 seconds was used. Dynamic exclusion was applied with a repeat count of 1 and an exclusion duration of 120 seconds; singly charged precursors were excluded from selection. Minimum signal threshold for precursor selection was set to 50,000. Predictive AGC was used with AGC a target value of 5e5 for MS scans and 5e4 for MS/MS scans. EASY-IC was used for internal calibration. Mascot Distiller 2. 5 was used for raw data processing and for generating peak lists, essentially with standard settings for the Orbitrap (high/high settings). Mascot Server 2. 5 was used for database searching with the following parameters: peptide mass tolerance: 7 ppm, MS/MS mass tolerance: 0. 02 Da, enzyme: “semi-trypsin” for tryptic digests and “none” for elastase digests; fixed modifications: carbamidomethyl (C); variable modifications: Gln->pyroGlu (N-term Q), oxidation (M), acetyl (protein N-term), phosphorylation (STY), HexNAc (ST). Searches containing both HCD and ETD spectra (Fusion) were searched separately for either b and y ions (HCD) or c and z ions (ETD). For ETD-searches different modifications definitions (without neutral losses) for phosphorylation and HexNAc were applied. Separate Mascot searches were performed for light peptides (quantitation: “none”) and heavy peptides (quantitation: 15N-labeling). Database searching was performed against a small custom database containing 187 of the most abundant proteins identified in these samples before in a first round search (without PTMs) against UniProt Drosophila database. This was necessary to limit search space and processing times. The results from the different Mascot searches (different time-points, proteases and fragmentation techniques) were merged (separately for light and heavy peptides) and filtered for phosphorylated and HexNAc-modified peptides using a custom software tool (A. Schlosser). A Mascot score cut-off of 15 and a delta score cut-off of 10 [62] were applied, and only “rank 1” peptides were accepted. For one modification site and one type of modification, only the peptide spectrum match (PSM) with the highest score was kept, all other PSMs were filtered out. All remaining spectra were verified manually, e. g. by checking the presence of modification specific marker ions. HexNAc-modified peptides were only accepted when at least one of the HexNAc-specific fragment ions (204,186 and 168) [63] was present in the corresponding HCD spectra. After manually filtering, all remaining peptides were exported to generate a summary of the results (Table 1). The N14/N15 MS data have been submitted to the Chorus repository (project ID 1424): (https: //chorusproject. org/anonymous/download/experiment/e47a30f7f2c749aba438652d7d88ef04) and (https: //chorusproject. org/anonymous/download/experiment/e6d6163b31bf40288606f827c6f18371). All flies were reared on standard Drosophila medium (Bloomington Drosophila Stock Center standard recipe). Entrainment and collection of flies at the appropriate time-points were described as above. Roughly 4ml of fly heads were grinded into fine powder using chilled ceramic mortar and pestle and mixed in 30ml of lysis buffer (20mM HEPES pH 7. 5,1mM DTT, 1x protease inhibitor). Homogenate was dounced and poured over a cell strainer as described above prior to centrifugation at 7000xg for 45 minutes at 4°C to separate nuclear and cytoplasmic lysates, repeated once. Lower layer (pellet) as the nuclear fraction from both spins was resuspended in 10ml Nuclear Extraction buffer (20mM HEPES pH 7. 5,10% Glycerol, 350mM NaCl, 0. 1% Triton X-100,1mM DTT, 1mM MgCl2,0. 5mM EDTA, 1x protease inhibitor, 10mM NaF) with the addition of MG132 (Sigma) and DNAse (Promega). Upper layer (supernatant) as the cytoplasmic fraction was supplemented with Lysis buffer with the addition of MG132 (Sigma) and DNAse (Promega). Nuclear and cytoplasmic fractions were incubated at 4°C for 30 minutes over a nutator. After incubation, nuclear fraction was diluted to 150mM NaCl with Lysis buffer. Nuclear and cytoplasmic fractions were centrifuged at 27,000rpm for 15 minutes at 4°C. Supernatant of nuclear and cytoplasmic samples was recovered before incubation with 200μl α-FLAG M2 resin (Sigma) at 4°C overnight. Beads were washed three times with Lysis buffer for 15 minutes each and subsequently eluted in 200μl of 3XFLAG peptide (Sigma) at a dilution of 250μg/ml at room temperature for 15 minutes. Eluates were resolved on a Tris-Tricine gel and PER bands were excised for protease digestion and mass spectrometry as described in [64]. The label-free MS proteomics data for PER phosphorylation site mapping have been deposited into ProteomeXchange (PXD008281) (ProteomeXchange: http: //proteomecentral. proteomexchange. org/cgi/GetDataset? ID=PXD008281), MassIVE repository (MSV000081736) (MassIVE: https: //massive. ucsd. edu/ProteoSAFe/dataset. jsp? task=384c7750b3004b7eac91054935a4e038), and Chorus repository (Project ID 1424) (https: //chorusproject. org/anonymous/download/experiment/1e0023a15da84e51bb18c55146104b32). Locomotor activity rhythms were measured as previously described [30]. 3-4-day old male flies were collected and subjected to 12hr light: 12hr dark (LD) cycles at 25°C for four days followed by seven days of constant darkness (DD) to measure free-running period using the Drosophila Activity Monitor System (DAMS) (TriKinetics, Waltham, MA). Data analysis was performed using FaasX as described in [30]. pAc-per-V5, pAc-3XFLAG-per-6Xcmyc, and pAc-clk-V5 were previously described [20,58]. For generating Serine/Threonine (S/T) to Alanine (A) O-GlcNAc site mutants (S2 Table for primer sequences), pAc-per-V5 or pAc-3XFLAG-per-6Xcmyc served as the template for site-directed PCR mutagenesis using QuikChange site-directed PCR mutagenesis (Agilent Technologies). All O-GlcNAc mutations were verified by Sanger sequencing. ogt cDNA (described above) was subcloned into a pMT-3XFLAG-6XHIS vector described in [20], with the epitope at the C-terminus of the ORF. pCopia-renilla luciferase and per-E-luc constructs were described previously [37]. S2 cells and DES expression medium were obtained from Life technologies (Carlsbad, CA), and transient transfections were performed using Effectene (Qiagen, Valencia, CA) according to manufacturer’s instructions and as previously described [17,20]. Luciferase reporter assays were performed as described [3,37]. Measurements of luciferase activity were performed using the Dual-Glo luciferase assay system (Promega, Madison, WI) according to the manufacturer’s recommendation on a TriStar LD 941 microplate reader (Berthold Technologies, Oak Ridge, TN). pAc-per-V5 (WT) or mutant variants were transfected into S2 cells with pMT-FLAG-ogt or an empty plasmid. 20 hours after transfection, ogt expression was induced for 16 hours. Cycloheximide was then added to a final concentration of 10 μg/ml. Cells were harvested and lysed with EB2 [20] at the times indicated. Proteins were analyzed by western blotting as detailed below and in [65]. S2 cell and adult fly head protein extractions, western blotting, and image analysis, were performed as previously described [58,65] with modifications. Primary antibodies α-V5 (Invitrogen, Carlsbad, CA) (1: 5000) was used to detect CLK-V5 and PER-V5, α-cmyc (9E10, Sigma, St. Louis, MA) (1: 5000) to detect PER-CMYC, α-FLAG (Sigma) (1: 7000) to detect FLAG-OGT, α-HA 3F10 (Roche, Indianapolis, IN) (1: 1000) to detect PER-HA, α-PER (GP5620; RRID: AB_2747405) [57] (1: 3000) to detect PER, and α-HSP70 (Sigma) (1: 10,000) was used for normalization. Secondary antibodies conjugated with HRP were added at final dilution as follows: α-mouse IgG at 1: 5000 for α-V5 detection, 1: 7000 for α-FLAG detection, or 1: 10,000 for α-HSP70 detection, α-guinea pig IgG at 1: 2000 for α-PER detection, and α-rat IgG (1: 1000) for α-HA detection of PER-HA. Membranes were imaged and protein levels were quantified using the ChemiDoc MP system with Image Lab software (Bio-Rad). To calculate PER degradation rate, PER intensity was normalized to HSP70 intensity at each time-point, and was then converted to a fraction of the peak value (peak = 1). For quantifying PER levels from fly heads, PER values were normalized against HSP70 intensity at each time-point, and subsequently expressed as a fraction of the peak PER levels. Co-IP assays using protein extracts from S2 cells and fly heads were performed as described [58,65] with modifications. Proteins were extracted using modified RIPA buffer with the addition of 100μM PUGNAc to preserve O-GlcNAcylation of proteins prior to input analysis by western blotting or Co-IP with appropriate antibodies. Samples were pre-cleared using sepharose beads (Sigma) to reduce nonspecific binding. For co-IP in S2 cells, CLK IP samples were incubated with 15μl α-V5 resin (Sigma) and negative non-specific control IP samples were incubated with 15μl α-HA resin (Sigma). For co-IP in fly heads, CLK IP samples were incubated with 4μl α-CLK antibody (Santa Cruz Biotechnology, Dallas, TX) for 3 hours prior to incubation with 20μl gamma sepharose beads (GE, Pittsburgh, PA) for 1 hour; PER IP samples were incubated with 20μl α-HA resin (Sigma). Immune complexes were resolved by SDS-PAGE as described [58,65]. IP signal intensity was normalized to intensity of the bait protein. These values were then converted as relative to the peak value of the dataset (peak = 1). Representative data shown are averages of normalized PER or CLK interactions from at least three independent experiments. Total RNA was extracted from fly heads and abdominal fat bodies using TRI-Reagent (Sigma). cDNA synthesis from total RNA and real-time PCR analysis was performed as previously described [58,65]. For isolating abdominal fat bodies, flies were collected at the indicated time-points on the first day of DD and immediately transferred in TRI-Reagent (Sigma) for 40 minutes at room temperature with agitation following dissection in RNAlater buffer (Thermo Fischer Scientific). At least 16 flies were dissected for each genotype and time-point. After dissection, fat bodies were rinsed with nuclease-free water twice and resuspended in TRI-Reagent prior to RNA isolation. ChIP assays were essentially performed as described [58]. qPCR of an intergenic region (FBgn0003638) of the Drosophila genome representing background CLK binding was subtracted from input samples. Technical triplicates for the qPCR step were performed for each of the three biological replicates. Two-tailed t-tests were used to determine statistical differences between genotypes at each time-point. Adult fly brain immunohistochemistry was performed as described previously [65] with modifications. Briefly, adult flies were entrained in LD cycles for 3 days and collected at the appropriate time-point following incubation with fixative solution (4% paraformaldehyde, 0. 2% Triton X-100 in PBS) for at least 40 minutes in the dark at room temperature with gentle rotation. Fixative solution was removed, and then wash solution (0. 2% Triton X-100 in PBS) was added to transfer flies into an embryo dish. Brains were dissected using #5 Rubis nano tweezers (Electron Microscopy Sciences, Hatfield, PA). Approximately 10 brains were dissected for each time-point. After dissection, brains were incubated in fixative solution at room temperature for 40 minutes with gentle rotation. Brains were rinsed quickly with wash solution three times prior to three slow washes in wash solution for 10 minutes each with gentle rotation. Wash solution was removed, and blocking solution (0. 2% Triton X-100,5% horse serum in PBS) was added to the brains for 40 minutes at 4°C with gentle rotation. Brains were then incubated with new blocking solution at 4°C overnight with primary antibodies at the following dilutions: α-HA 3F10 antibody (Roche) (1: 100), and α-PDF antibody (Developmental Studies Hybridoma Bank, Iowa City, IA) (1: 100). After ~18 hr, brains were rinsed quickly with wash solution three times prior to four slow washes with wash solution for 10 minutes each with gentle rotation. Brains were then incubated in secondary antibodies in blocking solution overnight. Secondary antibodies used at the following dilutions were Dylight88-conjugated α-rat (1: 100) and Alexa647-conjugated α-mouse (1: 100) (Thermo Fischer Scientific). After ~18 hr, brains were rinsed quickly with wash solution three times prior to four slow washes with wash solution for 10 minutes each with gentle rotation. Brains were rinsed quickly with PBS following incubation with 85% glycerol for 15 minutes. Brains were mounted on microscope slides in VectaShield mounting medium (Vector Laboratories, Burlingame, CA) under a #1. 5 (17-mm) coverslip. Confocal images were obtained with an Olympus FV1000 Confocal Microscope (Olympus Life Science, Center Valley, PA) and processed with the FV1000 software (Olympus Life Science). Scoring of PER subcellular localization was performed as previously described [66,67]. At least five brains were used for each genotype and time-point. For statistical analysis, scored LNvs from each brain served as one biological replicate. Two-tailed t-tests were used to determine statistical differences between genotypes at each time-point. | Circadian clocks are self-sustained, endogenous pacemakers that enable organisms to anticipate daily environmental changes and resource abundance to perform specific time-of-day activities and achieve optimal survival. Multiple time cues are interpreted by circadian clocks to facilitate synchrony between organisms and their environment. A large body of work have identified light and temperature as important zeitgebers. More recent works highlight the significance of metabolic cues as signals to entrain and modulate circadian clocks to drive proper rhythms of physiology and behavior. Metabolic input, primarily through clock-controlled feeding activity, can regulate circadian physiology through multiple pathways. Some of these pathways are unknown while others, such as the O-GlcNAcylation of clock proteins, are just emerging. In this study, we utilized mass spectrometry proteomics to identify O-GlcNAcylation sites of the Drosophila PERIOD (PER) protein, a key regulator of the clock, and performed site-specific functional characterization of PER O-GlcNAcylation. Our results support that PER (Ser942) O-GlcNAcylation, a nutrient-sensitive protein modification that is expected to be more abundant during feeding period, prevents newly synthesized PER from prematurely performing its function during daytime and therefore restricts its activity to nighttime when flies are fasting. This study provides new insights into the mechanisms linking nutrient input and circadian physiology. | Abstract
Introduction
Results
Discussion
Materials and methods | phosphorylation
invertebrates
gene regulation
regulatory proteins
dna-binding proteins
animals
circadian oscillators
animal models
drosophila melanogaster
model organisms
immunoprecipitation
experimental organism systems
transcription factors
chronobiology
extraction techniques
drosophila
research and analysis methods
protein extraction
animal studies
proteins
gene expression
insects
precipitation techniques
arthropoda
biochemistry
circadian rhythms
eukaryota
post-translational modification
genetics
biology and life sciences
organisms | 2019 | O-GlcNAcylation of PERIOD regulates its interaction with CLOCK and timing of circadian transcriptional repression | 19,414 | 340 |
Mycobacterium tuberculosis (Mtb) and helminth infections elicit antagonistic immune effector functions and are co-endemic in several regions of the world. We therefore hypothesized that helminth infection may influence Mtb-specific T-cell immune responses. We evaluated the cytokine profile of Mtb-specific T cells in 72 individuals with pulmonary TB disease recruited from two Sub-Saharan regions with high and moderate helminth burden i. e. 55 from Tanzania (TZ) and 17 from South Africa (SA), respectively. We showed that Mtb-specific CD4 T-cell functional profile of TB patients from Tanzania are primarily composed of polyfunctional Th1 and Th2 cells, associated with increased expression of Gata-3 and reduced expression of T-bet in memory CD4 T cells. In contrast, the cytokine profile of Mtb-specific CD4 T cells of TB patients from SA was dominated by single IFN-γ and dual IFN-γ/TNF-α and associated with TB-induced systemic inflammation and elevated serum levels of type I IFNs. Of note, the proportion of patients with Mtb-specific CD8 T cells was significantly reduced in Mtb/helminth co-infected patients from TZ. It is likely that the underlying helminth infection and possibly genetic and other unknown environmental factors may have caused the induction of mixed Th1/Th2 Mtb-specific CD4 T cell responses in patients from TZ. Taken together, these results indicate that the generation of Mtb-specific CD4 and CD8 T cell responses may be substantially influenced by environmental factors in vivo. These observations may have major impact in the identification of immune biomarkers of disease status and correlates of protection.
Helminth infections are endemic in many African countries with different prevalence depending on the geographic region helminth species and age of population [1]. Soil transmitted helminth infections are among the most common infections, transmitted via soil contaminated by eggs excreted from human faeces [2]. Of note, helminth infections are co-endemic in many geographic areas endemic for Mycobacterium tuberculosis (Mtb), HIV-1 and Plasmodium falciparum. Therefore, co-infections of helminths with Mtb, HIV-1 and/or Plasmodium falciparum occur in a large proportion of the subjects [3]. Mycobacterium tuberculosis (Mtb) is a facultative intracellular organism, obligate aerobe, infecting primarily lungs via the aerogenic route [4]. It has been recently estimated that 1. 7 billion people are infected with Mtb among which 5–15% will develop tuberculosis disease (TB) [5]. To date, the only vaccine available to prevent TB disease consist of an attenuate strain of M. bovis, the Bacillus of Calmette et Guérin (BCG). While BCG immunization protects from life-threatening disseminated forms of TB disease in children, its efficacy in adults is inconsistent [6]. The protective components of Mtb-specific immunity are partially defined. Several studies have underscored the essential role of IL-12/IFN-γ axis in the protection against Mtb infection [7–10]. In addition, an efficient CD4 T-cell response probably associated with type 1 cytokine secretion is associated with the control of Mtb infection, since a severe reduction in the CD4 T cell number during HIV infection or the suppression of their function following anti-TNF-α therapy are associated with increased risk of TB reactivation [11,12]. The current paradigm of human cellular immunity indicates that functionally-distinct CD4 T-cell populations are specifically involved against a variety of pathogens depending on their size and their intra- or extra-cellular localization. In this model, type 1 helper CD4 T cells (Th1 cells) intervene against viruses and intracellular pathogens, Th2 cells against parasites such as worms and Th17 cells against extracellular pathogens [13,14] including bacteria and fungi [14,15]. Consistent with this paradigm, the protective Mtb-specific T-cell response is usually ascribed to typical Th1 response with CD4 T cells producing cytokines such as IFN-γ or TNF-α that contribute to the recruitment of monocytes and granulocytes and activate the anti-microbial activity of macrophages [16,17]. By contrast, helminth infections induce IL-4/IL-5 and IL-13 producing CD4 T cells and regulatory T cells [18–21] associated with the alternative activation of macrophages for repair of tissues injured upon migration of worms across different body compartments [1]. With regard to the generation of the functionally distinct T helper CD4 T cell populations, the pioneering studies from Romagnani and others [22,23], clearly demonstrated the critical importance of the cytokine environment, rather than the specific antigen in the development of distinct T helper antigen-specific CD4 T cells [24,25]. In this regard, the presence of an IL-4 cytokine background favored the development of Th2 specific to pathogens that usually induce the classical Th1 responses [23,26]. Multiple studies performed in Mtb/helminth co-infected individuals have focused on the impact of helminth infection on 1) TB diagnosis [27–35], 2) reactivation of TB from latently infected individuals (LTBI) [36,37] and 3) BCG vaccine immunogenicity [38–40]. In addition, it was recently shown that helminth infection may interfere and/or influence innate [41], cellular [42–44] and humoral [45] immune responses to Mtb. However limited information is available on the cytokine profile of Mtb-specific T-cell immune responses in subjects with Mtb/helminth co-infection. In the present study, we investigated the cytokine profile of Mtb-specific immune response in patients with clinically active TB from two countries i. e. Tanzania (TZ) and South Africa (SA) in the presence or absence of active helminth infections. We provide evidence that the functional profile of Mtb-specific CD4 T cells of TB patients from TZ was characterized by a mixed Th1/Th2 cytokine profile, while that from SA was associated with a typical single IFN-γ and dual IFN-γ/TNF-α Th1 profile. These results demonstrate that distinct functional profiles of CD4 T-cell responses can be directed against the same pathogen, i. e. Mtb, in human populations from different geographic areas. Ad hoc designed studies will be needed to define the factors driving the distinct functional profiles as well as to determine whether the distinct functional profiles are associated with variation of the TB pathology and/or response to drug therapy.
The functional profiles of Mtb-specific CD4 T-cell responses were assessed by intracellular cytokine staining (ICS) according to the gating strategy shown in S2 Fig. In particular, the ability of Mtb-specific CD4 T cells to produce IFN-γ, TNF-α, IL-2, IL-4, IL-5 and/or IL-13 in response to ESAT-6 and CFP-10 peptide pools stimulation was assessed by multi-parametric flow cytometry in 25 TB patients and 30 Mtb/helminth co-infected patients from TZ and compared to 17 TB patients from SA. Of note, Th2 cytokines i. e. IL-4, IL-5 and IL-13 were all assessed in the same flow cytometry fluorescence channel, which allowed the assessment of total Th2 cytokine production but prevented direct identification of individual IL-4, IL-5 or IL-13 Mtb-specific CD4 T-cell responses. Cytokine profiles of Mtb-specific CD4 T cells from three representative TB patients from SA (#08) and TZ (TB (#60062) and Mtb/helminth co-infected patient (#60031) are shown in Fig 1A. We first compared the frequencies of cytokine-producing Mtb-specific memory CD4 T cells from TB patients from SA versus TB patients from TZ (Fig 1B). The cumulative data showed a significantly higher IL-2+ and IL-4/IL-5/IL-13+ Mtb-specific memory CD4 T-cell frequencies in TB patients from TZ compared with TB patients from SA (P<0. 05), while the frequencies of IFN-γ and TNF-α producing Mtb-specific memory CD4 T cells were not significantly different between TB patients from TZ and from SA (P>0. 05; Fig 1B). Interestingly, no significant differences were observed for Th1 and Th2 cytokine producing CD4 T cells between TB patients from TZ with and without ongoing helminth co-infections (P>0. 05) (Fig 1B). We next analyzed the cytokine profile of Mtb-specific memory CD4 T cells of TB patients from SA and TZ (Fig 1C, pie charts). The cytokine profile of Mtb-specific memory CD4 T cells of TB patients from SA was significantly different from of TB patients from TZ (P<0. 05; Fig 1C, pie charts). Again, no significant differences were observed between TB patients and Mtb/helminth co-infected patients from TZ (P>0. 05) (Fig 1C, pie charts). In depth analysis showed that Mtb-specific CD4 T-cell responses of TB patients from TZ were significantly enriched in polyfunctional IFN-γ+IL-2+TNF-α+IL-4/5/13- CD4 T cells (triple IFN-γ/IL-2/TNF-α Mtb-specific CD4 T cells) and in IFN-γ-IL-2-TNF-α-IL-4/5/13+ CD4 T-cell populations (single IL-4/5/13 Mtb-specific CD4 T cells) as compared to TB patients from SA (43–42% versus 20% for triple IFN-γ/IL-2/TNF-α and 14–16% versus 1% for single IL-4/5/13; P<0. 05) (Fig 1C). In contrast, Mtb-specific CD4 T-cell responses of TB patients from SA were significantly enriched in IFN-γ+IL-2-TNF-α+IL-4/5/13- (dual IFN-γ/TNF-α Mtb-specific CD4 T cells) and in IFN-γ+IL-2-TNF-α-IL-4/5/13- CD4 T-cell populations (single IFN-γ Mtb-specific CD4 T cells) as compared to TB patients from TZ (about 45% versus 20–18% for dual IFN-γ/TNF-α and 25% versus 7–4% for single IFN-γ+; P<0. 05) (Fig 1C). No significant differences were observed between the Mtb-specific CD4 T-cell cytokine profile of TB patients and Mtb/helminth co-infected patients from TZ (P>0. 05) (Fig 1C). Of note, Schistosoma mansoni-specific CD4 T-cell responses were evaluated on Mtb/helminth co-infected individuals from TZ using S. mansoni soluble egg antigens (SEA) by polychromatic flow cytometry (n = 7). The results obtained showed that SEA-specific CD4 T-cell responses were dominated by single TNF-α and single IL-4/IL-5/IL-13-producing CD4 T cells, while polyfunctional IFN-γ+IL-2+TNF-α+ CD4 T cells represented less than 5% of total SEA-specific CD4 T-cell responses confirming previous observations [46,47] (S3 Fig). To better estimate the influence of 1) different helminth species, 2) infection caused by more than one helminth (polyparasitism), 3) differences in helminth lung migration capacity and 4) past helminth exposure on the generation of Th2 Mtb-specific CD4 T-cells, the proportion of IL-4/5/13-producing Mtb-specific CD4 T cells was compared between TB patients from TZ with helminth infection caused by one or more helminth species (S4A Fig), or between TB patients from TZ coinfected with helminth species exhibiting (hookworms and Strongyloides stercoralis) or not (S. mansoni, Schistosoma haematobium and Wuchereria bancrofti) lung migration capacity (S4B Fig). Similarly, the proportion of IL-4/5/13-producing Mtb-specific CD4 T cells was compared between TB patients from SA or TZ with evidence of past exposure to helminths versus TB patients with no sign of ongoing or past exposure to helminths (S4C and S4D Fig). The cumulative data showed that the proportion of single IL-4/5/13-producing Mtb-specific CD4 T cells from active TB cases was not influenced by 1) polyparasitism (including W. bancrofti, hookworm, S. mansoni, S. haematobium or S. stercoralis) (P>0. 05) (S4A Fig), 2) by helminth infections caused by helminth species exhibiting lung migration capacity, or 3) by past exposure to helminths (S4B and S4C Fig). To further characterize the cytokine profile of Mtb-specific T-cells, multiplex bead array analyses (luminex) were performed on supernatants of ESAT-6/CFP-10 peptide pool (Mtb) -stimulated cell cultures. The cumulative data showed that Mtb-stimulated cell culture supernatants of TB patients from TZ secreted similar levels of IFN-γ, TNF-α and IL-10 (P>0. 05), but significantly higher levels of IL-2, IL-4, IL-5, IL-13, IL-17A and IL-17F than Mtb-stimulated cell culture supernatants of TB patients from SA (P<0. 05; Fig 1D). However, no significant differences were observed between TB patients and Mtb/helminth co-infected patients from TZ (P>0. 05; Fig 1D), confirming our flow cytometry analyses. In addition, the levels of Th2 cytokines secreted in Mtb-stimulated cell culture supernatants of Mtb/helminth patients from TZ was not influenced by individual species of helminth infections, by polyparasitism (P>0. 05) (S4E Fig), by helminth species exhibiting lung migration capacity (S4F Fig), or by past exposure to helminths (S4G and S4H Fig). Taken together, our data indicate that a high proportion of Mtb-specific CD4 T cells from Tanzanian TB patients have a mixed Th1/Th2 cytokine profile which is observed either in patients with active helminth infection or in a large proportion of patients with positive helminth serology. In contrast, Mtb-specific CD4 T cells from South African TB cases have a classical, previously described Th1 cytokine profile. We then determined whether the development of Mtb-specific Th2 CD4 T cells was associated with changes in the expression of Th1 and Th2 cell lineage transcription factors, T-bet and Gata-3 respectively [48,49]. Representative examples and cumulative data showed that the percentage of memory CD4 T cells expressing Gata-3 was significantly increased in TB and Mtb/helminth co-infected patients from TZ (7% and 7. 9%, respectively) compared with TB patients from SA (2. 4%; P<0. 05) (Fig 2A and 2B). In contrast, the percentage of memory CD4 T cells expressing high levels of T-bet (T-bethigh) was significantly lower in TB and Mtb/helminth patients from TZ (3. 6% and 3. 2%, respectively) as compared to TB patients from SA (12. 3%; P<0. 05) (Fig 2A and 2C). However, the frequencies of memory CD4 T cells expressing Gata-3 or T-bethigh did not differ between TB patients and Mtb/helminth co-infected patients from TZ (P>0. 05) (Fig 2B and 2C). In addition, the frequency of memory CD4 T cells expressing Gata-3 of Mtb/helminth co-infected patients from TZ was not influenced by the helminth species and by polyparasitism (P>0. 05) (S4I Fig), by helminth species exhibiting lung migration capacity (S4J Fig), or by past exposure to helminths (S4K and S4L Fig). Interestingly, the percentage of T-bethigh memory CD4 T cells negatively correlated with the percentage of memory CD4 T cells expressing Gata-3 (r = -0. 6745; P<0. 0001) (S5 Fig) thus supporting previous observations [50]. We then determined whether the expression of T-bet or Gata-3 by memory CD4 T cells was associated with Mtb-specific CD4 T-cell cytokine profile of TB patients and Mtb/helminth co-infected patients from SA or TZ. To address this issue, we plotted the percentage of Mtb-specific CD4 T cells producing IFN-γ or IL-4/5/13 against the percentage memory CD4 T cells expressing T-bethigh or Gata-3 from the same patients (Fig 2D and 2E). The cumulative data showed that the percentage of IFN-γ-producing Mtb-specific CD4 T cells directly correlated with the percentage of T-bethigh memory CD4 T cells (r = 0. 3802, P = 0. 0085) (Fig 2D) and the percentage of IL-4/5/13-producing Mtb-specific CD4 T cells directly correlated with the percentage of memory CD4 T cells expressing Gata-3 (r = 0. 3782, P<0. 0088) (Fig 2E). Taken together, these data indicate that TB patients from TZ have a mixed Th1/Th2 cytokine profile associated with increased Gata-3 and reduced T-bethigh expression. Ongoing helminth infections has been shown to interfere with CD8 T cells responses targeting viruses in mouse models [51,52]. Our group and others has recently shown that Mtb-specific CD8 T cells were more frequently detected in patients with TB disease as compared to those with latent Mtb infection [53]. Based on this observation, a recent diagnostic test i. e. QuantiFERON TB PLUS proposes optimized CD8-TB-specific-peptides stimulation [54,55]. This prompted us to investigate whether ongoing helminth infections would influence the proportion of TB patients with detectable Mtb-specific CD8 T cells. To address this issue, the ability of Mtb-specific CD8 T cells to produce IFN-γ, TNF-α, IL-2 and perforin was assessed in 16 TB patients and 23 Mtb/helminth co-infected patients from TZ by flow cytometry. As shown in Fig 3A, the proportion of subjects with detectable Mtb-specific CD8 T cells was significantly reduced in Mtb/helminth co-infected patients from TZ as compared to TB patients from TZ (43. 7% versus 80%, respectively; P<0. 05; Fig 3A). The frequency of cytokine- and perforin-producing Mtb-specific CD8 T cells and the functional profile of Mtb-specific CD8 T-cell responses did not differ significantly between TB and Mtb/helminth co-infected patients from TZ with detectable Mtb-specific CD8 T cells (Fig 3B) and the predominant CD8 T-cell population was IFN-γ+IL-2-TNF-α+Perforin- (Fig 3C). In summary, these data indicate that ongoing helminth infection reduced the proportion of TB cases with detectable Mtb-specific CD8 T cells. One of the objectives of the present study was to determine whether ongoing helminth infection may influence the levels of systemic inflammation markers. In order to address this issue, we first assessed the serum levels of IL-1α, IL-6, TNF-α, IL-10, IL-12p70, IFN-α2, IFN-β, IFN-ω, IFN-γ, IL-23 and CRP of TB patients from TZ and SA by multiplex bead array analyses (Fig 4). The cumulative data indicated that the serum levels of IL-1α, TNF-α, IL-12p70, IFN-α2, IFN-β, IFN-ω, IL-23 and CRP were significantly increased in TB patients from SA as compared to TB patients from TZ (P<0. 05) (Fig 4). However, the levels of IFN-γ, IL-6 and IL-10 were not statistically different between TB patients from SA and TZ (P>0. 05; Fig 4). Interestingly, ongoing helminth infection did not further influence the serum levels of cytokines and CRP of TB patients from TZ (P>0. 05) (Fig 4). Taken together, these data suggest that TB patients from SA and TZ showed differences in serum cytokine profile. In comparison to Tanzanian TB cases, South African patients showed a more pronounced pro-inflammatory serum cytokine profile associated with high levels of type I IFNs. To better identify the immunological parameters associated with Mtb-specific immune signature in TB patients and the influence of ongoing helminth infection, a principal component analysis (PCA) was performed. The results indicated that the Mtb-specific immune profile of TB and Mtb/helminth co-infected patients from TZ clustered away from that of TB patients from SA with a percentage of discrimination reaching about 70% (Fig 5A and 5B). In depth analysis revealed that the differentially expressed immune parameters contributing the most to discriminate Mtb-specific immune response of TB patients from TZ and SA were the percentages of CD4 T cells expressing Gata-3 and T-bethigh, the proportion of polyfunctional TNF-α/IFN-γ/IL-2, dual TNF-α/IFN-γ and single IFN-γ Mtb-specific CD4 T cells among total Mtb-specific CD4 T-cell responses, levels of IL-5 in Mtb-stimulated culture supernatants and IFN-β serum levels (Fig 5A and 5B and Table 2). Again, this analysis did not allow discriminating Mtb-specific immune response of TB and Mtb/helminth patients from TZ (Fig 5A and 5B). We next performed multiparametric statistical analysis to investigate the potential associations between the four major represented Mtb-specific CD4 T-cell populations i. e. triple IFN-γ/IL-2/TNF-α, dual IFN-γ/TNF-α, single IFN-γ and single IL-4/5/13 Mtb-specific CD4 T cells, the levels of cytokine detected in Mtb-stimulated culture supernatants, i. e. IL-2, IL-4, IL-5 and IL-13, the percentages of memory CD4 T cells expressing Gata-3 or T-bethigh, and the serum levels of IL-1α, TNF-α, IL-12p70, IFN-β and IL-23 in TB patients from SA and TZ (Fig 5C). The combined data indicated that IFN-β serum concentrations positively correlated with i) T-bethigh memory CD4 T cells, ii) higher proportion of Mtb-specific Th1 cells (single IFN-γ and dual IFN-γ/TNF-α), iii) higher serum levels of pro-inflammatory cytokines (IL-1α and TNF-α), IL-12p70 and IL-23 and negatively correlated with a) Gata-3 expression on memory CD4 T cells, b) higher proportion of polyfunctional Mtb-specific CD4 T cells and c) type 2 cytokine secretion (IL-5 and IL-13) (P<0. 05) (Fig 5C). Taken together, the data indicate that the serum cytokine profile and the Mtb-specific immune signatures of TB patients from SA and TZ are significantly different and that Th1 cytokine profiles are positively associated with TB-induced systemic inflammation and higher serum levels of type I IFNs. We then assessed the influence of ongoing helminth infection on TB drug treatment efficiency in the Tanzanian cohort. To address this issue, we first assessed the presence of Mtb in the sputum of treated individuals based on sputum smear microscopy. The cumulative data showed a significant reduction in the proportion of patients with detectable Mtb in the sputum following 60 days of drug treatment (P<0. 05; Fig 6A). The presence of an ongoing helminth infection did not influence the effect of anti-mycobacterial drug treatment in relation to sputum detectable Mtb (P>0. 05) (Fig 6A), suggesting that ongoing helminth infection was not associated with reduced anti-mycobacterial treatment efficiency. In addition, 60 days of anti-mycobacterial treatment was associated with a significant increase of the body mass index (BMI) and significant decrease of serum levels of CRP and pro-inflammatory cytokines i. e. IL-1α, IL-6, and IL-1β (P<0. 05; Fig 6B–6D), irrespective of the presence or absence of ongoing helminth infection (P>0. 05) (Fig 6B–6D). We next assessed the impact of TB treatment initiation on Mtb-specific T-cell immunity in TB and Mtb/helminth patients. The functional profiles of Mtb-specific CD4 T-cell responses and the transcription factor expression profiles were assessed using aforementioned experimental strategies. The cumulative data indicated that the initiation of efficient TB treatment did not significantly influence 1) the Mtb-specific CD4 T-cell functional profiles (P>0. 05) (Fig 6E), 2) the nature and the amount of Th2 cytokines produced i. e. IL-4 and/or IL-5 and/or IL-13 (P>0. 05) (S6A Fig) and 3) the transcription factor profile of TB and Mtb/helminth infected patients from TZ (P>0. 05) (S6B Fig). Indeed, two months after TB treatment, the cytokine profile of Mtb-specific CD4 T cells was still dominated by polyfunctional IFN-γ/IL-2/TNF-α and Th2 Mtb-specific CD4 T cells (Fig 6E). In summary, two months of TB treatment did not change the functional profile of Mtb-specific CD4 T cells. These data provide evidence that TB treatment outcome during this follow up period was not influenced by presence or absence of ongoing helminth infection.
Helminth infections have been shown to impact the control of virus replication in mouse models [52,56] and interfere with vaccine induced immune responses [57,58]. Since helminths and Mtb are co-endemic in several regions of the world, including Tanzania [1], we hypothesized that ongoing helminth infection may influence and potentially modulate the functional profile of Mtb-specific T-cell responses. Our hypothesis is founded on the observations that helminths stimulate Th2 type immune responses and previous studies that demonstrated that the cytokine micro-environment may influence, at least in vitro, the functional profile of antigen-specific CD4 T cell responses [59,60]. In the present study, we provide evidence that a significant proportion of Mtb-specific CD4 T cells in patients with active TB disease from TZ have a Th2 cytokine profile as indicated by the production of IL-4/IL-5/IL-13 and expression of the Th2 cell lineage transcription factor Gata-3. Mtb-specific CD4 T-cells comprised Th1 and Th2 cells with a polyfunctional cytokine profile, and the increased frequency of Gata-3+ memory CD4 T cells was associated with reduced frequency of T-bethigh+ memory CD4 T cells. Interestingly, Mtb-specific CD4 T cells in patients with active TB disease from SA had a typical Th1 profile encompassing single IFN-γ and dual IFN-γ/TNF-α cells. Taken together, these results demonstrate that active TB disease induced the generation of mixed Th1/Th2 Mtb-specific CD4 T cells in patients from TZ whereas a typical Th1 Mtb-specific CD4 T cell response was generated in patients from SA. Further investigations would be needed to determine whether TB-uninfected individuals from TZ and SA harbor different basic response to TB antigens. These observations are of high interest since the assessment of Mtb-specific CD4 T-cell cytokine profile is consistently proposed to allow the discrimination between active and latent Mtb infections [61–64]. Among, these parameters, high proportions of IL-2-producing Mtb-specific CD4 T cells (in association with Th1 cytokines i. e IFN-γ and/or TNF-α) are associated with individuals with LTBI and therefore Mtb containment, while high proportion of Mtb-specific CD4 T cells producing Th1 cytokines in absence of IL-2 is associated with patients suffering from active TB disease [62–65]. Indeed, reduced capacity to produce IL-2 is usually associated with high antigen load, persistent T-cell stimulation, expression of co-inhibitory molecules and differentiation towards effector memory and/or terminally differentiated effector memory [66,67]. Interestingly, CD4 T cells coexpressing IFN-γ and TNF-α and harboring a phenotype of effector-memory response were associated with active tuberculosis in HIV-uninfected [68,69] and HIV-infected TB patients [67]. Of note, polyfunctional helminth-specific CD4 T-cell responses were also recently associated to helminth biological activity [47,70]. As mentioned above, the study hypothesis was that helminth infection might influence the functional profile of Mtb-specific CD4 T cells. In this context, about 67% of patients from TZ either had an ongoing active helminth infection or had evidence of previous helminth exposure/infection. However, about 11% of the patients from SA also showed a positive serology for previous helminth exposure/infection with helminth but no sign of generation of Mtb-specific Th2 CD4 T cells. Despite this discordance between positive serology for helminth exposure and lack of the generation of Mtb-specific Th2 CD4 T cells in patients from SA, we cannot exclude that the frequency and/or severity of helminth re-infection/exposure in patients from TZ is higher as when compared to South African patients and therefore responsible for the mixed Th1/Th2 functional profile of Mtb-specific CD4 T cells [71,72]. In addition, we cannot exclude that human or Mtb genetic diversity and other yet unknown environmental factors may contribute to the generation of mixed Th1/Th2 functional cytokine profile and/or influence the durability of the Th2 response after parasite infection is cleared [73] in patients from TZ. Consistent with previous studies [53,74], about 80% of TB patients from TZ had Mtb-specific CD8 T cell responses detected by polychromatic flow cytometry. Interestingly, this proportion was significantly reduced (about two fold) in the presence of ongoing helminth infections. These results provide additional evidence that helminth infections may significantly influence the generation of distinct T cell subsets in active TB disease. This important finding echoes with recent data demonstrating that ongoing helminth infection impairs virus-specific T-cell immunity via a STAT-6-dependent alternative activation of macrophages differentiation [52]. In contrast to Tanzanian TB cases, the South African TB patients were characterized by significantly higher serum levels of pro-inflammatory cytokines (IL-1α and TNF-α) in combination with IL-12p70 and type I IFNs. Interestingly, statistical analyses revealed that elevated serum levels of IFN-β were associated with elevated serum levels of pro-inflammatory cytokines and Th1 Mtb-specific CD4 T cells lacking IL-2 co-production. These IFN-β serum levels were inversely correlated with polyfunctional Mtb-specific CD4 T cells and Th2 cytokines detected by polychromatic flow cytometry and in cell culture supernatants. Based on these findings, the so far unappreciated role of IFN-β (and type I IFNs in general) in coordinating TB specific immunity needs to be further explored. The role of Th1, Th2 and CD8 T cells in the control of Mtb infection and the progression of TB disease is under intense debate. It has been clearly demonstrated that functional impairment of the IL-12p70/IFN-γ axis predisposes to the development of mycobacterial disease [7–10], probably by compromising the phagocytic and cytolytic capacity of macrophages primed with IFN-γ [75]. The role of Mtb-specific CD8 T cells is however still controversial. Some studies indicate that Mtb-specific CD8 T cells may play an important role in protective immunity against TB via the production of perforin and/or cytolysin [76–80], while others indicate that the presence of Mtb-specific CD8 T cells may be detrimental [81], since Mtb-specific CD8 T cells were enriched in TB patients as compared to individuals with LTBI [53]. Excessive IFN-γ production by CD4 or CD8 T cells may in fact favour Mtb transmission via inflammation mediated mucosal damage enabling access of Mtb bacilli to airways [81]. The dual role of IFN-γ in the protection or progression of TB disease may be linked to macrophage hypo-responsiveness to IFN-γ [82], also called progressive exhaustion, which can be mediated by Mtb-induced type I IFNs [83,84]. Indeed, recent studies demonstrated that Mtb-induced type I IFN might be detrimental, since TB patients with reduced/absent type I IFN signature had reduced bacterial load and/or improved host survival [85–87]. Interestingly, other studies indicate that type I IFN responses enhance CD4 T-cell differentiation towards Th1 [88], enhance CD8 T-cell responses [89] and interfere with IL-23-mediated Th17 cell differentiation and IL-4-mediated Th2-cell differentiation by inhibiting Gata-3 expression [88,90,91], supporting our observations. The reason why TB patients from TZ had lower levels of type I IFNs remains unclear and needs to be further investigated. However, one could postulate that the genetic background of the Mtb strains isolated from TZ and SA, the antigen load and the genetic background of the individuals living in TZ versus SA might be associated with this profile, since the level of Mtb-induced type I IFN production might be strain dependent [92]. Of note, these parameters were not evaluated in the present study and would require further evaluation. Finally, we did not observe any influence of ongoing helminth infection on the efficacy of TB therapy. Interestingly, after sixty days of treatment, we did not observe changes in Mtb-specific CD4 T-cell cytokine and transcription factor expression profile but strongly reduced CRP serum levels and pro-inflammatory cytokine circulation in combination with improved BMI. In conclusion, we provide evidence that the generation of Mtb-specific CD4 and CD8 T cell responses, may be substantially influenced by co-infectious agents and possibly genetic and environmental factors resulting in pronounced variations in the qualitative and quantitative profile of pathogen-specific responding T cells in human populations.
In total, 72 subjects were recruited to participate in this study. No statistical method was used to predetermine sample size. Fifty-five subjects were recruited at the Mwananyamala Hospital, Dar es Salaam, and the TB clinic of Bagamoyo (TZ). TB patients (n = 55) were selected based on sputum smear microscopy confirmed by GeneXpert assay and HIV infection was ruled out by rapid serological test (Alere Determine HIV-1/2 test). The diagnosis of ongoing and/or past helminth infection was based on assays performed on feces (Kato-Katz thick smear, FLOTAC and Baermann assays), urine (urine filtration), whole blood (Immuno-chromatography) and serum (ELISA) samples at date of blood sample collection and the assay used depended on the helminth species and on the site of sample collection. All TB patients from TZ were screened for active soil-transmitted helminths (hookworms, Ascaris lumbricoides, Trichuris trichiura) and S. mansoni infections using the Kato-Katz thick smear and FLOTAC methods performed on one stool sample at date of blood sample collection, for active S. stercoralis infection using the Baermann technique and for active S. haematobium infection using the urine filtration method. Binax NOW ICT test card were used to detect W. Bancrofti antigen on blood sample [93,94]. In addition, serology for 7 different helminths (Echinococcus spp, Fasciola hepatica, Filaria, Schistosoma spp, S. stercoralis, Toxocara spp and Trichinella spp) was performed by ELISA. The serodiagnostic helminth screening ELISA is routinely performed at the diagnostic centre of the Swiss Tropical and Public Health Institute and detects helminth specific IgG. A total of 30 Mtb/helminth co-infected patients were recruited. TB patient co-infected with only one helminth species were infected with S. mansoni (n = 7), W. bancrofti (n = 6), hookworms (n = 2), S. haematobium (n = 2) or S. stercoralis (n = 2). Eleven TB patients (36%) were co-infected with multiple helminth species. Blood samples were collected prior to and following 60 days of anti-mycobacterial treatment (“fixed dose combination” consisting of Rifampicin, Isoniazid, Pyrazinamide and Ethambutol, (RHZE) ) from 20 TB and 24 Mtb/helminth co-infected patients recruited in TZ. In addition, seventeen subjects were enrolled at the field site of the South African Tuberculosis Vaccine Initiative in the Boland Overberg region of the Western Cape Province of SA (SA). TB disease was diagnosed by positive sputum Xpert MTB/RIF and HIV infection was ruled out by rapid serological test. PBMCs were collected as part of a cross-sectional study, in HIV-negative participants before commencing treatment for TB. Diagnosis of helminth exposure was performed using ELISA detecting helminth-specific IgG (Echinococcus spp, Fasciola hepatica, Filaria, Schistosoma spp, S. stercoralis, Toxocara spp and Trichinella spp). All participants were adults and provided written informed consent and the study protocol was approved for TZ by the Ethikkomission beider Basel (EKBB; Basel, Switzerland; reference number 257/08), the Ifakara Health Institute Institutional Review Board and the National Institute for Medical Research (NIMR; Dar es Salaam, United Republic of Tanzania; reference number NIMR/HQ/R. 8a/Vol. IX/1098). For SA, the Human Research Ethics Committee of the University of Cape Town granted the study protocol approval. The following monoclonal antibodies (mAbs) were used in different combinations. CD3-APC-H7 (CloneSK7), CD4-PECF594 or CD4-APC (Clone RPA-T4), CD8-PB (Clone RPA-T8), IFN-γ-AF700 or IFN-γ-APC (Clone B27), TNF-α-PeCy-7 (Clone MAb11), IL-4-PE (Clone 3010. 211), IL-2-PE (Clone MQ1-17H12), Gata-3-PeCy-7 (Clone L50-823), all from Becton Dickinson (BD); CD45RA-BV711 (Clone HI100), IL-2-PerCpCy5. 5 (Clone MQ1-17H12), IL-5-PE (Clone TRFK5), IL-13-PE (Clone JES10-5A2), T-bet-PerCpCy5. 5 (Clone 4B10) were purchased from BioLegend; CD8-Efluor625NC (Clone RPA-T8) from eBioscience; perforin-FITC (Clone B-D48) from Diaclone. Mtb-derived CFP-10 and ESAT-6 peptide pools are composed of 15-mers overlapping by 11 amino-acids encompassing the entire sequences of the proteins and all peptides were HPLC purified (>90% purity). PBMCs were stimulated overnight in complete media (RPMI (Invitrogen), 10% fetal calf serum (FCS; Invitrogen), 100 µg/ml penicillin, 100 unit/ml streptomycin (BioConcept) ) with ESAT-6 and CFP-10 peptide pools (1 µg/ml) or with Staphyloccocus enterotoxin B (SEB; 250 ng/mL) or unstimulated in the presence of Golgiplug (1 μl/ml; BD) as previously described [95]. At the end of the stimulation period, cells were washed and stained (20 min; 4°C) for dead cells using the Aqua LIVE/DEAD stain kit (Invitrogen), washed and stained (20 min; 4°C) with mAbs to CD3, CD4, CD8 and CD45RA. Cells were then permeabilized (30 min; 20°C) (Cytofix/Cytoperm, BD) and stained (20 min; 20°C) with mAbs to TNF-α, IFN-γ, IL-2, IL-4, IL-5 and IL-13. PBMCs were stimulated overnight in complete media (RPMI (Invitrogen), 10% fetal calf serum (FCS; Invitrogen), 100 µg/ml penicillin, 100 unit/ml streptomycin (BioConcept) ) with ESAT-6 and CFP-10 peptide pools (1 µg/ml) or Staphyloccocus enterotoxin B (SEB; 250 ng/mL) or unstimulated in the presence of Golgiplug (1 μl/ml; BD). At the end of the stimulation period, cells were washed and stained (20 min; 4°C) for dead cells using the Aqua LIVE/DEAD stain kit, then permeabilized (30 min; 20°C) (Cytofix/Cytoperm, BD) and stained (20 min; 20°C) with mAbs to CD3, CD4, CD8, TNF-α, IFN-γ, IL-2 and perforin. PBMCs were washed, stained (20 min; 4°C) for dead cells using the Aqua LIVE/DEAD stain kit, then washed and stained (20 min; 4°C) for CD3, CD4, CD8, CD45RA. Cells were then washed, permeabilized (45 min; 4°C) (Foxp3 Fixation/Permeabilization Kit; eBioscience) and stained (20 min; 4°C) with mAbs to T-bet and Gata-3. PBMCs (2x105 cells) were stimulated for 24 hours in complete media (RPMI (Invitrogen), 10% fetal calf serum (FCS; Invitrogen), 100 µg/ml penicillin, 100 unit/ml streptomycin (BioConcept) ) with ESAT-6 and CFP-10 peptide pools (1 µg/ml) or with Staphyloccocus enterotoxin B (SEB; 250 ng/mL) or left unstimulated (negative control). At the end of the stimulation period, culture supernatants were collected and levels of TNF-α, IFN-γ, IL-2, IL-4, IL-5, IL-13, IL-10, IL-17A and IL-17F were assessed cells by luminex assay (ProcartaPlex Mix&Match Human plex, eBioscience). Serum levels of IL-1α, IL-6, TNF-α, IL-12p70, IL-23, IL-10, IFN-α2, IFN-β, IFN-ω and IFN-γ was assessed by luminex assay (ProcartaPlex Mix&Match Human plex, eBioscience) and CRP was assessed by nephelemetry (CardioPhasehsCRP, Siemens Healthcare Diagnostics Products GmbH) as previously described [96]. Cells were fixed with CellFix (BD), acquired on an LSRII SORP (4 lasers: 405,488,532 and 633 nm) and analyzed using FlowJo (version 9. 7. 7) (Tree star Inc, Ashland, OR, USA). Frequencies of cytokine-producing Mtb-specific T cells and cytokine profile of Mtb-specific T-cell responses were analyzed using SPICE software (version 5. 34) following background subtraction. When required, analysis and presentation of distributions was performed using SPICE, downloaded from <http: //exon. niaid. nih. gov/spice> [97]. Statistical significance (P values) was obtained either using two-tailed Chi-square analysis for comparison of positive proportions or using one-way ANOVA (Kruskal-Wallis test) followed by Mann–Whitney test or Wilcoxon Matched-pairs two-tailed Signed Rank test for multiple comparisons or Spearman rank test for correlations using GraphPad Prism version 7 (San Diego, CA). Statistical analyses of global cytokine profiles (pie charts) were performed by partial permutation tests using the SPICE software as described [97]. Principal component analysis (PCA) was performed using the R ‘‘stats” package. To normalize data distribution, the values of each parameter were first log transformed. Data were then filtered using Lasso method [98]. | Mycobacterium tuberculosis (Mtb) and helminth infections are co-endemic in several regions of the world and their immune responses may be mutually antagonistic. We therefore hypothesized that helminth infection would impact and potentially shape Mtb-specific T-cell responses and systemic inflammation in patients suffering from active pulmonary tuberculosis (TB) enrolled from two helminth endemic regions i. e. Tanzania (TZ) and South Africa (SA). In this study, we demonstrate for the first time that TB patients from SA and TZ harbor distinct immune responses to Mtb antigens. Indeed, we showed that Mtb-specific CD4 T-cell responses of TB patients from TZ were composed by a mixed T helper type 1 (Th1) and Th2 responses. In contrast, the cytokine profile of Mtb-specific CD4 T cells of TB patients from SA was dominated by Th1 cells and associated with TB-induced systemic inflammation and elevated serum levels of type I IFN. Taken together, these data indicate that Mtb-specific T-cell responses are diverse in human populations and can be strongly influenced by host and pathogen genetic background, co-infections and yet unknown environmental factors. Identification of correlates of risk and protection from TB disease will help in the rational development of protective T-cell based vaccines against TB, early monitoring TB treatment outcomes and focused follow up of high risk populations. | Abstract
Introduction
Results
Discussion
Materials and methods | blood cells
t helper cells
invertebrates
innate immune system
medicine and health sciences
immune cells
immune physiology
cytokines
helminths
immunology
tropical diseases
parasitic diseases
animals
bacterial diseases
developmental biology
molecular development
bacteria
infectious diseases
white blood cells
memory t cells
animal cells
tuberculosis
t cells
actinobacteria
immune system
helminth infections
cell biology
mycobacterium tuberculosis
physiology
biology and life sciences
cellular types
organisms | 2017 | Mixed Th1 and Th2 Mycobacterium tuberculosis-specific CD4 T cell responses in patients with active pulmonary tuberculosis from Tanzania | 11,254 | 335 |
Subsets and Splits